adding CFFI just in case. Need to make into a submodule at somepoint.

[CommonLispStat.git] / external / cffi.darcs / doc / cffi-manual.texinfo

\input texinfo   @c -*- Mode: Texinfo; Mode: auto-fill -*-
@c %**start of header
@setfilename cffi.info
@settitle CFFI User Manual
@exampleindent 2

@c @documentencoding utf-8

@ignore
Style notes:

* The reference section names and "See Also" list are roman, not
  @code.  This is to follow the format of CLHS.

* How it looks in HTML is the priority.
@end ignore

@c ============================= Macros =============================
@c The following macros are used throughout this manual.

@macro Function {args}
@defun \args\
@end defun
@end macro

@macro Macro {args}
@defmac \args\
@end defmac
@end macro

@macro Accessor {args}
@deffn {Accessor} \args\
@end deffn
@end macro

@macro GenericFunction {args}
@deffn {Generic Function} \args\
@end deffn
@end macro

@macro ForeignType {args}
@deftp {Foreign Type} \args\
@end deftp
@end macro

@macro Variable {args}
@defvr {Special Variable} \args\
@end defvr
@end macro

@macro Condition {args}
@deftp {Condition Type} \args\
@end deftp
@end macro

@macro cffi
@acronym{CFFI}
@end macro

@macro impnote {text}
@quotation
@strong{Implementor's note:} @emph{\text\}
@end quotation
@end macro

@c Info "requires" that x-refs end in a period or comma, or ) in the
@c case of @pxref.  So the following implements that requirement for
@c the "See also" subheadings that permeate this manual, but only in
@c Info mode.
@ifinfo
@macro seealso {name}
@ref{\name\}.
@end macro
@end ifinfo

@ifnotinfo
@alias seealso = ref
@end ifnotinfo

@c Set ROMANCOMMENTS to get comments in roman font.
@ifset ROMANCOMMENTS
@alias lispcmt = r
@end ifset
@ifclear ROMANCOMMENTS
@alias lispcmt = asis
@end ifclear


@c ============================= Macros =============================


@c Show types, functions, and concepts in the same index.
@syncodeindex tp cp
@syncodeindex fn cp

@copying
Copyright @copyright{} 2005 James Bielman <jamesjb at jamesjb.com> @*
Copyright @copyright{} 2005-2007 Lu@'{@dotless{i}}s Oliveira
  <loliveira at common-lisp.net> @*
Copyright @copyright{} 2006 Stephen Compall <s11 at member.fsf.org>

@quotation
Permission is hereby granted, free of charge, to any person obtaining
a copy of this software and associated documentation files (the
``Software''), to deal in the Software without restriction, including
without limitation the rights to use, copy, modify, merge, publish,
distribute, sublicense, and/or sell copies of the Software, and to
permit persons to whom the Software is furnished to do so, subject to
the following conditions:

The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software.

@sc{The software is provided ``as is'', without warranty of any kind,
express or implied, including but not limited to the warranties of
merchantability, fitness for a particular purpose and noninfringement.
In no event shall the authors or copyright holders be liable for any
claim, damages or other liability, whether in an action of contract,
tort or otherwise, arising from, out of or in connection with the
software or the use or other dealings in the software.}
@end quotation
@end copying
@c %**end of header

@titlepage
@title CFFI User Manual
@c @subtitle Version X.X
@c @author James Bielman

@page
@vskip 0pt plus 1filll
@insertcopying
@end titlepage

@contents

@ifnottex
@node Top
@top cffi
@insertcopying
@end ifnottex

@menu
* Introduction::                What is CFFI?
* Implementation Support::      
* Tutorial::                    Interactive intro to using CFFI.
* Wrapper generators::          CFFI forms from munging C source code.
* Foreign Types::               
* Pointers::                    
* Strings::                     
* Variables::                   
* Functions::                   
* Libraries::                   
* Callbacks::                   
* Limitations::                 
* Platform-specific features::  Details about the underlying system.
* Glossary::                    List of CFFI-specific terms and meanings.
* Comprehensive Index::         

@detailmenu
 --- Dictionary ---

Foreign Types

* convert-from-foreign::        Outside interface to backward type translator.
* convert-to-foreign::          Outside interface to forward type translator.
* defbitfield::                 Defines a bitfield.
* defcstruct::                  Defines a C structure type.
* defcunion::                   Defines a C union type.
* defctype::                    Defines a foreign typedef.
* defcenum::                    Defines a C enumeration.
* define-foreign-type::         Defines a foreign type specifier.
* define-parse-method::         Specifies how a type should be parsed.
@c * explain-foreign-slot-value::  <unimplemented>
* foreign-bitfield-symbols::    Returns a list of symbols for a bitfield type.
* foreign-bitfield-value::      Calculates a value for a bitfield type.
* foreign-enum-keyword::        Finds a keyword in an enum type.
* foreign-enum-value::          Finds a value in an enum type.
* foreign-slot-names::          Returns a list of slot names in a foreign struct.
* foreign-slot-offset::         Returns the offset of a slot in a foreign struct.
* foreign-slot-pointer::        Returns a pointer to a slot in a foreign struct.
* foreign-slot-value::          Returns the value of a slot in a foreign struct.
* foreign-type-alignment::      Returns the alignment of a foreign type.
* foreign-type-size::           Returns the size of a foreign type.
* free-converted-object::       Outside interface to typed object deallocators.
* free-translated-object::      Free a type translated foreign object.
* translate-from-foreign::      Translate a foreign object to a Lisp object.
* translate-to-foreign::        Translate a Lisp object to a foreign object.
* with-foreign-object::         Allocates a foreign object with dynamic extent.
* with-foreign-slots::          Access the slots of a foreign structure.

Pointers

* foreign-free::                Deallocates memory.
* foreign-alloc::               Allocates memory.
* foreign-symbol-pointer::      Returns a pointer to a foreign symbol.
* inc-pointer::                 Increments the address held by a pointer.
* incf-pointer::                Increments the pointer address in a place.
* make-pointer::                Returns a pointer to a given address.
* mem-aref::                    Accesses the value of an index in an array.
* mem-ref::                     Dereferences a pointer.
* null-pointer::                Returns a NULL pointer.
* null-pointer-p::              Tests a pointer for NULL value.
* pointerp::                    Tests whether an object is a pointer or not.
* pointer-address::             Returns the address pointed to by a pointer.
* pointer-eq::                  Tests if two pointers point to the same address.
* with-foreign-pointer::        Allocates memory with dynamic extent.

Strings

* foreign-string-alloc::        Converts a Lisp string to a foreign string.
* foreign-string-free::         Deallocates memory used by a foreign string.
* foreign-string-to-lisp::      Converts a foreign string to a Lisp string.
* lisp-string-to-foreign::      Copies a Lisp string into a foreign string.
* with-foreign-string::         Allocates a foreign string with dynamic extent.
* with-foreign-pointer-as-string::  Similar to CL's with-output-to-string.

Variables

* defcvar::                     Defines a C global variable.
* get-var-pointer::             Returns a pointer to a defined global variable.

Functions

* defcfun::                     Defines a foreign function.
* foreign-funcall::             Performs a call to a foreign function.
* foreign-funcall-pointer::     Performs a call through a foreign pointer.

Libraries

* close-foreign-library::       Closes a foreign library.
* *darwin-framework-directories*::  Search path for Darwin frameworks.
* define-foreign-library::      Explain how to load a foreign library.
* *foreign-library-directories*::  Search path for shared libraries.
* load-foreign-library::        Load a foreign library.
* load-foreign-library-error::  Signalled on failure of its namesake.
* use-foreign-library::         Load a foreign library when needed.

Callbacks

* callback::                    Returns a pointer to a defined callback.
* defcallback::                 Defines a Lisp callback.
* get-callback::                Returns a pointer to a defined callback.

@end detailmenu
@end menu




@c ===================================================================
@c CHAPTER: Introduction

@node Introduction
@chapter Introduction

@cffi{} is the Common Foreign Function Interface for @acronym{ANSI}
Common Lisp systems.  By @dfn{foreign function} we mean a function
written in another programming language and having different data and
calling conventions than Common Lisp, namely, C.  @cffi{} allows you
to call foreign functions and access foreign variables, all without
leaving the Lisp image.

We consider this manual ever a work in progress.  If you have
difficulty with anything @cffi{}-specific presented in the manual,
please contact @email{cffi-devel@@common-lisp.net,the developers} with
details.


@heading Motivation

@xref{Tutorial-Comparison,, What makes Lisp different}, for
an argument in favor of @acronym{FFI} in general.

@cffi{}'s primary role in any image is to mediate between Lisp
developers and the widely varying @acronym{FFI}s present in the
various Lisp implementations it supports.  With @cffi{}, you can
define foreign function interfaces while still maintaining portability
between implementations.  It is not the first Common Lisp package with
this objective; however, it is meant to be a more malleable framework
than similar packages.


@heading Design Philosophy

@itemize
@item
Pointers do not carry around type information. Instead, type
information is supplied when pointers are dereferenced.

@item
A type safe pointer interface can be developed on top of an
untyped one.  It is difficult to do the opposite.

@item
Functions are better than macros.  When a macro could be used
for performance, use a compiler-macro instead.
@end itemize


@c ===================================================================
@c CHAPTER: Implementation Support

@node Implementation Support
@chapter Implementation Support

@cffi{} supports various free and commercial Lisp implementations:
Allegro CL, Corman CL, @sc{clisp}, @acronym{CMUCL}, @acronym{ECL},
LispWorks, Open@acronym{MCL}, @acronym{SBCL} and the Scieneer CL.

There are also plans to support Digitool @acronym{MCL}, and @acronym{GCL}.


@section Allegro CL

@strong{Tested platforms:} linux/x86, linux/ppc, win32/x86, darwin/ppc.

Version 7.0 is supported. The 8.0 beta is also known to work. Earlier
versions are untested and unsupported but patches to support them
are welcome.

@subheading Limitations

@itemize
@item
Does not support the @code{:long-long} type.
@end itemize

@section Corman CL

@strong{Tested platforms:} win32/x86.

Versions prior to 2.51 are untested and unsupported. Also, you will
need to avoid Corman's buggy @code{COMPILE-FILE} and fasl
loader. Please follow @uref{http://www.weitz.de/corman-asdf/, these
instructions} by Edi Weitz to setup ASDF for Corman CL in a way that
works around these issues.

@subheading Limitations

@itemize
@item
Does not support @code{foreign-funcall}.
@end itemize


@section @sc{clisp}

@strong{Tested platforms:} linux/x86, linux/ppc, win32/x86, darwin/ppc.

Version is 2.34 or newer is required on win32/x86. For other platforms
version 2.35 or newer is required.


@section @acronym{CMUCL}

@strong{Tested platforms:} linux/x86, darwin/ppc.

Versions prior to 19B are untested. For darwin/ppc, the 2006-02 (19C)
snapshot or later is recommended.


@section @acronym{ECL}

@strong{Tested platforms:} @emph{needs testing...}

As of November 2005, the CVS version of ECL is required. It is
reported to pass all tests.

@subheading Limitations
@itemize
@item
Does not support the @code{:long-long} type.

@item
On platforms where ECL's dynamic FFI is not supported (ie. when
@code{:dffi} is not present in @code{*features*}),
@code{cffi:load-foreign-library} does not work and you must use ECL's
own @code{ffi:load-foreign-library} with a constant string argument.
@end itemize


@section Lispworks

@strong{Tested platforms:} linux/x86, win32/x86, darwin/ppc.

Versions prior to 4.4 are untested.

@subheading Limitations
@itemize
@item
Does not support the @code{:long-long} type.
@end itemize


@section Open@acronym{MCL}

@strong{Tested platforms:} darwin/ppc, linux/ppc.

Open@acronym{MCL} 1.0 or newer is recommended.


@section @acronym{SBCL}

@strong{Tested platforms:} linux/x86, linux/ppc, darwin/ppc.

Version 0.9.6 or newer is recommended.

@subheading Limitations

@itemize
@item
Not all platforms support callbacks.
@end itemize


@section Scieneer CL

@strong{Tested platforms:} linux/x86, linux/amd64.

Version 1.2.10 or newer is recommended.  Passes all tests.
The x86 and AMD64 ports feature long-double support.


@c ===================================================================
@c CHAPTER: An Introduction to Foreign Interfaces and CFFI

@c This macro is merely a marker that I don't think I'll use after
@c all.
@macro tutorialsource {text}
@c \text\
@end macro

@c because I don't want to type this over and over
@macro clikicffi
http://www.cliki.net/CFFI
@end macro
@c TeX puts spurious newlines in when you use the above macro
@c in @examples &c.  So it is expanded below in some places.


@node Tutorial
@chapter An Introduction to Foreign Interfaces and @acronym{CFFI}

@c Above, I don't use the cffi macro because it breaks TeX.

@cindex tutorial, @cffi{}
Users of many popular languages bearing semantic similarity to Lisp,
such as Perl and Python, are accustomed to having access to popular C
libraries, such as @acronym{GTK}, by way of ``bindings''.  In Lisp, we
do something similar, but take a fundamentally different approach.
This tutorial first explains this difference, then explains how you
can use @cffi{}, a powerful system for calling out to C and C++ and
access C data from many Common Lisp implementations.

@cindex foreign functions and data
The concept can be generalized to other languages; at the time of
writing, only @cffi{}'s C support is fairly complete, but C++
support is being worked on.  Therefore, we will interchangeably refer
to @dfn{foreign functions} and @dfn{foreign data}, and ``C functions''
and ``C data''.  At no time will the word ``foreign'' carry its usual,
non-programming meaning.

This tutorial expects you to have a working understanding of both
Common Lisp and C, including the Common Lisp macro system.

@menu
* Tutorial-Comparison::         Why FFI?
* Tutorial-Getting a URL::      An FFI use case.
* Tutorial-Loading::            Load libcurl.so.
* Tutorial-Initializing::       Call a function in libcurl.so.
* Tutorial-easy_setopt::        An advanced libcurl function.
* Tutorial-Abstraction::        Why breaking it is necessary.
* Tutorial-Lisp easy_setopt::   Semi-Lispy option interface.
* Tutorial-Memory::             In C, you collect the garbage.
* Tutorial-Callbacks::          Make useful C function pointers.
* Tutorial-Completion::         Minimal get-url functionality.
* Tutorial-Types::              Defining new foreign types.
* Tutorial-Conclusion::         What's next?
@end menu


@node Tutorial-Comparison
@section What makes Lisp different

The following sums up how bindings to foreign libraries are usually
implemented in other languages, then in Common Lisp:

@table @asis
@item Perl, Python, Java, other one-implementation languages
@cindex @acronym{SWIG}
@cindex Perl
@cindex Python
Bindings are implemented as shared objects written in C.  In some
cases, the C code is generated by a tool, such as @acronym{SWIG}, but
the result is the same: a new C library that manually translates
between the language implementation's objects, such as @code{PyObject}
in Python, and whatever C object is called for, often using C
functions provided by the implementation.  It also translates between
the calling conventions of the language and C.

@item Common Lisp
@cindex @acronym{SLIME}
Bindings are written in Lisp.  They can be created at-will by Lisp
programs.  Lisp programmers can write new bindings and add them to the
image, using a listener such as @acronym{SLIME}, as easily as with
regular Lisp definitions.  The only foreign library to load is the one
being wrapped---the one with the pure C interface; no C or other
non-Lisp compilation is required.
@end table

@cindex advantages of @acronym{FFI}
@cindex benefits of @acronym{FFI}
We believe the advantages of the Common Lisp approach far outweigh any
disadvantages.  Incremental development with a listener can be as
productive for C binding development as it is with other Lisp
development.  Keeping it ``in the [Lisp] family'', as it were, makes
it much easier for you and other Lisp programmers to load and use the
bindings.  Common Lisp implementations such as @acronym{CMUCL}, freed
from having to provide a C interface to their own objects, are thus
freed to be implemented in another language (as @acronym{CMUCL} is)
while still allowing programmers to call foreign functions.

@cindex minimal bindings
Perhaps the greatest advantage is that using an @acronym{FFI} doesn't
obligate you to become a professional binding developer.  Writers of
bindings for other languages usually end up maintaining or failing to
maintain complete bindings to the foreign library.  Using an
@acronym{FFI}, however, means if you only need one or two functions,
you can write bindings for only those functions, and be assured that
you can just as easily add to the bindings if need be.

@cindex C abstractions
@cindex abstractions in C
The removal of the C compiler, or C interpretation of any kind,
creates the main disadvantage: some of C's ``abstractions'' are not
available, violating information encapsulation.  For example,
@code{struct}s that must be passed on the stack, or used as return
values, without corresponding functional abstractions to create and
manage the @code{struct}s, must be declared explicitly in Lisp.  This
is fine for structs whose contents are ``public'', but is not so
pleasant when a struct is supposed to be ``opaque'' by convention,
even though it is not so defined.@footnote{Admittedly, this is an
advanced issue, and we encourage you to leave this text until you are
more familiar with how @cffi{} works.}

Without an abstraction to create the struct, Lisp needs to be able to
lay out the struct in memory, so must know its internal details.

@cindex workaround for C
In these cases, you can create a minimal C library to provide the
missing abstractions, without destroying all the advantages of the
Common Lisp approach discussed above.  In the case of @code{struct}s,
you can write simple, pure C functions that tell you how many bytes a
struct requires or allocate new structs, read and write fields of the
struct, or whatever operations are supposed to be
public.@footnote{This does not apply to structs whose contents are
intended to be part of the public library interface.  In those cases,
a pure Lisp struct definition is always preferred.  In fact, many
prefer to stay in Lisp and break the encapsulation anyway, placing the
burden of correct library interface definition on the library.}

@impnote{cffi-grovel, a project not yet part of @cffi{}, automates
this and other processes.}

Another disadvantage appears when you would rather use the foreign
language than Lisp.  However, someone who prefers C to Lisp is not a
likely candidate for developing a Lisp interface to a C library.


@node Tutorial-Getting a URL
@section Getting a @acronym{URL}

@cindex c@acronym{URL}
The widely available @code{libcurl} is a library for downloading files
over protocols like @acronym{HTTP}.  We will use @code{libcurl} with
@cffi{} to download a web page.

Please note that there are many other ways to download files from the
web, not least the @sc{cl-curl} project to provide bindings to
@code{libcurl} via a similar @acronym{FFI}.@footnote{Specifically,
@acronym{UFFI}, an older @acronym{FFI} that takes a somewhat different
approach compared to @cffi{}.  I believe that these days (December
2005) @cffi{} is more portable and actively developed, though not as
mature yet.  Consensus in the free @sc{unix} Common Lisp community
seems to be that @cffi{} is preferred for new development, though
@acronym{UFFI} will likely go on for quite some time as many projects
already use it.  @cffi{} includes the @code{UFFI-COMPAT} package for
complete compatibility with @acronym{UFFI}.}

@uref{http://curl.haxx.se/libcurl/c/libcurl-tutorial.html,,libcurl-tutorial(3)}
is a tutorial for @code{libcurl} programming in C.  We will follow
that to develop a binding to download a file.  We will also use
@file{curl.h}, @file{easy.h}, and the @command{man} pages for the
@code{libcurl} function, all available in the @samp{curl-dev} package
or equivalent for your system, or in the c@acronym{URL} source code
package.  If you have the development package, the headers should be
installed in @file{/usr/include/curl/}, and the @command{man} pages
may be accessed through your favorite @command{man} facility.


@node Tutorial-Loading
@section Loading foreign libraries

@cindex loading @cffi{}
@cindex requiring @cffi{}
First of all, we will create a package to work in.  You can save these
forms in a file, or just send them to the listener as they are.  If
creating bindings for an @acronym{ASDF} package of yours, you will
want to add @code{:cffi} to the @code{:depends-on} list in your
@file{.asd} file.  Otherwise, just use the @code{asdf:oos} function to
load @cffi{}.

@tutorialsource{Initialization}
@lisp
(asdf:oos 'asdf:load-op :cffi)

;;; @lispcmt{Nothing special about the "CFFI-USER" package.  We're just}
;;; @lispcmt{using it as a substitute for your own CL package.}
(defpackage :cffi-user
  (:use :common-lisp :cffi))

(in-package :cffi-user)

(define-foreign-library libcurl
  (:unix (:or "libcurl.so.3" "libcurl.so"))
  (t (:default "libcurl")))

(use-foreign-library libcurl)
@end lisp

@cindex foreign library load
@cindex library, foreign
Using @code{define-foreign-library} and @code{use-foreign-library}, we
have loaded @code{libcurl} into Lisp, much as the linker does when you
start a C program, or @code{common-lisp:load} does with a Lisp source
file or @acronym{FASL} file.  We special-cased for @sc{unix} machines
to always load a particular version, the one this tutorial was tested
with; for those who don't care, the @code{define-foreign-library}
clause @code{(t (:default "libcurl"))} should be satisfactory, and
will adapt to various operating systems.


@node Tutorial-Initializing
@section Initializing @code{libcurl}

@cindex function definition
After the introductory matter, the tutorial goes on to present the
first function you should use.

@example
CURLcode curl_global_init(long flags);
@end example

@noindent
Let's pick this apart into appropriate Lisp code:

@tutorialsource{First CURLcode}
@lisp
;;; @lispcmt{A CURLcode is the universal error code.  curl/curl.h says}
;;; @lispcmt{no return code will ever be removed, and new ones will be}
;;; @lispcmt{added to the end.}
(defctype curl-code :int)

;;; @lispcmt{Initialize libcurl with FLAGS.}
(defcfun "curl_global_init" curl-code
  (flags :long))
@end lisp

@impnote{CFFI currently assumes the UNIX viewpoint that there is one C
symbol namespace, containing all symbols in all loaded objects.  This
is not so on Windows and Darwin.  The interface may be changed to deal
with this.}

Note the parallels with the original C declaration.  We've defined
@code{curl-code} as a wrapping type for @code{:int}; right now, it
only marks it as special, but later we will do something more
interesting with it.  The point is that we don't have to do it yet.

@cindex calling foreign functions
Looking at @file{curl.h}, @code{CURL_GLOBAL_NOTHING}, a possible value
for @code{flags} above, is defined as @samp{0}.  So we can now call
the function:

@example
@sc{cffi-user>} (curl-global-init 0)
@result{} 0
@end example

@cindex looks like it worked
Looking at @file{curl.h} again, @code{0} means @code{CURLE_OK}, so it
looks like the call succeeded.  Note that @cffi{} converted the
function name to a Lisp-friendly name.  You can specify your own name
if you want; use @code{("curl_global_init" @var{your-name-here})} as
the @var{name} argument to @code{defcfun}.

The tutorial goes on to have us allocate a handle.  For good measure,
we should also include the deallocator.  Let's look at these
functions:

@example
CURL *curl_easy_init( );
void curl_easy_cleanup(CURL *handle);
@end example

Advanced users may want to define special pointer types; we will
explore this possibility later.  For now, just treat every pointer as
the same:

@tutorialsource{curl_easy handles}
@lisp
(defcfun "curl_easy_init" :pointer)

(defcfun "curl_easy_cleanup" :void
  (easy-handle :pointer))
@end lisp

Now we can continue with the tutorial:

@example
@sc{cffi-user>} (defparameter *easy-handle* (curl-easy-init))
@result{} *EASY-HANDLE*
@sc{cffi-user>} *easy-handle*
@result{} #<FOREIGN-ADDRESS #x09844EE0>
@end example

@cindex pointers in Lisp
Note the print representation of a pointer.  It changes depending on
what Lisp you are using, but that doesn't make any difference to
@cffi{}.


@node Tutorial-easy_setopt
@section Setting download options

The @code{libcurl} tutorial says we'll want to set many options before
performing any download actions.  This is done through
@code{curl_easy_setopt}:

@c That is literally ..., not an ellipsis.
@example
CURLcode curl_easy_setopt(CURL *curl, CURLoption option, ...);
@end example

@cindex varargs
@cindex foreign arguments
We've introduced a new twist: variable arguments.  There is no obvious
translation to the @code{defcfun} form, particularly as there are four
possible argument types.  Because of the way C works, we could define
four wrappers around @code{curl_easy_setopt}, one for each type; in
this case, however, we'll use the general-purpose macro
@code{foreign-funcall} to call this function.

@cindex enumeration, C
To make things easier on ourselves, we'll create an enumeration of the
kinds of options we want to set.  The @code{enum CURLoption} isn't the
most straightforward, but reading the @code{CINIT} C macro definition
should be enlightening.

@tutorialsource{CURLoption enumeration}
@lisp
(defmacro define-curl-options (name type-offsets &rest enum-args)
  "As with CFFI:DEFCENUM, except each of ENUM-ARGS is as follows:

    (NAME TYPE NUMBER)

Where the arguments are as they are with the CINIT macro defined
in curl.h, except NAME is a keyword.

TYPE-OFFSETS is a plist of TYPEs to their integer offsets, as
defined by the CURLOPTTYPE_LONG et al constants in curl.h."
  (flet ((enumerated-value (type offset)
           (+ (getf type-offsets type) offset)))
    `(progn
       (defcenum ,name
         ,@@(loop for (name type number) in enum-args
              collect (list name (enumerated-value type number))))
       ',name)))                ;@lispcmt{for REPL users' sanity}

(define-curl-options curl-option
    (long 0 objectpoint 10000 functionpoint 20000 off-t 30000)
  (:noprogress long 43)
  (:nosignal long 99)
  (:errorbuffer objectpoint 10)
  (:url objectpoint 2))
@end lisp

With some well-placed Emacs @code{query-replace-regexp}s, you could
probably similarly define the entire @code{CURLoption} enumeration.  I
have selected to transcribe a few that we will use in this tutorial.

If you're having trouble following the macrology, just macroexpand the
@code{curl-option} definition, or see the following macroexpansion,
conveniently downcased and reformatted:

@tutorialsource{DEFINE-CURL-OPTIONS macroexpansion}
@lisp
(progn
  (defcenum curl-option
    (:noprogress 43)
    (:nosignal 99)
    (:errorbuffer 10010)
    (:url 10002))
  'curl-option)
@end lisp

@noindent
That seems more than reasonable.  You may notice that we only use the
@var{type} to compute the real enumeration offset; we will also need
the type information later.

First, however, let's make sure a simple call to the foreign function
works:

@example
@sc{cffi-user>} (foreign-funcall "curl_easy_setopt"
               :pointer *easy-handle*
               curl-option :nosignal :long 1 curl-code)
@result{} 0
@end example

@code{foreign-funcall}, despite its surface simplicity, can be used to
call any C function.  Its first argument is a string, naming the
function to be called.  Next, for each argument, we pass the name of
the C type, which is the same as in @code{defcfun}, followed by a Lisp
object representing the data to be passed as the argument.  The final
argument is the return type, for which we use the @code{curl-code}
type defined earlier.

@code{defcfun} just puts a convenient fa@,cade on
@code{foreign-funcall}.@footnote{This isn't entirely true; some Lisps
don't support @code{foreign-funcall}, so @code{defcfun} is implemented
without it.  @code{defcfun} may also perform optimizations that
@code{foreign-funcall} cannot.}  Our earlier call to
@code{curl-global-init} could have been written as follows:

@example
@sc{cffi-user>} (foreign-funcall "curl_global_init" :long 0
                            curl-code)
@result{} 0
@end example

Before we continue, we will take a look at what @cffi{} can and can't
do, and why this is so.


@node Tutorial-Abstraction
@section Breaking the abstraction

@cindex breaking the abstraction
@cindex abstraction breaking
In @ref{Tutorial-Comparison,, What makes Lisp different}, we mentioned
that writing an @acronym{FFI} sometimes requires depending on
information not provided as part of the interface.  The easy option
@code{CURLOPT_WRITEDATA}, which we will not provide as part of the
Lisp interface, illustrates this issue.

Strictly speaking, the @code{curl-option} enumeration is not
necessary; we could have used @code{:int 99} instead of
@code{curl-option :nosignal} in our call to @code{curl_easy_setopt}
above.  We defined it anyway, in part to hide the fact that we are
breaking the abstraction that the C @code{enum} provides.  If the
c@acronym{URL} developers decide to change those numbers later, we
must change the Lisp enumeration, because enumeration values are not
provided in the compiled C library, @code{libcurl.so.3}.

@cffi{} works because the most useful things in C libraries ---
non-static functions and non-static variables --- are included
accessibly in @code{libcurl.so.3}.  A C compiler that violated this
would be considered a worthless compiler.

The other thing @code{define-curl-options} does is give the ``type''
of the third argument passed to @code{curl_easy_setopt}.  Using this
information, we can tell that the @code{:nosignal} option should
accept a long integer argument.  We can implicitly assume @code{t}
@equiv{} 1 and @code{nil} @equiv{} 0, as it is in C, which takes care
of the fact that @code{CURLOPT_NOSIGNAL} is really asking for a
boolean.

The ``type'' of @code{CURLOPT_WRITEDATA} is @code{objectpoint}.
However, it is really looking for a @code{FILE*}.
@code{CURLOPT_ERRORBUFFER} is looking for a @code{char*}, so there is
no obvious @cffi{} type but @code{:pointer}.

The first thing to note is that nowhere in the C interface includes
this information; it can only be found in the manual.  We could
disjoin these clearly different types ourselves, by splitting
@code{objectpoint} into @code{filepoint} and @code{charpoint}, but we
are still breaking the abstraction, because we have to augment the
entire enumeration form with this additional
information.@footnote{Another possibility is to allow the caller to
specify the desired C type of the third argument.  This is essentially
what happens in a call to the function written in C.}

@cindex streams and C
@cindex @sc{file}* and streams
The second is that the @code{CURLOPT_WRITEDATA} argument is completely
incompatible with the desired Lisp data, a
stream.@footnote{@xref{Other Kinds of Streams,,, libc, GNU C Library
Reference}, for a @acronym{GNU}-only way to extend the @code{FILE*}
type.  You could use this to convert Lisp streams to the needed C
data.  This would be quite involved and far outside the scope of this
tutorial.}  It is probably acceptable if we are controlling every file
we might want to use as this argument, in which case we can just call
the foreign function @code{fopen}.  Regardless, though, we can't write
to arbitrary streams, which is exactly what we want to do for this
application.

Finally, note that the @code{curl_easy_setopt} interface itself is a
hack, intended to work around some of the drawbacks of C.  The
definition of @code{Curl_setopt}, while long, is far less cluttered
than the equivalent disjoint-function set would be; in addition,
setting a new option in an old @code{libcurl} can generate a run-time
error rather than breaking the compile.  Lisp can just as concisely
generate functions as compare values, and the ``undefined function''
error is just as useful as any explicit error we could define here
might be.


@node Tutorial-Lisp easy_setopt
@section Option functions in Lisp

We could use @code{foreign-funcall} directly every time we wanted to
call @code{curl_easy_setopt}.  However, we can encapsulate some of the
necessary information with the following.

@lisp
;;; @lispcmt{We will use this type later in a more creative way.  For}
;;; @lispcmt{now, just consider it a marker that this isn't just any}
;;; @lispcmt{pointer.}
(defctype easy-handle :pointer)

(defmacro curl-easy-setopt (easy-handle enumerated-name
                            value-type new-value)
  "Call `curl_easy_setopt' on EASY-HANDLE, using ENUMERATED-NAME
as the OPTION.  VALUE-TYPE is the CFFI foreign type of the third
argument, and NEW-VALUE is the Lisp data to be translated to the
third argument.  VALUE-TYPE is not evaluated."
  `(foreign-funcall "curl_easy_setopt" easy-handle ,easy-handle
                    curl-option ,enumerated-name
                    ,value-type ,new-value curl-code))
@end lisp

Now we define a function for each kind of argument that encodes the
correct @code{value-type} in the above.  This can be done reasonably
in the @code{define-curl-options} macroexpansion; after all, that is
where the different options are listed!

@cindex Lispy C functions
We could make @code{cl:defun} forms in the expansion that simply call
@code{curl-easy-setopt}; however, it is probably easier and clearer to
use @code{defcfun}.  @code{define-curl-options} was becoming unwieldy,
so I defined some helpers in this new definition.

@smalllisp
(defun curry-curl-option-setter (function-name option-keyword)
  "Wrap the function named by FUNCTION-NAME with a version that
curries the second argument as OPTION-KEYWORD.

This function is intended for use in DEFINE-CURL-OPTION-SETTER."
  (setf (symbol-function function-name)
          (let ((c-function (symbol-function function-name)))
            (lambda (easy-handle new-value)
              (funcall c-function easy-handle option-keyword
                       new-value)))))

(defmacro define-curl-option-setter (name option-type
                                     option-value foreign-type)
  "Define (with DEFCFUN) a function NAME that calls
curl_easy_setopt.  OPTION-TYPE and OPTION-VALUE are the CFFI
foreign type and value to be passed as the second argument to
easy_setopt, and FOREIGN-TYPE is the CFFI foreign type to be used
for the resultant function's third argument.

This macro is intended for use in DEFINE-CURL-OPTIONS."
  `(progn
     (defcfun ("curl_easy_setopt" ,name) curl-code
       (easy-handle easy-handle)
       (option ,option-type)
       (new-value ,foreign-type))
     (curry-curl-option-setter ',name ',option-value)))

(defmacro define-curl-options (type-name type-offsets &rest enum-args)
  "As with CFFI:DEFCENUM, except each of ENUM-ARGS is as follows:

    (NAME TYPE NUMBER)

Where the arguments are as they are with the CINIT macro defined
in curl.h, except NAME is a keyword.

TYPE-OFFSETS is a plist of TYPEs to their integer offsets, as
defined by the CURLOPTTYPE_LONG et al constants in curl.h.

Also, define functions for each option named
set-`TYPE-NAME'-`OPTION-NAME', where OPTION-NAME is the NAME from
the above destructuring."
  (flet ((enumerated-value (type offset)
           (+ (getf type-offsets type) offset))
         ;;@lispcmt{map PROCEDURE, destructuring each of ENUM-ARGS}
         (map-enum-args (procedure)
           (mapcar (lambda (arg) (apply procedure arg)) enum-args))
         ;;@lispcmt{build a name like SET-CURL-OPTION-NOSIGNAL}
         (make-setter-name (option-name)
           (intern (concatenate
                    'string "SET-" (symbol-name type-name)
                    "-" (symbol-name option-name)))))
    `(progn
       (defcenum ,type-name
         ,@@(map-enum-args
            (lambda (name type number)
              (list name (enumerated-value type number)))))
       ,@@(map-enum-args
          (lambda (name type number)
            (declare (ignore number))
            `(define-curl-option-setter ,(make-setter-name name)
               ,type-name ,name ,(ecase type
                                   (long :long)
                                   (objectpoint :pointer)
                                   (functionpoint :pointer)
                                   (off-t :long)))))
       ',type-name)))
@end smalllisp

@noindent
Macroexpanding our @code{define-curl-options} form once more, we
see something different:

@lisp
(progn
  (defcenum curl-option
    (:noprogress 43)
    (:nosignal 99)
    (:errorbuffer 10010)
    (:url 10002))
  (define-curl-option-setter set-curl-option-noprogress
    curl-option :noprogress :long)
  (define-curl-option-setter set-curl-option-nosignal
    curl-option :nosignal :long)
  (define-curl-option-setter set-curl-option-errorbuffer
    curl-option :errorbuffer :pointer)
  (define-curl-option-setter set-curl-option-url
    curl-option :url :pointer)
  'curl-option)
@end lisp

@noindent
Macroexpanding one of the new @code{define-curl-option-setter}
forms yields the following:

@lisp
(progn
  (defcfun ("curl_easy_setopt" set-curl-option-nosignal) curl-code
    (easy-handle easy-handle)
    (option curl-option)
    (new-value :long))
  (curry-curl-option-setter 'set-curl-option-nosignal ':nosignal))
@end lisp

@noindent
Finally, let's try this out:

@example
@sc{cffi-user>} (set-curl-option-nosignal *easy-handle* 1)
@result{} 0
@end example

@noindent
Looks like it works just as well.  This interface is now reasonably
high-level to wash out some of the ugliness of the thinnest possible
@code{curl_easy_setopt} @acronym{FFI}, without obscuring the remaining
C bookkeeping details we will explore.


@node Tutorial-Memory
@section Memory management

According to the documentation for @code{curl_easy_setopt}, the type
of the third argument when @var{option} is @code{CURLOPT_ERRORBUFFER}
is @code{char*}.  Above, we've defined
@code{set-curl-option-errorbuffer} to accept a @code{:pointer} as the
new option value.  However, there is a @cffi{} type @code{:string},
which translates Lisp strings to C strings when passed as arguments to
foreign function calls.  Why not, then, use @code{:string} as the
@cffi{} type of the third argument?  There are two reasons, both
related to the necessity of breaking abstraction described in
@ref{Tutorial-Abstraction,, Breaking the abstraction}.

The first reason also applies to @code{CURLOPT_URL}, which we will use
to illustrate the point.  Assuming we have changed the type of the
third argument underlying @code{set-curl-option-url} to
@code{:string}, look at these two equivalent forms.

@lisp
(set-curl-option-url *easy-handle* "http://www.cliki.net/CFFI")

@equiv{} (with-foreign-string (url "http://www.cliki.net/CFFI")
     (foreign-funcall "curl_easy_setopt" easy-handle *easy-handle*
                      curl-option :url :pointer url curl-code))
@end lisp

@noindent
The latter, in fact, is mostly equivalent to what a foreign function
call's macroexpansion actually does.  As you can see, the Lisp string
@code{"@clikicffi{}"} is copied into a @code{char} array and
null-terminated; the pointer to beginning of this array, now a C
string, is passed as a @cffi{} @code{:pointer} to the foreign
function.

@cindex dynamic extent
@cindex foreign values with dynamic extent
Unfortunately, the C abstraction has failed us, and we must break it.
While @code{:string} works well for many @code{char*} arguments, it
does not for cases like this.  As the @code{curl_easy_setopt}
documentation explains, ``The string must remain present until curl no
longer needs it, as it doesn't copy the string.''  The C string
created by @code{with-foreign-string}, however, only has dynamic
extent: it is ``deallocated'' when the body (above containing the
@code{foreign-funcall} form) exits.

@cindex premature deallocation
If we are supposed to keep the C string around, but it goes away, what
happens when some @code{libcurl} function tries to access the
@acronym{URL} string?  We have reentered the dreaded world of C
``undefined behavior''.  In some Lisps, it will probably get a chunk
of the Lisp/C stack.  You may segfault.  You may get some random piece
of other data from the heap.  Maybe, in a world where ``dynamic
extent'' is defined to be ``infinite extent'', everything will turn
out fine.  Regardless, results are likely to be almost universally
unpleasant.@footnote{``@i{But I thought Lisp was supposed to protect
me from all that buggy C crap!}''  Before asking a question like that,
remember that you are a stranger in a foreign land, whose residents
have a completely different set of values.}

Returning to the current @code{set-curl-option-url} interface, here is
what we must do:

@lisp
(let (easy-handle)
  (unwind-protect
    (with-foreign-string (url "http://www.cliki.net/CFFI")
      (setf easy-handle (curl-easy-init))
      (set-curl-option-url easy-handle url)
      #|@lispcmt{do more with the easy-handle, like actually get the URL}|#)
    (when easy-handle
      (curl-easy-cleanup easy-handle))))
@end lisp

@c old comment to luis: I go on to say that this isn't obviously
@c extensible to new option settings that require C strings to stick
@c around, as it would involve re-evaluating the unwind-protect form
@c with more dynamic memory allocation.  So I plan to show how to
@c write something similar to ObjC's NSAutoreleasePool, to be managed
@c with a simple unwind-protect form.

@noindent
That is fine for the single string defined here, but for every string
option we want to pass, we have to surround the body of
@code{with-foreign-string} with another @code{with-foreign-string}
wrapper, or else do some extremely error-prone pointer manipulation
and size calculation in advance.  We could alleviate some of the pain
with a recursively expanding macro, but this would not remove the need
to modify the block every time we want to add an option, anathema as
it is to a modular interface.

Before modifying the code to account for this case, consider the other
reason we can't simply use @code{:string} as the foreign type.  In C,
a @code{char *} is a @code{char *}, not necessarily a string.  The
option @code{CURLOPT_ERRORBUFFER} accepts a @code{char *}, but does
not expect anything about the data there.  However, it does expect
that some @code{libcurl} function we call later can write a C string
of up to 255 characters there.  We, the callers of the function, are
expected to read the C string at a later time, exactly the opposite of
what @code{:string} implies.

With the semantics for an input string in mind --- namely, that the
string should be kept around until we @code{curl_easy_cleanup} the
easy handle --- we are ready to extend the Lisp interface:

@lisp
(defvar *easy-handle-cstrings* (make-hash-table)
  "Hashtable of easy handles to lists of C strings that may be
safely freed after the handle is freed.")

(defun make-easy-handle ()
  "Answer a new CURL easy interface handle, to which the lifetime
of C strings may be tied.  See `add-curl-handle-cstring'."
  (let ((easy-handle (curl-easy-init)))
    (setf (gethash easy-handle *easy-handle-cstrings*) '())
    easy-handle))

(defun free-easy-handle (handle)
  "Free CURL easy interface HANDLE and any C strings created to
be its options."
  (curl-easy-cleanup handle)
  (mapc #'foreign-string-free
        (gethash handle *easy-handle-cstrings*))
  (remhash handle *easy-handle-cstrings*))

(defun add-curl-handle-cstring (handle cstring)
  "Add CSTRING to be freed when HANDLE is, answering CSTRING."
  (car (push cstring (gethash handle *easy-handle-cstrings*))))
@end lisp

@noindent
Here we have redefined the interface to create and free handles, to
associate a list of allocated C strings with each handle while it
exists.  The strategy of using different function names to wrap around
simple foreign functions is more common than the solution implemented
earlier with @code{curry-curl-option-setter}, which was to modify the
function name's function slot.@footnote{There are advantages and
disadvantages to each approach; I chose to @code{(setf
symbol-function)} earlier because it entailed generating fewer magic
function names.}

Incidentally, the next step is to redefine
@code{curry-curl-option-setter} to allocate C strings for the
appropriate length of time, given a Lisp string as the
@code{new-value} argument:

@lisp
(defun curry-curl-option-setter (function-name option-keyword)
  "Wrap the function named by FUNCTION-NAME with a version that
curries the second argument as OPTION-KEYWORD.

This function is intended for use in DEFINE-CURL-OPTION-SETTER."
  (setf (symbol-function function-name)
          (let ((c-function (symbol-function function-name)))
            (lambda (easy-handle new-value)
              (funcall c-function easy-handle option-keyword
                       (if (stringp new-value)
                         (add-curl-handle-cstring
                          easy-handle
                          (foreign-string-alloc new-value))
                         new-value))))))
@end lisp

@noindent
A quick analysis of the code shows that you need only reevaluate the
@code{curl-option} enumeration definition to take advantage of these
new semantics.  Now, for good measure, let's reallocate the handle
with the new functions we just defined, and set its @acronym{URL}:

@example
@sc{cffi-user>} (curl-easy-cleanup *easy-handle*)
@result{} NIL
@sc{cffi-user>} (setf *easy-handle* (make-easy-handle))
@result{} #<FOREIGN-ADDRESS #x09844EE0>
@sc{cffi-user>} (set-curl-option-nosignal *easy-handle* 1)
@result{} 0
@sc{cffi-user>} (set-curl-option-url *easy-handle*
                                "http://www.cliki.net/CFFI")
@result{} 0
@end example

@cindex strings
For fun, let's inspect the Lisp value of the C string that was created
to hold @code{"@clikicffi{}"}.  By virtue of the implementation of
@code{add-curl-handle-cstring}, it should be accessible through the
hash table defined:

@example
@sc{cffi-user>} (foreign-string-to-lisp
            (car (gethash *easy-handle* *easy-handle-cstrings*)))
@result{} "http://www.cliki.net/CFFI"
@end example

@noindent
Looks like that worked, and @code{libcurl} now knows what
@acronym{URL} we want to retrieve.

Finally, we turn back to the @code{:errorbuffer} option mentioned at
the beginning of this section.  Whereas the abstraction added to
support string inputs works fine for cases like @code{CURLOPT_URL}, it
hides the detail of keeping the C string; for @code{:errorbuffer},
however, we need that C string.

In a moment, we'll define something slightly cleaner, but for now,
remember that you can always hack around anything.  We're modifying
handle creation, so make sure you free the old handle before
redefining @code{free-easy-handle}.

@smalllisp
(defvar *easy-handle-errorbuffers* (make-hash-table)
  "Hashtable of easy handles to C strings serving as error
writeback buffers.")

;;; @lispcmt{An extra byte is very little to pay for peace of mind.}
(defparameter *curl-error-size* 257
  "Minimum char[] size used by cURL to report errors.")

(defun make-easy-handle ()
  "Answer a new CURL easy interface handle, to which the lifetime
of C strings may be tied.  See `add-curl-handle-cstring'."
  (let ((easy-handle (curl-easy-init)))
    (setf (gethash easy-handle *easy-handle-cstrings*) '())
    (setf (gethash easy-handle *easy-handle-errorbuffers*)
            (foreign-alloc :char :count *curl-error-size*
                           :initial-element 0))
    easy-handle))

(defun free-easy-handle (handle)
  "Free CURL easy interface HANDLE and any C strings created to
be its options."
  (curl-easy-cleanup handle)
  (foreign-free (gethash handle *easy-handle-errorbuffers*))
  (remhash handle *easy-handle-errorbuffers*)
  (mapc #'foreign-string-free
        (gethash handle *easy-handle-cstrings*))
  (remhash handle *easy-handle-cstrings*))

(defun get-easy-handle-error (handle)
  "Answer a string containing HANDLE's current error message."
  (foreign-string-to-lisp
   (gethash handle *easy-handle-errorbuffers*)))
@end smalllisp

Be sure to once again set the options we've set thus far.  You may
wish to define yet another wrapper function to do this.


@node Tutorial-Callbacks
@section Calling Lisp from C

If you have been reading
@uref{http://curl.haxx.se/libcurl/c/curl_easy_setopt.html,,
@code{curl_easy_setopt(3)}}, you should have noticed that some options
accept a function pointer.  In particular, we need one function
pointer to set as @code{CURLOPT_WRITEFUNCTION}, to be called by
@code{libcurl} rather than the reverse, in order to receive data as it
is downloaded.

A binding writer without the aid of @acronym{FFI} usually approaches
this problem by writing a C function that accepts C data, converts to
the language's internal objects, and calls the callback provided by
the user, again in a reverse of usual practices.

The @cffi{} approach to callbacks precisely mirrors its differences
with the non-@acronym{FFI} approach on the ``calling C from Lisp''
side, which we have dealt with exclusively up to now.  That is, you
define a callback function in Lisp using @code{defcallback}, and
@cffi{} effectively creates a C function to be passed as a function
pointer.

@impnote{This is much trickier than calling C functions from Lisp, as
it literally involves somehow generating a new C function that is as
good as any created by the compiler.  Therefore, not all Lisps support
them.  @xref{Implementation Support}, for information about @cffi{}
support issues in this and other areas.  You may want to consider
changing to a Lisp that supports callbacks in order to continue with
this tutorial.}

@cindex callback definition
@cindex defining callbacks
Defining a callback is very similar to defining a callout; the main
difference is that we must provide some Lisp forms to be evaluated as
part of the callback.  Here is the signature for the function the
@code{:writefunction} option takes:

@example
size_t
@var{function}(void *ptr, size_t size, size_t nmemb, void *stream);
@end example

@impnote{size_t is almost always an unsigned int.  You can get this
and many other types using feature tests for your system by using
cffi-grovel.}

The above signature trivially translates into a @cffi{}
@code{defcallback} form, as follows.

@lisp
;;; @lispcmt{Alias in case size_t changes.}
(defctype size :unsigned-int)

;;; @lispcmt{To be set as the CURLOPT_WRITEFUNCTION of every easy handle.}
(defcallback easy-write size ((ptr :pointer) (size size)
                              (nmemb size) (stream :pointer))
  (let ((data-size (* size nmemb)))
    (handler-case
      ;; @lispcmt{We use the dynamically-bound *easy-write-procedure* to}
      ;; @lispcmt{call a closure with useful lexical context.}
      (progn (funcall (symbol-value '*easy-write-procedure*)
                      (foreign-string-to-lisp ptr data-size nil))
             data-size)         ;@lispcmt{indicates success}
      ;; @lispcmt{The WRITEFUNCTION should return something other than the}
      ;; @lispcmt{#bytes available to signal an error.}
      (error () (if (zerop data-size) 1 0)))))
@end lisp

First, note the correlation of the first few forms, used to declare
the C function's signature, with the signature in C syntax.  We
provide a Lisp name for the function, its return type, and a name and
type for each argument.

In the body, we call the dynamically-bound
@code{*easy-write-procedure*} with a ``finished'' translation, of
pulling together the raw data and size into a Lisp string, rather than
deal with the data directly.  As part of calling
@code{curl_easy_perform} later, we'll bind that variable to a closure
with more useful lexical bindings than the top-level
@code{defcallback} form.

Finally, we make a halfhearted effort to prevent non-local exits from
unwinding the C stack, covering the most likely case with an
@code{error} handler, which is usually triggered
unexpectedly.@footnote{Unfortunately, we can't protect against
@emph{all} non-local exits, such as @code{return}s and @code{throw}s,
because @code{unwind-protect} cannot be used to ``short-circuit'' a
non-local exit in Common Lisp, due to proposal @code{minimal} in
@uref{http://www.lisp.org/HyperSpec/Issues/iss152-writeup.html,
@acronym{ANSI} issue @sc{Exit-Extent}}.  Furthermore, binding an
@code{error} handler prevents higher-up code from invoking restarts
that may be provided under the callback's dynamic context.  Such is
the way of compromise.}  The reason is that most C code is written to
understand its own idiosyncratic error condition, implemented above in
the case of @code{curl_easy_perform}, and more ``undefined behavior''
can result if we just wipe C stack frames without allowing them to
execute whatever cleanup actions as they like.

Using the @code{CURLoption} enumeration in @file{curl.h} once more, we
can describe the new option by modifying and reevaluating
@code{define-curl-options}.

@lisp
(define-curl-options curl-option
    (long 0 objectpoint 10000 functionpoint 20000 off-t 30000)
  (:noprogress long 43)
  (:nosignal long 99)
  (:errorbuffer objectpoint 10)
  (:url objectpoint 2)
  (:writefunction functionpoint 11)) ;@lispcmt{new item here}
@end lisp

Finally, we can use the defined callback and the new
@code{set-curl-option-writefunction} to finish configuring the easy
handle, using the @code{callback} macro to retrieve a @cffi{}
@code{:pointer}, which works like a function pointer in C code.

@example
@sc{cffi-user>} (set-curl-option-writefunction
            *easy-handle* (callback easy-write))
@result{} 0
@end example


@node Tutorial-Completion
@section A complete @acronym{FFI}?

@c TeX goes insane on @uref{@clikicffi{}}

With all options finally set and a medium-level interface developed,
we can finish the definition and retrieve
@uref{http://www.cliki.net/CFFI}, as is done in the tutorial.

@lisp
(defcfun "curl_easy_perform" curl-code
  (handle easy-handle))
@end lisp

@example
@sc{cffi-user>} (with-output-to-string (contents)
             (let ((*easy-write-procedure*
                     (lambda (string)
                       (write-string string contents))))
               (declare (special *easy-write-procedure*))
               (curl-easy-perform *easy-handle*)))
@result{} "<!DOCTYPE HTML PUBLIC \"-//W3C//DTD HTML 4.01//EN\"
@enddots{}
Now fear, comprehensively</P>
"
@end example

Of course, that itself is slightly unwieldy, so you may want to define
a function around it that simply retrieves a @acronym{URL}.  I will
leave synthesis of all the relevant @acronym{REPL} forms presented
thus far into a single function as an exercise for the reader.

The remaining sections of this tutorial explore some advanced features
of @cffi{}; the definition of new types will receive special
attention.  Some of these features are essential for particular
foreign function calls; some are very helpful when trying to develop a
Lispy interface to C.


@node Tutorial-Types
@section Defining new types

We've occasionally used the @code{defctype} macro in previous sections
as a kind of documentation, much what you'd use @code{typedef} for in
C.  We also tried one special kind of type definition, the
@code{defcenum} type.  @xref{defcstruct}, for a definition macro that
may come in handy if you need to use C @code{struct}s as data.

@cindex type definition
@cindex data in Lisp and C
@cindex translating types
However, all of these are mostly sugar for the powerful underlying
foreign type interface called @dfn{type translators}.  You can easily
define new translators for any simple named foreign type.  Since we've
defined the new type @code{curl-code} to use as the return type for
various @code{libcurl} functions, we can use that to directly convert
c@acronym{URL} errors to Lisp errors.

@code{defctype}'s purpose is to define simple @code{typedef}-like
aliases.  In order to use @dfn{type translators} we must use the
@code{define-foreign-type} macro.  So let's redefine @code{curl-code}
using it.

@lisp
(define-foreign-type curl-code-type ()
  ()
  (:actual-type :int)
  (:simple-parser curl-code))
@end lisp

@code{define-foreign-type} is a thin wrapper around @code{defclass}.
For now, all you need to know in the context of this example is that
it does what @code{(defctype curl-code :int)} would do and,
additionally, defines a new class @code{curl-code-type} which we will
take advantage of shortly.

The @code{CURLcode} enumeration seems to follow the typical error code
convention of @samp{0} meaning all is well, and each non-zero integer
indicating a different kind of error.  We can apply that trivially to
differentiate between normal exits and error exits.

@lisp
(define-condition curl-code-error (error)
  (($code :initarg :curl-code :reader curl-error-code))
  (:report (lambda (c stream)
             (format stream "libcurl function returned error ~A"
                            (curl-error-code c))))
  (:documentation "Signalled when a libcurl function answers
a code other than CURLE_OK."))

(defmethod translate-from-foreign (value (type curl-code-type))
  "Raise a CURL-CODE-ERROR if VALUE, a curl-code, is non-zero."
  (if (zerop value)
      :curle-ok
      (error 'curl-code-error :curl-code value)))
@end lisp

@noindent
The heart of this translator is new method
@code{translate-from-foreign}.  By specializing the @var{type}
parameter on @code{curl-code-type}, we immediately modify the behavior
of every function that returns a @code{curl-code} to pass the result
through this new method.

To see the translator in action, try invoking a function that returns
a @code{curl-code}.  You need to reevaluate the respective
@code{defcfun} form so that it picks up the new @code{curl-code}
definition.

@example
@sc{cffi-user>} (set-curl-option-nosignal *easy-handle* 1)
@result{} :CURLE-OK
@end example

@noindent
As the result was @samp{0}, the new method returned @code{:curle-ok},
just as specified.@footnote{It might be better to return
@code{(values)} than @code{:curle-ok} in real code, but this is good
for illustration.}  I will leave disjoining the separate
@code{CURLcode}s into condition types and improving the @code{:report}
function as an exercise for you.

The creation of @code{*easy-handle-cstrings*} and
@code{*easy-handle-errorbuffers*} as properties of @code{easy-handle}s
is a kluge.  What we really want is a Lisp structure that stores these
properties along with the C pointer.  Unfortunately,
@code{easy-handle} is currently just a fancy name for the foreign type
@code{:pointer}; the actual pointer object varies from Common Lisp
implementation to implementation, needing only to satisfy
@code{pointerp} and be returned from @code{make-pointer} and friends.

One solution that would allow us to define a new Lisp structure to
represent @code{easy-handle}s would be to write a wrapper around every
function that currently takes an @code{easy-handle}; the wrapper would
extract the pointer and pass it to the foreign function.  However, we
can use type translators to more elegantly integrate this
``translation'' into the foreign function calling framework, using
@code{translate-to-foreign}.

@smalllisp
(defclass easy-handle ()
  ((pointer :initform (curl-easy-init)
            :documentation "Foreign pointer from curl_easy_init")
   (error-buffer
    :initform (foreign-alloc :char :count *curl-error-size*
                             :initial-element 0)
    :documentation "C string describing last error")
   (c-strings :initform '()
              :documentation "C strings set as options"))
  (:documentation "I am a parameterization you may pass to
curl-easy-perform to perform a cURL network protocol request."))

(defmethod initialize-instance :after ((self easy-handle) &key)
  (set-curl-option-errorbuffer self (slot-value self 'error-buffer)))

(defun add-curl-handle-cstring (handle cstring)
  "Add CSTRING to be freed when HANDLE is, answering CSTRING."
  (car (push cstring (slot-value handle 'c-strings))))

(defun get-easy-handle-error (handle)
  "Answer a string containing HANDLE's current error message."
  (foreign-string-to-lisp
   (slot-value handle 'error-buffer)))

(defun free-easy-handle (handle)
  "Free CURL easy interface HANDLE and any C strings created to
be its options."
  (with-slots (pointer error-buffer c-strings) handle
    (curl-easy-cleanup pointer)
    (foreign-free error-buffer)
    (mapc #'foreign-string-free c-strings)))

(define-foreign-type easy-handle-type ()
  ()
  (:actual-type :pointer)
  (:simple-parser easy-handle))

(defmethod translate-to-foreign (handle (type easy-handle-type))
  "Extract the pointer from an easy-HANDLE."
  (slot-value handle 'pointer))
@end smalllisp

While we changed some of the Lisp functions defined earlier to use
@acronym{CLOS} slots rather than hash tables, the foreign functions
work just as well as they did before.

@cindex limitations of type translators
The greatest strength, and the greatest limitation, of the type
translator comes from its generalized interface.  As stated
previously, we could define all foreign function calls in terms of the
primitive foreign types provided by @cffi{}.  The type translator
interface allows us to cleanly specify the relationship between Lisp
and C data, independent of where it appears in a function call.  This
independence comes at a price; for example, it cannot be used to
modify translation semantics based on other arguments to a function
call.  In these cases, you should rely on other features of Lisp,
rather than the powerful, yet domain-specific, type translator
interface.


@node Tutorial-Conclusion
@section What's next?

@cffi{} provides a rich and powerful foundation for communicating with
foreign libraries; as we have seen, it is up to you to make that
experience a pleasantly Lispy one.  This tutorial does not cover all
the features of @cffi{}; please see the rest of the manual for
details.  In particular, if something seems obviously missing, it is
likely that either code or a good reason for lack of code is already
present.

@impnote{There are some other things in @cffi{} that might deserve
tutorial sections, such as define-foreign-type,
free-translated-object, or structs.  Let us know which ones you care
about.}


@c ===================================================================
@c CHAPTER: Wrapper generators

@node Wrapper generators
@chapter Wrapper generators

@cffi{}'s interface is designed for human programmers, being aimed at
aesthetic as well as technical sophistication.  However, there are a
few programs aimed at translating C and C++ header files, or
approximations thereof, into @cffi{} forms constituting a foreign
interface to the symbols in those files.

These wrapper generators are known to support output of @cffi{} forms.

@table @asis
@item @uref{http://www.cliki.net/Verrazano,Verrazano}
Designed specifically for Common Lisp.  Uses @acronym{GCC}'s parser
output in @acronym{XML} format to discover functions, variables, and
other header file data.  This means you need @acronym{GCC} to generate
forms; on the other hand, the parser employed is mostly compliant with
@acronym{ANSI} C.

@item @uref{http://www.cliki.net/SWIG,SWIG}
A foreign interface generator originally designed to generate Python
bindings, it has been ported to many other systems, including @cffi{}
in version 1.3.28.  Includes its own C declaration munger, not
intended to be fully-compliant with @acronym{ANSI} C.
@end table

First, this manual does not describe use of these other programs; they
have documentation of their own.  If you have problems using a
generated interface, please look at the output @cffi{} forms and
verify that they are a correct @cffi{} interface to the library in
question; if they are correct, contact @cffi{} developers with
details, keeping in mind that they communicate in terms of those forms
rather than any particular wrapper generator.  Otherwise, contact the
maintainers of the wrapper generator you are using, provided you can
reasonably expect more accuracy from the generator.

When is more accuracy an unreasonable expectation?  As described in
the tutorial (@pxref{Tutorial-Abstraction,, Breaking the
abstraction}), the information in C declarations is insufficient to
completely describe every interface.  In fact, it is quite common to
run into an interface that cannot be handled automatically, and
generators should be excused from generating a complete interface in
these cases.

As further described in the tutorial, the thinnest Lisp interface to a
C function is not always the most pleasant one.  In many cases, you
will want to manually write a Lispier interface to the C functions
that interest you.

Wrapper generators should be treated as time-savers, not complete
automation of the full foreign interface writing job.  Reports of the
amount of work done by generators vary from 30% to 90%.  The
incremental development style enabled by @cffi{} generally reduces
this proportion below that for languages like Python.

@c Where I got the above 30-90% figures:
@c 30%: lemonodor's post about SWIG
@c 90%: Balooga on #lisp.  He said 99%, but that's probably an
@c      exaggeration (leave it to me to pass judgement :)
@c -stephen


@c ===================================================================
@c CHAPTER: Foreign Types

@node Foreign Types
@chapter Foreign Types

Foreign types describe how data is translated back and forth between C
and Lisp. @cffi{} provides various built-in types and allows the user to
define new types.

@menu
* Built-In Types::              
* Other Types::                 
* Defining Foreign Types::      
* Foreign Type Translators::    
* Optimizing Type Translators:: 
* Foreign Structure Types::     
* Operations on Types::         
* Allocating Foreign Objects::  

Dictionary

* convert-from-foreign::        
* convert-to-foreign::          
* defbitfield::                 
* defcstruct::                  
* defcunion::                   
* defctype::                    
* defcenum::                    
@c * define-type-spec-parser::  
* define-foreign-type::         
* define-parse-method::         
@c * explain-foreign-slot-value:
* foreign-bitfield-symbols::    
* foreign-bitfield-value::      
* foreign-enum-keyword::        
* foreign-enum-value::          
* foreign-slot-names::          
* foreign-slot-offset::         
* foreign-slot-pointer::        
* foreign-slot-value::          
* foreign-type-alignment::      
* foreign-type-size::           
* free-converted-object::       
* free-translated-object::      
* translate-from-foreign::      
* translate-to-foreign::        
* with-foreign-slots::          
@end menu

@node Built-In Types
@section Built-In Types

@ForeignType{:char}
@ForeignType{:unsigned-char}
@ForeignType{:short}
@ForeignType{:unsigned-short}
@ForeignType{:int}
@ForeignType{:unsigned-int}
@ForeignType{:long}
@ForeignType{:unsigned-long}
@ForeignType{:long-long}
@ForeignType{:unsigned-long-long}

These types correspond to the native C integer types according to the
@acronym{ABI} of the Lisp implementation's host system.

@ForeignType{:uchar}
@ForeignType{:ushort}
@ForeignType{:uint}
@ForeignType{:ulong}
@ForeignType{:llong}
@ForeignType{:ullong}

For convenience, the above types are provided as shortcuts for
@code{unsigned-char}, @code{unsigned-short}, @code{unsigned-int},
@code{unsigned-long}, @code{long-long} and @code{unsigned-long-long},
respectively.

@code{:long-long} and @code{:unsigned-long-long} are not supported on
all implementations. When those types are @strong{not} available, the
symbol @code{cffi-features:no-long-long} is pushed into
@code{*features*}.

@ForeignType{:int8}
@ForeignType{:uint8}
@ForeignType{:int16}
@ForeignType{:uint16}
@ForeignType{:int32}
@ForeignType{:uint32}
@ForeignType{:int64}
@ForeignType{:uint64}

Foreign integer types of specific sizes, corresponding to the C types
defined in @code{stdint.h}.

@c @ForeignType{:size}
@c @ForeignType{:ssize}
@c @ForeignType{:ptrdiff}
@c @ForeignType{:time}

@c Foreign integer types corresponding to the standard C types (without
@c the @code{_t} suffix).

@c @impnote{These are not implemented yet. --luis}

@c @impnote{I'm sure there are more of these that could be useful, let's
@c add any types that can't be defined portably to this list as
@c necessary. --james}

@ForeignType{:float}
@ForeignType{:double}

On all systems, the @code{:float} and @code{:double} types represent a
C @code{float} and @code{double}, respectively. On most but not all
systems, @code{:float} and @code{:double} represent a Lisp
@code{single-float} and @code{double-float}, respectively. It is not
so useful to consider the relationship between Lisp types and C types
as isomorphic, as simply to recognize the relationship, and relative
precision, among each respective category.

@ForeignType{:long-double}

This type is only supported on SCL.

@ForeignType{:pointer &optional type}

A foreign pointer to an object of any type, corresponding to
@code{void *}.  You can optionally specify type of pointer
(e.g. @code{(:pointer :char)}).  Although @cffi{} won't do anything
with that information yet, it is useful for documentation purposes.

@ForeignType{:void}

No type at all. Only valid as the return type of a function.

@node Other Types
@section Other Types

@cffi{} also provides a few useful types that aren't built-in C
types.

@ForeignType{:string}

The @code{:string} type performs automatic conversion between Lisp and
C strings. Note that, in the case of functions the converted C string
will have dynamic extent (i.e.@: it will be automatically freed after
the foreign function returns).

In addition to Lisp strings, this type will also convert
Lisp arrays of element type @code{(unsigned-byte 8)} and will pass
foreign pointers unmodified.

A method for @ref{free-translated-object} is specialized for this
type. So, for example, foreign strings allocated by this type and
passed to a foreign function will be freed after the function
returns.

@lisp
CFFI> (foreign-funcall "getenv" :string "SHELL" :string)
@result{} "/bin/bash"

CFFI> (with-foreign-string (str "abcdef")
        (foreign-funcall "strlen" :string str :int))
@result{} 6

CFFI> (let ((str (make-array 4 :element-type '(unsigned-byte 8)
                             :initial-element 65)))
        (foreign-funcall "strlen" :string str :int))
@result{} 4
@end lisp

@ForeignType{:string+ptr}

Like @code{:string} but returns a list with two values when convert
from C to Lisp: a Lisp string and the C string's foreign pointer.

@lisp
CFFI> (foreign-funcall "getenv" :string "SHELL" :string+ptr)
@result{} ("/bin/bash" #.(SB-SYS:INT-SAP #XBFFFFC6F))
@end lisp

@ForeignType{:boolean &optional (base-type :int)}

The @code{:boolean} type converts between a Lisp boolean and a C
boolean. It canonicalizes to @var{base-type} which is @code{:int} by
default.

@lisp
(convert-to-foreign nil :boolean) @result{} 0
(convert-to-foreign t :boolean) @result{} 1
(convert-from-foreign 0 :boolean) @result{} nil
(convert-from-foreign 1 :boolean) @result{} t
@end lisp

@ForeignType{:wrapper base-type &key to-c from-c}

The @code{:wrapper} type stores two symbols passed to the @var{to-c}
and @var{from-c} arguments. When a value is being translated to or
from C, this type @code{funcall}s the respective symbol.

@code{:wrapper} types will be typedefs for @var{base-type} and will
inherit its translators, if any.

Here's an example of how the @code{:boolean} type could be defined in
terms of @code{:wrapper}.

@lisp
(defun bool-c-to-lisp (value)
  (not (zerop value)))

(defun bool-lisp-to-c (value)
  (if value 1 0))

(defctype my-bool (:wrapper :int :from-c bool-c-to-lisp
                                 :to-c bool-lisp-to-c))

(convert-to-foreign nil 'my-bool) @result{} 0
(convert-from-foreign 1 'my-bool) @result{} t
@end lisp

@node Defining Foreign Types
@section Defining Foreign Types

You can define simple C-like @code{typedef}s through the
@code{defctype} macro. Defining a typedef is as simple as giving
@code{defctype} a new name and the name of the type to be wrapped.

@lisp
;;; @lispcmt{Define MY-INT as an alias for the built-in type :INT.}
(defctype my-int :int)
@end lisp

With this type definition, one can, for instance, declare arguments to
foreign functions as having the type @code{my-int}, and they will be
passed as integers.

@subheading More complex types

@cffi{} offers another way to define types through
@code{define-foreign-type}, a thin wrapper macro around
@code{defclass}. As an example, let's go through the steps needed to
define a @code{(my-string &key encoding)} type. First, we need to
define our type class:

@lisp
(define-foreign-type my-string-type ()
  ((encoding :reader string-type-encoding :initarg :encoding))
  (:actual-type :pointer))
@end lisp

The @code{:actual-type} class option tells CFFI that this type will
ultimately be passed to and received from foreign code as a
@code{:pointer}. Now you need to tell CFFI how to parse a type
specification such as @code{(my-string :encoding :utf8)} into an
instance of @code{my-string-type}.  We do that with
@code{define-parse-method}:

@lisp
(define-parse-method my-string (&key (encoding :utf-8))
  (make-instance 'my-string-type :encoding encoding))
@end lisp

The next section describes how make this type actually translate
between C and Lisp strings.

@node Foreign Type Translators
@section Foreign Type Translators

Type translators are used to automatically convert Lisp values to or
from foreign values.  For example, using type translators, one can
take the @code{my-string} type defined in the previous section and
specify that it should:

@itemize
@item
convert C strings to Lisp strings;
@item
convert Lisp strings to newly allocated C strings;
@item
free said C strings when they are no longer needed.
@end itemize

In order to tell @cffi{} how to automatically convert Lisp values to
foreign values, define a specialized method for the
@code{translate-to-foreign} generic function:

@lisp
;;; @lispcmt{Define a method that converts Lisp strings to C strings.}
(defmethod translate-to-foreign (string (type my-string-type))
  (foreign-string-alloc string :encoding (string-type-encoding type)))
@end lisp

@noindent
From now on, whenever an object is passed as a @code{my-string} to a
foreign function, this method will be invoked to convert the Lisp
value. To perform the inverse operation, which is needed for functions
that return a @code{my-string}, specialize the
@code{translate-from-foreign} generic function in the same manner:

@lisp
;;; @lispcmt{Define a method that converts C strings to Lisp strings.}
(defmethod translate-from-foreign (pointer (type my-string-type))
  (foreign-string-to-lisp pointer :encoding (string-type-encoding type)))
@end lisp

@noindent
When a @code{translate-to-foreign} method requires allocation of
foreign memory, you must also define a @code{free-translated-object}
method to free the memory once the foreign object is no longer needed,
otherwise you'll be faced with memory leaks.  This generic function is
called automatically by @cffi{} when passing objects to foreign
functions. Let's do that:

@lisp
;;; @lispcmt{Free strings allocated by translate-to-foreign.}
(defmethod free-translated-object (pointer (type my-string-type) param)
  (declare (ignore param))
  (foreign-string-free pointer))
@end lisp

@noindent
In this specific example, we don't need the @var{param} argument, so
we ignore it. See @ref{free-translated-object}, for an explanation of
its purpose and how you can use it.

A type translator does not necessarily need to convert the value.  For
example, one could define a typedef for @code{:pointer} that ensures,
in the @code{translate-to-foreign} method, that the value is not a
null pointer, signalling an error if a null pointer is passed.  This
would prevent some pointer errors when calling foreign functions that
cannot handle null pointers.

@strong{Please note:} these methods are meant as extensible hooks
only, and you should not call them directly.  Use
@code{convert-to-foreign}, @code{convert-from-foreign} and
@code{free-converted-object} instead.

@xref{Tutorial-Types,, Defining new types}, for another example of
type translators.

@node Optimizing Type Translators
@section Optimizing Type Translators

@cindex type translators, optimizing
@cindex compiler macros for type translation
@cindex defining type-translation compiler macros
Being based on generic functions, the type translation mechanism
described above can add a bit of overhead.  This is usually not
significant, but we nevertheless provide a way of getting rid of the
overhead for the cases where it matters.

A good way to understand this issue is to look at the code generated
by @code{defcfun}. Consider the following example using the previously
defined @code{my-string} type:

@lisp
CFFI> (macroexpand-1 '(defcfun foo my-string (x my-string)))
;; (simplified, downcased, etc...)
(defun foo (x)
  (multiple-value-bind (#:G2019 #:PARAM3149)
      (translate-to-foreign x #<MY-STRING-TYPE @{11ED5A79@}>)
    (unwind-protect
        (translate-from-foreign
         (foreign-funcall "foo" :pointer #:G2019 :pointer)
         #<MY-STRING-TYPE @{11ED5659@}>)
      (free-translated-object #:G2019 #<MY-STRING-TYPE @{11ED51A79@}>
                              #:PARAM3149))))
@end lisp

@noindent
In order to get rid of those generic function calls, @cffi{} has
another set of extensible generic functions that provide functionality
similar to @acronym{CL}'s compiler macros:
@code{expand-to-foreign-dyn}, @code{expand-to-foreign} and
@code{expand-from-foreign}. Here's how one could define a
@code{my-boolean} with them:

@lisp
(define-foreign-type my-boolean-type ()
  ()
  (:actual-type :int)
  (:simple-parser my-boolean))

(defmethod expand-to-foreign (value (type my-boolean-type))
  `(if ,value 1 0))

(defmethod expand-from-foreign (value (type my-boolean-type))
  `(not (zerop ,value)))
@end lisp

@noindent
And here's what the macroexpansion of a function using this type would
look like:

@lisp
CFFI> (macroexpand-1 '(defcfun bar my-boolean (x my-boolean)))
;; (simplified, downcased, etc...)
(defun bar (x)
  (let ((#:g3182 (if x 1 0)))
    (not (zerop (foreign-funcall "bar" :int #:g3182 :int)))))
@end lisp

@noindent
No generic function overhead.

Let's go back to our @code{my-string} type.  The expansion interface
has no equivalent of @code{free-translated-object}; you must instead
define a method on @code{expand-to-foreign-dyn}, the third generic
function in this interface.  This is especially useful when you can
allocate something much more efficiently if you know the object has
dynamic extent, as is the case with function calls that don't save the
relevant allocated arguments.

This exactly what we need for the @code{my-string} type:

@lisp
(defmethod expand-from-foreign (form (type my-string-type))
  `(foreign-string-to-lisp ,form))

(defmethod expand-to-foreign-dyn (value var body (type my-string-type))
  (let ((encoding (string-type-encoding type)))
    `(with-foreign-string (,var ,value :encoding ',encoding)
       ,@@body)))
@end lisp

@noindent
So let's look at the macro expansion:

@lisp
CFFI> (macroexpand-1 '(defcfun foo my-string (x my-string)))
;; (simplified, downcased, etc...)
(defun foo (x)
  (with-foreign-string (#:G2021 X :encoding ':utf-8)
    (foreign-string-to-lisp
     (foreign-funcall "foo" :pointer #:g2021 :pointer))))
@end lisp

@noindent
Again, no generic function overhead.

@subheading Other details

To short-circuit expansion and use the @code{translate-*} functions
instead, simply call the next method.  Return its result in cases
where your method cannot generate an appropriate replacement for it.
This analogous to the @code{&whole form} mechanism compiler macros
provide.

The @code{expand-*} methods have precedence over their
@code{translate-*} counterparts and are guaranteed to be used in
@code{defcfun}, @code{foreign-funcall}, @code{defcvar} and
@code{defcallback}.  If you define a method on each of the
@code{expand-*} generic functions, you are guaranteed to have full
control over the expressions generated for type translation in these
macros.

They may or may not be used in other @cffi{} operators that need to
translate between Lisp and C data; you may only assume that
@code{expand-*} methods will probably only be called during Lisp
compilation.

@code{expand-to-foreign-dyn} has precedence over
@code{expand-to-foreign} and is only used in @code{defcfun} and
@code{foreign-funcall}, only making sense in those contexts.

@strong{Important note:} this set of generic functions is called at
macroexpansion time.  Methods are defined when loaded or evaluated,
not compiled.  You are responsible for ensuring that your
@code{expand-*} methods are defined when the @code{foreign-funcall} or
other forms that use them are compiled.  One way to do this is to put
the method definitions earlier in the file and inside an appropriate
@code{eval-when} form; another way is to always load a separate Lisp
or @acronym{FASL} file containing your @code{expand-*} definitions
before compiling files with forms that ought to use them.  Otherwise,
they will not be found and the runtime translators will be used
instead.

@node Foreign Structure Types
@section Foreign Structure Types

For more involved C types than simple aliases to built-in types, such
as you can make with @code{defctype}, @cffi{} allows declaration of
structures and unions with @code{defcstruct} and @code{defcunion}.

For example, consider this fictional C structure declaration holding
some personal information:

@example
struct person @{
  int number;
  char* reason;
@};
@end example

@noindent
The equivalent @code{defcstruct} form follows:

@lisp
(defcstruct person
  (number :int)
  (reason :string))
@end lisp

@cffi{} knows how to align C @code{struct}s, and how to figure in
padding between struct elements.

Please note that this interface is only for those that must know about
the values contained in a relevant struct.  If the library you are
interfacing returns an opaque pointer that needs only be passed to
other C library functions, by all means just use @code{:pointer} or a
type-safe definition munged together with @code{defctype} and type
translation.

@node Operations on Types
@section Operations on Types

@impnote{Which ``operations'' are worth going over here? --stephen}

@node Allocating Foreign Objects
@section Allocating Foreign Objects

@c I moved this because I moved with-foreign-object to the Pointers
@c chapter, where foreign-alloc is.

@xref{Allocating Foreign Memory}.


@c ===================================================================
@c CONVERT-FROM-FOREIGN

@node convert-from-foreign
@unnumberedsec convert-from-foreign
@subheading Syntax
@Function{convert-from-foreign foreign-value type @result{} value}

@subheading Arguments and Values

@table @var
@item foreign-value
The primitive C value as returned from a primitive foreign function or
from @code{convert-to-foreign}.

@item type
A @cffi{} type specifier.

@item value
The Lisp value translated from @var{foreign-value}.
@end table

@subheading Description

This is an external interface to the type translation facility.  In
the implementation, all foreign functions are ultimately defined as
type translation wrappers around primitive foreign function
invocations.

This function is available mostly for inspection of the type
translation process, and possibly optimization of special cases of
your foreign function calls.

Its behavior is better described under @code{translate-from-foreign}'s
documentation.

@subheading Examples

@lisp
CFFI-USER> (convert-to-foreign "a boat" :string)
@result{} #<FOREIGN-ADDRESS #x097ACDC0>
@result{} (T)
CFFI-USER> (convert-from-foreign * :string)
@result{} "a boat"
@end lisp

@subheading See Also
@seealso{convert-to-foreign} @*
@seealso{translate-from-foreign}


@c ===================================================================
@c CONVERT-TO-FOREIGN

@node convert-to-foreign
@unnumberedsec convert-to-foreign
@subheading Syntax
@Function{convert-to-foreign value type @result{} foreign-value, alloc-params}

@subheading Arguments and Values

@table @var
@item value
The Lisp object to be translated to a foreign object.

@item type
A @cffi{} type specifier.

@item foreign-value
The primitive C value, ready to be passed to a primitive foreign
function.

@item alloc-params
Something of a translation state; you must pass it to
@code{free-converted-object} along with the foreign value for that to
work.
@end table

@subheading Description

This is an external interface to the type translation facility.  In
the implementation, all foreign functions are ultimately defined as
type translation wrappers around primitive foreign function
invocations.

This function is available mostly for inspection of the type
translation process, and possibly optimization of special cases of
your foreign function calls.

Its behavior is better described under @code{translate-to-foreign}'s
documentation.

@subheading Examples

@lisp
CFFI-USER> (convert-to-foreign t :boolean)
@result{} 1
@result{} (NIL)
CFFI-USER> (convert-to-foreign "hello, world" :string)
@result{} #<FOREIGN-ADDRESS #x097C5F80>
@result{} (T)
CFFI-USER> (code-char (mem-aref * :char 5))
@result{} #\,
@end lisp

@subheading See Also
@seealso{convert-from-foreign} @*
@seealso{free-converted-object} @*
@seealso{translate-to-foreign}


@c ===================================================================
@c DEFBITFIELD

@node defbitfield
@unnumberedsec defbitfield
@subheading Syntax
@Macro{defbitfield name-and-options &body masks}

masks ::= [docstring] @{ (symbol value) @}* @*
name-and-options ::= name | (name &optional (base-type :int))

@subheading Arguments and Values

@table @var
@item name
The name of the new bitfield type.

@item docstring
A documentation string, ignored.

@item base-type
A symbol denoting a foreign type.

@item symbol
A Lisp symbol.

@item value
An integer representing a bitmask.
@end table

@subheading Description
The @code{defbitfield} macro is used to define foreign types that map
lists of symbols to integer values.

If @var{value} is omitted, it will be computed as follows: find the
greatest @var{value} previously used, including those so computed,
with only a single 1-bit in its binary representation (that is, powers
of two), and left-shift it by one.  This rule guarantees that a
computed @var{value} cannot clash with previous values, but may clash
with future explicitly specified values.

Symbol lists will be automatically converted to values and vice versa
when being passed as arguments to or returned from foreign functions,
respectively. The same applies to any other situations where an object
of a bitfield type is expected.

Types defined with @code{defbitfield} canonicalize to @var{base-type}
which is @code{:int} by default.

@subheading Examples
@lisp
(defbitfield open-flags
  (:rdonly #x0000)
  :wronly               ;@lispcmt{#x0001}
  :rdwr                 ;@lispcmt{@dots{}}
  :nonblock
  :append
  (:creat    #x0200))
  ;; @lispcmt{etc@dots{}}

CFFI> (foreign-bitfield-symbols 'open-flags #b1101)
@result{} (:RDONLY :WRONLY :NONBLOCK :APPEND)

CFFI> (foreign-bitfield-value 'open-flags '(:rdwr :creat))
@result{} 514   ; #x0202

(defcfun ("open" unix-open) :int
  (path :string)
  (flags open-flags)
  (mode :uint16)) ; unportable

CFFI> (unix-open "/tmp/foo" '(:wronly :creat) #o644)
@result{} #<an fd>

;;; @lispcmt{Consider also the following lispier wrapper around open()}
(defun lispier-open (path mode &rest flags)
  (unix-open path flags mode))
@end lisp

@subheading See Also
@seealso{foreign-bitfield-value} @*
@seealso{foreign-bitfield-symbols}


@c ===================================================================
@c DEFCSTRUCT

@node defcstruct
@unnumberedsec defcstruct
@subheading Syntax
@Macro{defcstruct name-and-options &body doc-and-slots @result{} name}

name-and-options ::= structure-name | (structure-name &key size)

doc-and-slots ::= [docstring] @{ (slot-name slot-type &key count offset) @}*

@subheading Arguments and Values

@table @var
@item structure-name
The name of new structure type.

@item docstring
A documentation string, ignored.

@item slot-name
A symbol naming the slot.  It must be unique among slot names in this
structure.

@item size
Use this option to override the size (in bytes) of the struct.

@item slot-type
The type specifier for the slot.

@item count
Used to declare an array of size @var{count} inside the
structure.  Defaults to @code{1} as such an array and a single element
are semantically equivalent.

@item offset
Overrides the slot's offset. The next slot's offset is calculated
based on this one.
@end table

@subheading Description
This defines a new @cffi{} aggregate type akin to C @code{struct}s.
In other words, it specifies that foreign objects of the type
@var{structure-name} are groups of different pieces of data, or
``slots'', of the @var{slot-type}s, distinguished from each other by
the @var{slot-name}s.  Each structure is located in memory at a
position, and the slots are allocated sequentially beginning at that
point in memory (with some padding allowances as defined by the C
@acronym{ABI}, unless otherwise requested by specifying an
@var{offset} from the beginning of the structure (offset 0).

In other words, it is isomorphic to the C @code{struct}, giving
several extra features.

There are two kinds of slots, for the two kinds of @cffi{} types:

@table @dfn
@item Simple
Contain a single instance of a type that canonicalizes to a built-in
type, such as @code{:long} or @code{:pointer}.  Used for simple
@cffi{} types.

@item Aggregate
Contain an embedded structure or union, or an array of objects.  Used
for aggregate @cffi{} types.
@end table

The use of @acronym{CLOS} terminology for the structure-related
features is intentional; structure definitions are very much like
classes with (far) fewer features.

@subheading Examples
@lisp
(defcstruct point
  "Pointer structure."
  (x :int)
  (y :int))

CFFI> (with-foreign-object (ptr 'point)
        ;; @lispcmt{Initialize the slots}
        (setf (foreign-slot-value ptr 'point 'x) 42
              (foreign-slot-value ptr 'point 'y) 42)
        ;; @lispcmt{Return a list with the coordinates}
        (with-foreign-slots ((x y) ptr point)
          (list x y)))
@result{} (42 42)
@end lisp

@lisp
;; @lispcmt{Using the :size and :offset options to define a partial structure.}
;; @lispcmt{(this is useful when you are interested in only a few slots}
;; @lispcmt{of a big foreign structure)}

(defcstruct (foo :size 32)
  "Some struct with 32 bytes."
                        ; @lispcmt{<16 bytes we don't care about>}
  (x :int :offset 16)   ; @lispcmt{an int at offset 16}
  (y :int)              ; @lispcmt{another int at offset 16+sizeof(int)}
                        ; @lispcmt{<a couple more bytes we don't care about>}
  (z :char :offset 24)) ; @lispcmt{a char at offset 24}
                        ; @lispcmt{<7 more bytes ignored (since size is 32)>}

CFFI> (foreign-type-size 'foo)
@result{} 32
@end lisp

@lisp
;;; @lispcmt{Using :count to define arrays inside of a struct.}
(defcstruct video_tuner
  (name :char :count 32))
@end lisp

@subheading See Also
@seealso{foreign-slot-pointer} @*
@seealso{foreign-slot-value} @*
@seealso{with-foreign-slots}


@c ===================================================================
@c DEFCUNION

@node defcunion
@unnumberedsec defcunion
@subheading Syntax
@Macro{defcunion name &body doc-and-slots @result{} name}

doc-and-slots ::= [docstring] @{ (slot-name slot-type &key count) @}*

@subheading Arguments and Values

@table @var
@item name
The name of new union type.

@item docstring
A documentation string, ignored.

@item slot-name
A symbol naming the slot.

@item slot-type
The type specifier for the slot.

@item count
Used to declare an array of size @var{count} inside the
structure.
@end table

@subheading Description
A union is a structure in which all slots have an offset of zero.  It
is isomorphic to the C @code{union}.  Therefore, you should use the
usual foreign structure operations for accessing a union's slots.

@subheading Examples
@lisp
(defcunion uint32-bytes
  (int-value :unsigned-int)
  (bytes :unsigned-char :count 4))
@end lisp

@subheading See Also
@seealso{foreign-slot-pointer} @*
@seealso{foreign-slot-value}


@c ===================================================================
@c DEFCTYPE

@node defctype
@unnumberedsec defctype
@subheading Syntax
@Macro{defctype name base-type &optional documentation}

@subheading Arguments and Values

@table @var
@item name
The name of the new foreign type.

@item base-type
A symbol or a list defining the new type.

@item documentation
A documentation string, currently ignored.
@end table

@subheading Description
The @code{defctype} macro provides a mechanism similar to C's
@code{typedef} to define new types. The new type inherits
@var{base-type}'s translators, if any.

There is no way to define translations for types for types defined
with @code{defctype}.  For that, you should use
@ref{define-foreign-type}.

@subheading Examples
@lisp
(defctype my-string :string
  "My own string type.")

(defctype long-bools (:boolean :long)
  "Booleans that map to C longs.")
@end lisp

@subheading See Also
@seealso{define-foreign-type}


@c ===================================================================
@c DEFCENUM

@node defcenum
@unnumberedsec defcenum
@subheading Syntax
@Macro{defcenum name-and-options &body enum-list}

enum-list ::= [docstring] @{ keyword | (keyword value) @}*
name-and-options ::= name | (name &optional (base-type :int))

@subheading Arguments and Values

@table @var
@item name
The name of the new enum type.

@item docstring
A documentation string, ignored.

@item base-type
A symbol denoting a foreign type.

@item keyword
A keyword symbol.

@item value
An index value for a keyword.
@end table

@subheading Description
The @code{defcenum} macro is used to define foreign types that map
keyword symbols to integer values, similar to the C @code{enum} type.

If @var{value} is omitted its value will either be 0, if it's the
first entry, or it it will continue the progression from the last
specified value.

Keywords will be automatically converted to values and vice-versa when
being passed as arguments to or returned from foreign functions,
respectively. The same applies to any other situations where an object
of an @code{enum} type is expected.

Types defined with @code{defcenum} canonicalize to @var{base-type}
which is @code{:int} by default.

@subheading Examples
@lisp
(defcenum boolean
  :no
  :yes)

CFFI> (foreign-enum-value 'boolean :no)
@result{} 0
@end lisp

@lisp
(defcenum numbers
  (:one 1)
  :two
  (:four 4))

CFFI> (foreign-enum-keyword 'numbers 2)
@result{} :TWO
@end lisp

@subheading See Also
@seealso{foreign-enum-value} @*
@seealso{foreign-enum-keyword}


@c ===================================================================
@c DEFINE-FOREIGN-TYPE

@node define-foreign-type
@unnumberedsec define-foreign-type
@subheading Syntax
@Macro{define-foreign-type class-name supers slots &rest options @result{} class-name}

options ::= (@code{:actual-type} @var{type}) | @
    (@code{:simple-parser} @var{symbol}) | @
    @emph{regular defclass option}

@subheading Arguments and Values

@table @var
@item class-name
A symbol naming the new foreign type class.

@item supers
A list of symbols naming the super classes.

@item slots
A list of slot definitions, passed to @code{defclass}.
@end table

@subheading Description

@c TODO rewrite

The macro @code{define-foreign-type} defines a new class
@var{class-name}. It is a thin wrapper around @code{defclass}. Among
other things, it ensures that @var{class-name} becomes a subclass of
@var{foreign-type}, what you need to know about that is that there's
an initarg @code{:actual-type} which serves the same purpose as
@code{defctype}'s @var{base-type} argument.

@c TODO mention the type translators here
@c FIX FIX

@subheading Examples
Taken from @cffi{}'s @code{:boolean} type definition:

@lisp
(define-foreign-type :boolean (&optional (base-type :int))
  "Boolean type. Maps to an :int by default. Only accepts integer types."
  (ecase base-type
    ((:char
      :unsigned-char
      :int
      :unsigned-int
      :long
      :unsigned-long) base-type)))

CFFI> (canonicalize-foreign-type :boolean)
@result{} :INT
CFFI> (canonicalize-foreign-type '(:boolean :long))
@result{} :LONG
CFFI> (canonicalize-foreign-type '(:boolean :float))
;; @lispcmt{@error{} signalled by ECASE.}
@end lisp

@subheading See Also
@seealso{defctype} @*
@seealso{define-parse-method}


@c ===================================================================
@c DEFINE-PARSE-METHOD

@node define-parse-method
@unnumberedsec define-parse-method
@subheading Syntax
@Macro{define-parse-method name lambda-list &body body @result{} name}

@subheading Arguments and Values

@table @var
@item type-name
A symbol naming the new foreign type.

@item lambda-list
A lambda list which is the argument list of the new foreign type.

@item body
One or more forms that provide a definition of the new foreign type.
@end table

@subheading Description


@c TODO: update example. The boolean type is probably a good choice.

@subheading Examples
Taken from @cffi{}'s @code{:boolean} type definition:

@lisp
(define-foreign-type :boolean (&optional (base-type :int))
  "Boolean type. Maps to an :int by default. Only accepts integer types."
  (ecase base-type
    ((:char
      :unsigned-char
      :int
      :unsigned-int
      :long
      :unsigned-long) base-type)))

CFFI> (canonicalize-foreign-type :boolean)
@result{} :INT
CFFI> (canonicalize-foreign-type '(:boolean :long))
@result{} :LONG
CFFI> (canonicalize-foreign-type '(:boolean :float))
;; @lispcmt{@error{} signalled by ECASE.}
@end lisp

@subheading See Also
@seealso{define-foreign-type}


@c ===================================================================
@c EXPLAIN-FOREIGN-SLOT-VALUE

@c @node explain-foreign-slot-value
@c @unnumberedsec explain-foreign-slot-value
@c @subheading Syntax
@c @Macro{explain-foreign-slot-value ptr type &rest slot-names}

@c @subheading Arguments and Values

@c @table @var
@c @item ptr
@c ...

@c @item type
@c ...

@c @item slot-names
@c ...
@c @end table

@c @subheading Description
@c This macro translates the slot access that would occur by calling
@c @code{foreign-slot-value} with the same arguments into an equivalent
@c expression in C and prints it to @code{*standard-output*}.

@c @emph{Note: this is not implemented yet.}

@c @subheading Examples
@c @lisp
@c CFFI> (explain-foreign-slot-value ptr 'timeval 'tv-secs)
@c @result{} ptr->tv_secs

@c CFFI> (explain-foreign-slot-value emp 'employee 'hire-date 'tv-usecs)
@c @result{} emp->hire_date.tv_usecs
@c @end lisp

@c @subheading See Also


@c ===================================================================
@c FOREIGN-BITFIELD-SYMBOLS

@node foreign-bitfield-symbols
@unnumberedsec foreign-bitfield-symbols
@subheading Syntax
@Function{foreign-bitfield-symbols type value @result{} symbols}

@subheading Arguments and Values

@table @var
@item type
A bitfield type.

@item value
An integer.

@item symbols
A potentially shared list of symbols.
@code{nil}.
@end table

@subheading Description
The function @code{foreign-bitfield-symbols} returns a possibly shared
list of symbols that correspond to @var{value} in @var{type}.

@subheading Examples
@lisp
(defbitfield flags
  (flag-a 1)
  (flag-b 2)
  (flag-c 4))

CFFI> (foreign-bitfield-symbols 'boolean #b101)
@result{} (FLAG-A FLAG-C)
@end lisp

@subheading See Also
@seealso{defbitfield} @*
@seealso{foreign-bitfield-value}


@c ===================================================================
@c FOREIGN-BITFIELD-VALUE

@node foreign-bitfield-value
@unnumberedsec foreign-bitfield-value
@subheading Syntax
@Function{foreign-bitfield-value type symbols @result{} value}

@subheading Arguments and Values

@table @var
@item type
A @code{bitfield} type.

@item symbol
A Lisp symbol.

@item value
An integer.
@end table

@subheading Description
The function @code{foreign-bitfield-value} returns the @var{value} that
corresponds to the symbols in the @var{symbols} list.

@subheading Examples
@lisp
(defbitfield flags
  (flag-a 1)
  (flag-b 2)
  (flag-c 4))

CFFI> (foreign-bitfield-value 'flags '(flag-a flag-c))
@result{} 5  ; #b101
@end lisp

@subheading See Also
@seealso{defbitfield} @*
@seealso{foreign-bitfield-symbols}


@c ===================================================================
@c FOREIGN-ENUM-KEYWORD

@node foreign-enum-keyword
@unnumberedsec foreign-enum-keyword
@subheading Syntax
@Function{foreign-enum-keyword type value &key errorp @result{} keyword}

@subheading Arguments and Values

@table @var
@item type
An @code{enum} type.

@item value
An integer.

@item errorp
If true (the default), signal an error if @var{value} is not defined
in @var{type}.  If false, @code{foreign-enum-keyword} returns
@code{nil}.

@item keyword
A keyword symbol.
@end table

@subheading Description
The function @code{foreign-enum-keyword} returns the keyword symbol
that corresponds to @var{value} in @var{type}.

An error is signaled if @var{type} doesn't contain such @var{value}
and @var{errorp} is true.

@subheading Examples
@lisp
(defcenum boolean
  :no
  :yes)

CFFI> (foreign-enum-keyword 'boolean 1)
@result{} :YES
@end lisp

@subheading See Also
@seealso{defcenum} @*
@seealso{foreign-enum-value}


@c ===================================================================
@c FOREIGN-ENUM-VALUE

@node foreign-enum-value
@unnumberedsec foreign-enum-value
@subheading Syntax
@Function{foreign-enum-value type keyword &key errorp @result{} value}

@subheading Arguments and Values

@table @var
@item type
An @code{enum} type.

@item keyword
A keyword symbol.

@item errorp
If true (the default), signal an error if @var{keyword} is not
defined in @var{type}.  If false, @code{foreign-enum-value} returns
@code{nil}.

@item value
An integer.
@end table

@subheading Description
The function @code{foreign-enum-value} returns the @var{value} that
corresponds to @var{keyword} in @var{type}.

An error is signaled if @var{type} doesn't contain such
@var{keyword}, and @var{errorp} is true.

@subheading Examples
@lisp
(defcenum boolean
  :no
  :yes)

CFFI> (foreign-enum-value 'boolean :yes)
@result{} 1
@end lisp

@subheading See Also
@seealso{defcenum} @*
@seealso{foreign-enum-keyword}


@c ===================================================================
@c FOREIGN-SLOT-NAMES

@node foreign-slot-names
@unnumberedsec foreign-slot-names
@subheading Syntax
@Function{foreign-slot-names type @result{} names}

@subheading Arguments and Values

@table @var
@item type
A foreign struct type.

@item names
A list.
@end table

@subheading Description
The function @code{foreign-slot-names} returns a potentially shared
list of slot @var{names} for the given structure @var{type}. This list
has no particular order.

@subheading Examples
@lisp
(defcstruct timeval
  (tv-secs :long)
  (tv-usecs :long))

CFFI> (foreign-slot-names 'timeval)
@result{} (TV-SECS TV-USECS)
@end lisp

@subheading See Also
@seealso{defcstruct} @*
@seealso{foreign-slot-offset} @*
@seealso{foreign-slot-value} @*
@seealso{foreign-slot-pointer}


@c ===================================================================
@c FOREIGN-SLOT-OFFSET

@node foreign-slot-offset
@unnumberedsec foreign-slot-offset
@subheading Syntax
@Function{foreign-slot-offset type slot-name @result{} offset}

@subheading Arguments and Values

@table @var
@item type
A foreign struct type.

@item slot-name
A symbol.

@item offset
An integer.
@end table

@subheading Description
The function @code{foreign-slot-offset} returns the @var{offset} in
bytes of a slot in a foreign struct type.

@subheading Examples
@lisp
(defcstruct timeval
  (tv-secs :long)
  (tv-usecs :long))

CFFI> (foreign-slot-offset 'timeval 'tv-secs)
@result{} 0
CFFI> (foreign-slot-offset 'timeval 'tv-usecs)
@result{} 4
@end lisp

@subheading See Also
@seealso{defcstruct} @*
@seealso{foreign-slot-names} @*
@seealso{foreign-slot-pointer} @*
@seealso{foreign-slot-value}


@c ===================================================================
@c FOREIGN-SLOT-POINTER

@node foreign-slot-pointer
@unnumberedsec foreign-slot-pointer
@subheading Syntax
@Function{foreign-slot-pointer ptr type slot-name @result{} pointer}

@subheading Arguments and Values

@table @var
@item ptr
A pointer to a structure.

@item type
A foreign structure type.

@item slot-names
A slot name in the @var{type}.

@item pointer
A pointer to the slot @var{slot-name}.
@end table

@subheading Description
Returns a pointer to the location of the slot @var{slot-name} in a
foreign object of type @var{type} at @var{ptr}. The returned pointer
points inside the structure. Both the pointer and the memory it points
to have the same extent as @var{ptr}.

For aggregate slots, this is the same value returned by
@code{foreign-slot-value}.

@subheading Examples
@lisp
(defcstruct point
  "Pointer structure."
  (x :int)
  (y :int))

CFFI> (with-foreign-object (ptr 'point)
        (foreign-slot-pointer ptr 'point 'x))
@result{} #<FOREIGN-ADDRESS #xBFFF6E60>
;; @lispcmt{Note: the exact pointer representation varies from lisp to lisp.}
@end lisp

@subheading See Also
@seealso{defcstruct} @*
@seealso{foreign-slot-value} @*
@seealso{foreign-slot-names} @*
@seealso{foreign-slot-offset}


@c ===================================================================
@c FOREIGN-SLOT-VALUE

@node foreign-slot-value
@unnumberedsec foreign-slot-value
@subheading Syntax
@Accessor{foreign-slot-value ptr type slot-name @result{} object}

@subheading Arguments and Values

@table @var
@item ptr
A pointer to a structure.

@item type
A foreign structure type.

@item slot-name
A symbol naming a slot in the structure type.

@item object
The object contained in the slot specified by @var{slot-name}.
@end table

@subheading Description
For simple slots, @code{foreign-slot-value} returns the value of the
object, such as a Lisp integer or pointer. In C, this would be
expressed as @code{ptr->slot}.

For aggregate slots, a pointer inside the structure to the beginning
of the slot's data is returned. In C, this would be expressed as
@code{&ptr->slot}. This pointer and the memory it points to have the
same extent as @var{ptr}.

There are compiler macros for @code{foreign-slot-value} and its
@code{setf} expansion that open code the memory access when 
@var{type} and @var{slot-names} are constant at compile-time.

@subheading Examples
@lisp
(defcstruct point
  "Pointer structure."
  (x :int)
  (y :int))

CFFI> (with-foreign-object (ptr 'point)
        ;; @lispcmt{Initialize the slots}
        (setf (foreign-slot-value ptr 'point 'x) 42
              (foreign-slot-value ptr 'point 'y) 42)
        ;; @lispcmt{Return a list with the coordinates}
        (with-foreign-slots ((x y) ptr point)
          (list x y)))
@result{} (42 42)
@end lisp

@subheading See Also
@seealso{defcstruct} @*
@seealso{foreign-slot-names} @*
@seealso{foreign-slot-offset} @*
@seealso{foreign-slot-pointer} @*
@seealso{with-foreign-slots}


@c ===================================================================
@c FOREIGN-TYPE-ALIGNMENT

@node foreign-type-alignment
@unnumberedsec foreign-type-alignment
@subheading Syntax
@c XXX: This is actually a generic function.
@Function{foreign-type-alignment type @result{} alignment}

@subheading Arguments and Values

@table @var
@item type
A foreign type.

@item alignment
An integer.
@end table

@subheading Description
The function @code{foreign-type-alignment} returns the
@var{alignment} of @var{type} in bytes.

@subheading Examples
@lisp
CFFI> (foreign-type-alignment :char)
@result{} 1
CFFI> (foreign-type-alignment :short)
@result{} 2
CFFI> (foreign-type-alignment :int)
@result{} 4
@end lisp

@lisp
(defcstruct foo
  (a :char))

CFFI> (foreign-type-alignment 'foo)
@result{} 1
@end lisp

@subheading See Also
@seealso{foreign-type-size}


@c ===================================================================
@c FOREIGN-TYPE-SIZE

@node foreign-type-size
@unnumberedsec foreign-type-size
@subheading Syntax
@c XXX: this is actually a generic function.
@Function{foreign-type-size type @result{} size}

@subheading Arguments and Values

@table @var
@item type
A foreign type.

@item size
An integer.
@end table

@subheading Description
The function @code{foreign-type-size} return the @var{size} of
@var{type} in bytes.  This includes any padding within and following
the in-memory representation as needed to create an array of
@var{type} objects.

@subheading Examples
@lisp
(defcstruct foo
  (a :double)
  (c :char))

CFFI> (foreign-type-size :double)
@result{} 8
CFFI> (foreign-type-size :char)
@result{} 1
CFFI> (foreign-type-size 'foo)
@result{} 16
@end lisp

@subheading See Also
@seealso{foreign-type-alignment}


@c ===================================================================
@c FREE-CONVERTED-OBJECT

@node free-converted-object
@unnumberedsec free-converted-object
@subheading Syntax
@Function{free-converted-object foreign-value type params}

@subheading Arguments and Values

@table @var
@item foreign-value
The C object to be freed.

@item type
A @cffi{} type specifier.

@item params
The state returned as the second value from @code{convert-to-foreign};
used to implement the third argument to @code{free-translated-object}.
@end table

@subheading Description

The return value is unspecified.

This is an external interface to the type translation facility.  In
the implementation, all foreign functions are ultimately defined as
type translation wrappers around primitive foreign function
invocations.

This function is available mostly for inspection of the type
translation process, and possibly optimization of special cases of
your foreign function calls.

Its behavior is better described under @code{free-translated-object}'s
documentation.

@subheading Examples

@lisp
CFFI-USER> (convert-to-foreign "a boat" :string)
@result{} #<FOREIGN-ADDRESS #x097ACDC0>
@result{} (T)
CFFI-USER> (free-converted-object * :string '(t))
@result{} NIL
@end lisp

@subheading See Also
@seealso{convert-from-foreign} @*
@seealso{convert-to-foreign} @*
@seealso{free-translated-object}


@c ===================================================================
@c FREE-TRANSLATED-OBJECT

@c TODO: update

@node free-translated-object
@unnumberedsec free-translated-object
@subheading Syntax
@GenericFunction{free-translated-object value type-name param}

@subheading Arguments and Values

@table @var
@item pointer
The foreign value returned by @code{translate-to-foreign}.

@item type-name
A symbol naming a foreign type defined by @code{defctype}.

@item param
The second value, if any, returned by @code{translate-to-foreign}.
@end table

@subheading Description
This generic function may be specialized by user code to perform
automatic deallocation of foreign objects as they are passed to C
functions.

Any methods defined on this generic function must EQL-specialize the
@var{type-name} parameter on a symbol defined as a foreign type by
the @code{defctype} macro.

@subheading See Also
@seealso{Foreign Type Translators} @*
@seealso{translate-to-foreign}


@c ===================================================================
@c TRANSLATE-FROM-FOREIGN

@c TODO: update

@node translate-from-foreign
@unnumberedsec translate-from-foreign
@subheading Syntax
@GenericFunction{translate-from-foreign foreign-value type-name @
                                        @result{} lisp-value}

@subheading Arguments and Values

@table @var
@item foreign-value
The foreign value to convert to a Lisp object.

@item type-name
A symbol naming a foreign type defined by @code{defctype}.

@item lisp-value
The lisp value to pass in place of @code{foreign-value} to Lisp code.
@end table

@subheading Description
This generic function is invoked by @cffi{} to convert a foreign value to
a Lisp value, such as when returning from a foreign function, passing
arguments to a callback function, or accessing a foreign variable.

To extend the @cffi{} type system by performing custom translations, this
method may be specialized by @sc{eql}-specializing @code{type-name} on a
symbol naming a foreign type defined with @code{defctype}.  This
method should return the appropriate Lisp value to use in place of the
foreign value.

The results are undefined if the @code{type-name} parameter is
specialized in any way except an @sc{eql} specializer on a foreign type
defined with @code{defctype}.  Specifically, translations may not be
defined for built-in types.

@subheading See Also
@seealso{Foreign Type Translators} @*
@seealso{translate-to-foreign} @*
@seealso{free-translated-object}


@c ===================================================================
@c TRANSLATE-TO-FOREIGN

@c TODO: update

@node translate-to-foreign
@unnumberedsec translate-to-foreign
@subheading Syntax
@GenericFunction{translate-to-foreign lisp-value type-name @
                                      @result{} foreign-value, alloc-param}

@subheading Arguments and Values

@table @var
@item lisp-value
The Lisp value to convert to foreign representation.

@item type-name
A symbol naming a foreign type defined by @code{defctype}.

@item foreign-value
The foreign value to pass in place of @code{lisp-value} to foreign code.

@item alloc-param
If present, this value will be passed to
@code{free-translated-object}.
@end table

@subheading Description
This generic function is invoked by @cffi{} to convert a Lisp value to a
foreign value, such as when passing arguments to a foreign function,
returning a value from a callback, or setting a foreign variable.  A
``foreign value'' is one appropriate for passing to the next-lowest
translator, including the low-level translators that are ultimately
invoked invisibly with @cffi{}.

To extend the @cffi{} type system by performing custom translations, this
method may be specialized by @sc{eql}-specializing @code{type-name} on a
symbol naming a foreign type defined with @code{defctype}.  This
method should return the appropriate foreign value to use in place of
the Lisp value.

In cases where @cffi{} can determine the lifetime of the foreign object
returned by this method, it will invoke @code{free-translated-object}
on the foreign object at the appropriate time.  If
@code{translate-to-foreign} returns a second value, it will be passed
as the @code{param} argument to @code{free-translated-object}.  This
can be used to establish communication between the allocation and
deallocation methods.

The results are undefined if the @code{type-name} parameter is
specialized in any way except an @sc{eql} specializer on a foreign type
defined with @code{defctype}.  Specifically, translations may not be
defined for built-in types.

@subheading See Also
@seealso{Foreign Type Translators} @*
@seealso{translate-from-foreign} @*
@seealso{free-translated-object}


@c ===================================================================
@c WITH-FOREIGN-SLOTS

@node with-foreign-slots
@unnumberedsec with-foreign-slots
@subheading Syntax
@Macro{with-foreign-slots (vars ptr type) &body body}

@subheading Arguments and Values

@table @var
@item vars
A list of symbols.

@item ptr
A foreign pointer to a structure.

@item type
A structure type.

@item body
A list of forms to be executed.
@end table

@subheading Description
The @code{with-foreign-slots} macro creates local symbol macros for
each var in @var{vars} to reference foreign slots in @var{ptr} of
@var{type}.  It is similar to Common Lisp's @code{with-slots} macro.

@subheading Examples
@lisp
(defcstruct tm
  (sec :int)
  (min :int)
  (hour :int)
  (mday :int)
  (mon  :int)
  (year :int)
  (wday :int)
  (yday :int)
  (isdst  :boolean)
  (zone   :string)
  (gmtoff :long))

CFFI> (with-foreign-object (time :int)
        (setf (mem-ref time :int)
              (foreign-funcall "time" :pointer (null-pointer) :int))
        (foreign-funcall "gmtime" :pointer time tm))
@result{} #<A Mac Pointer #x102A30>
CFFI> (with-foreign-slots ((sec min hour mday mon year) * tm)
        (format nil "~A:~A:~A, ~A/~A/~A"
                hour min sec (+ 1900 year) mon mday))
@result{} "7:22:47, 2005/8/2"
@end lisp

@subheading See Also
@seealso{defcstruct} @*
@seealso{defcunion} @*
@seealso{foreign-slot-value}


@c ===================================================================
@c CHAPTER: Pointers

@node Pointers
@chapter Pointers

All C data in @cffi{} is referenced through pointers.  This includes
defined C variables that hold immediate values, and integers.

To see why this is, consider the case of the C integer.  It is not
only an arbitrary representation for an integer, congruent to Lisp's
fixnums; the C integer has a specific bit pattern in memory defined by
the C @acronym{ABI}.  Lisp has no such constraint on its fixnums;
therefore, it only makes sense to think of fixnums as C integers if
you assume that @cffi{} converts them when necessary, such as when
storing one for use in a C function call, or as the value of a C
variable.  This requires defining an area of memory@footnote{The
definition of @dfn{memory} includes the @acronym{CPU} registers.},
represented through an effective address, and storing it there.

Due to this compartmentalization, it only makes sense to manipulate
raw C data in Lisp through pointers to it.  For example, while there
may be a Lisp representation of a @code{struct} that is converted to C
at store time, you may only manipulate its raw data through a pointer.
The C compiler does this also, albeit informally.

@menu
* Basic Pointer Operations::    
* Allocating Foreign Memory::   
* Accessing Foreign Memory::    

Dictionary

* foreign-free::                
* foreign-alloc::               
* foreign-symbol-pointer::      
* inc-pointer::                 
* incf-pointer::                
* make-pointer::                
* mem-aref::                    
* mem-ref::                     
* null-pointer::                
* null-pointer-p::              
* pointerp::                    
* pointer-address::             
* pointer-eq::                  
* with-foreign-object::         
* with-foreign-pointer::        
@end menu

@node Basic Pointer Operations
@section Basic Pointer Operations

Manipulating pointers proper can be accomplished through most of the
other operations defined in the Pointers dictionary, such as
@code{make-pointer}, @code{pointer-address}, and @code{pointer-eq}.
When using them, keep in mind that they merely manipulate the Lisp
representation of pointers, not the values they point to.

@deftp {Lisp Type} foreign-pointer
The pointers' representations differ from implementation to
implementation and have different types.  @code{foreign-pointer}
provides a portable type alias to each of these types.
@end deftp


@node Allocating Foreign Memory
@section Allocating Foreign Memory

@cffi{} provides support for stack and heap C memory allocation.
Stack allocation, done with @code{with-foreign-object}, is sometimes
called ``dynamic'' allocation in Lisp, because memory allocated as
such has dynamic extent, much as with @code{let} bindings of special
variables.

This should not be confused with what C calls ``dynamic'' allocation,
or that done with @code{malloc} and friends.  This sort of heap
allocation is done with @code{foreign-alloc}, creating objects that
exist until freed with @code{foreign-free}.


@node Accessing Foreign Memory
@section Accessing Foreign Memory

When manipulating raw C data, consider that all pointers are pointing
to an array.  When you only want one C value, such as a single
@code{struct}, this array only has one such value.  It is worthwhile
to remember that everything is an array, though, because this is also
the semantic that C imposes natively.

C values are accessed as the @code{setf}-able places defined by
@code{mem-aref} and @code{mem-ref}.  Given a pointer and a @cffi{}
type (@pxref{Foreign Types}), either of these will dereference the
pointer, translate the C data there back to Lisp, and return the
result of said translation, performing the reverse operation when
@code{setf}-ing.  To decide which one to use, consider whether you
would use the array index operator @code{[@var{n}]} or the pointer
dereference @code{*} in C; use @code{mem-aref} for array indexing and
@code{mem-ref} for pointer dereferencing.


@c ===================================================================
@c FOREIGN-FREE

@node foreign-free
@unnumberedsec foreign-free
@subheading Syntax
@Function{foreign-free ptr @result{} undefined}

@subheading Arguments and Values

@table @var
@item ptr
A foreign pointer.
@end table

@subheading Description
The @code{foreign-free} function frees a @code{ptr} previously
allocated by @code{foreign-alloc}. The consequences of freeing a given
pointer twice are undefined.

@subheading Examples

@lisp
CFFI> (foreign-alloc :int)
@result{} #<A Mac Pointer #x1022E0>
CFFI> (foreign-free *)
@result{} NIL
@end lisp

@subheading See Also
@seealso{foreign-alloc} @*
@seealso{with-foreign-pointer}


@c ===================================================================
@c FOREIGN-ALLOC

@node foreign-alloc
@unnumberedsec foreign-alloc
@subheading Syntax
@Function{foreign-alloc type &key initial-element initial-contents (count 1) @
                        null-terminated-p @result{} pointer}

@subheading Arguments and Values

@table @var
@item type
A foreign type.

@item initial-element
A Lisp object.

@item initial-contents
A sequence.

@item count
An integer. Defaults to 1 or the length of @var{initial-contents} if
supplied.

@item null-terminated-p
A boolean, false by default.

@item pointer
A foreign pointer to the newly allocated memory.
@end table

@subheading Description
The @code{foreign-alloc} function allocates enough memory to hold
@var{count} objects of type @var{type} and returns a
@var{pointer}. This memory must be explicitly freed using
@code{foreign-free} once it is no longer needed.

If @var{initial-element} is supplied, it is used to initialize the
@var{count} objects the newly allocated memory holds.

If an @var{initial-contents} sequence is supplied, it must have a
length less than or equal to @var{count} and each of its elements
will be used to initialize the contents of the newly allocated
memory.

If @var{count} is omitted and @var{initial-contents} is specified, it
will default to @code{(length @var{initial-contents})}.

@var{initial-element} and @var{initial-contents} are mutually
exclusive.

When @var{null-terminated-p} is true,
@code{(1+ (max @var{count} (length @var{initial-contents})))} elements
are allocated and the last one is set to @code{NULL}. Note that in
this case @var{type} must be a pointer type (ie. a type that
canonicalizes to @code{:pointer}), otherwise an error is signaled.

@subheading Examples
@lisp
CFFI> (foreign-alloc :char)
@result{} #<A Mac Pointer #x102D80>     ; @lispcmt{A pointer to 1 byte of memory.}

CFFI> (foreign-alloc :char :count 20)
@result{} #<A Mac Pointer #x1024A0>     ; @lispcmt{A pointer to 20 bytes of memory.}

CFFI> (foreign-alloc :int :initial-element 12)
@result{} #<A Mac Pointer #x1028B0>
CFFI> (mem-ref * :int)
@result{} 12

CFFI> (foreign-alloc :int :initial-contents '(1 2 3))
@result{} #<A Mac Pointer #x102950>
CFFI> (loop for i from 0 below 3
            collect (mem-aref * :int i))
@result{} (1 2 3)

CFFI> (foreign-alloc :int :initial-contents #(1 2 3))
@result{} #<A Mac Pointer #x102960>
CFFI> (loop for i from 0 below 3
            collect (mem-aref * :int i))
@result{} (1 2 3)

;;; Allocate a char** pointer that points to newly allocated memory
;;; by the :string type translator for the string "foo".
CFFI> (foreign-alloc :string :initial-element "foo")
@result{} #<A Mac Pointer #x102C40>
@end lisp

@lisp
;;; Allocate a null-terminated array of strings.
;;; (Note: FOREIGN-STRING-TO-LISP returns NIL when passed a null pointer)
CFFI> (foreign-alloc :string
                     :initial-contents '("foo" "bar" "baz")
                     :null-terminated-p t)
@result{} #<A Mac Pointer #x102D20>
CFFI> (loop for i from 0 below 4
            collect (mem-aref * :string i))
@result{} ("foo" "bar" "baz" NIL)
CFFI> (progn
        (dotimes (i 3)
          (foreign-free (mem-aref ** :pointer i)))
        (foreign-free **))
@result{} nil
@end lisp

@subheading See Also
@seealso{foreign-free} @*
@seealso{with-foreign-object} @*
@seealso{with-foreign-pointer}


@c ===================================================================
@c FOREIGN-SYMBOL-POINTER

@node foreign-symbol-pointer
@unnumberedsec foreign-symbol-pointer
@subheading Syntax
@Function{foreign-symbol-pointer foreign-name &key library @result{} pointer}

@subheading Arguments and Values

@table @var
@item foreign-name
A string.

@item pointer
A foreign pointer, or @code{nil}.

@item library
A Lisp symbol or an instance of @code{foreign-library}.
@end table

@subheading Description
The function @code{foreign-symbol-pointer} will return a foreign
pointer corresponding to the foreign symbol denoted by the string
@var{foreign-name}.  If a foreign symbol named @var{foreign-name}
doesn't exist, @code{nil} is returned.

ABI name manglings will be performed on @var{foreign-name} by
@code{foreign-symbol-pointer} if necessary. (eg: adding a leading
underscore on darwin/ppc)

@var{library} should name a foreign library as defined by
@code{define-foreign-library}, @code{:default} (which is the default)
or an instance of @code{foreign-library} as returned by
@code{load-foreign-library}.

@strong{Important note:} do not keep these pointers across saved Lisp
cores as the foreign-library may move across sessions.

@subheading Examples

@lisp
CFFI> (foreign-symbol-pointer "errno")
@result{} #<A Mac Pointer #xA0008130>
CFFI> (foreign-symbol-pointer "strerror")
@result{} #<A Mac Pointer #x9002D0F8>
CFFI> (foreign-funcall-pointer * () :int (mem-ref ** :int) :string)
@result{} "No such file or directory"

CFFI> (foreign-symbol-pointer "inexistent symbol")
@result{} NIL
@end lisp

@subheading See Also
@seealso{defcvar}


@c ===================================================================
@c INC-POINTER

@node inc-pointer
@unnumberedsec inc-pointer
@subheading Syntax
@Function{inc-pointer pointer offset @result{} new-pointer}

@subheading Arguments and Values

@table @var
@item pointer
@itemx new-pointer
A foreign pointer.

@item offset
An integer.
@end table

@subheading Description
The function @code{inc-pointer} will return a @var{new-pointer} pointing
@var{offset} bytes past @var{pointer}.

@subheading Examples

@lisp
CFFI> (foreign-string-alloc "Common Lisp")
@result{} #<A Mac Pointer #x102EA0>
CFFI> (inc-pointer * 7)
@result{} #<A Mac Pointer #x102EA7>
CFFI> (foreign-string-to-lisp *)
@result{} "Lisp"
@end lisp

@subheading See Also
@seealso{incf-pointer} @*
@seealso{make-pointer} @*
@seealso{pointerp} @*
@seealso{null-pointer} @*
@seealso{null-pointer-p}


@c ===================================================================
@c INCF-POINTER

@node incf-pointer
@unnumberedsec inc-pointer
@subheading Syntax
@Macro{incf-pointer place &optional (offset 1) @result{} new-pointer}

@subheading Arguments and Values

@table @var
@item place
A @code{setf} place.

@item new-pointer
A foreign pointer.

@item offset
An integer.
@end table

@subheading Description
The @code{incf-pointer} macro takes the foreign pointer from
@var{place} and creates a @var{new-pointer} incremented by
@var{offset} bytes and which is stored in @var{place}.

@subheading Examples

@lisp
CFFI> (defparameter *two-words* (foreign-string-alloc "Common Lisp"))
@result{} *TWO-WORDS*
CFFI> (defparameter *one-word* *two-words*)
@result{} *ONE-WORD*
CFFI> (incf-pointer *one-word* 7)
@result{} #.(SB-SYS:INT-SAP #X00600457)
CFFI> (foreign-string-to-lisp *one-word*)
@result{} "Lisp"
CFFI> (foreign-string-to-lisp *two-words*)
@result{} "Common Lisp"
@end lisp

@subheading See Also
@seealso{inc-pointer} @*
@seealso{make-pointer} @*
@seealso{pointerp} @*
@seealso{null-pointer} @*
@seealso{null-pointer-p}


@c ===================================================================
@c MAKE-POINTER

@node make-pointer
@unnumberedsec make-pointer
@subheading Syntax
@Function{make-pointer address @result{} ptr}

@subheading Arguments and Values

@table @var
@item address
An integer.

@item ptr
A foreign pointer.
@end table

@subheading Description
The function @code{make-pointer} will return a foreign pointer
pointing to @var{address}.

@subheading Examples

@lisp
CFFI> (make-pointer 42)
@result{} #<FOREIGN-ADDRESS #x0000002A>
CFFI> (pointerp *)
@result{} T
CFFI> (pointer-address **)
@result{} 42
CFFI> (inc-pointer *** -42)
@result{} #<FOREIGN-ADDRESS #x00000000>
CFFI> (null-pointer-p *)
@result{} T
CFFI> (typep ** 'foreign-pointer)
@result{} T
@end lisp

@subheading See Also
@seealso{inc-pointer} @*
@seealso{null-pointer} @*
@seealso{null-pointer-p} @*
@seealso{pointerp} @*
@seealso{pointer-address} @*
@seealso{pointer-eq} @*
@seealso{mem-ref}


@c ===================================================================
@c MEM-AREF

@node mem-aref
@unnumberedsec mem-aref
@subheading Syntax
@Accessor{mem-aref ptr type &optional (index 0)}

(setf (@strong{mem-aref} @emph{ptr type &optional (index 0)) new-value})

@subheading Arguments and Values

@table @var
@item ptr
A foreign pointer.

@item type
A foreign type.

@item index
An integer.

@item new-value
A Lisp value compatible with @var{type}.
@end table

@subheading Description
The @code{mem-aref} function is similar to @code{mem-ref} but will
automatically calculate the offset from an @var{index}.

@lisp
(mem-aref ptr type n)

;; @lispcmt{is identical to:}

(mem-ref ptr type (* n (foreign-type-size type)))
@end lisp

@subheading Examples

@lisp
CFFI> (with-foreign-string (str "Hello, foreign world!")
        (mem-aref str :char 6))
@result{} 32
CFFI> (code-char *)
@result{} #\Space

CFFI> (with-foreign-object (array :int 10)
        (loop for i below 10
              do (setf (mem-aref array :int i) (random 100)))
        (loop for i below 10 collect (mem-aref array :int i)))
@result{} (22 7 22 52 69 1 46 93 90 65)
@end lisp

@subheading See Also
@seealso{mem-ref}


@c ===================================================================
@c MEM-REF

@node mem-ref
@unnumberedsec mem-ref
@subheading Syntax
@Accessor{mem-ref ptr type &optional offset @result{} object}

@subheading Arguments and Values

@table @var
@item ptr
A pointer.

@item type
A foreign type.

@item offset
An integer (in byte units).

@item object
The value @var{ptr} points to.
@end table

@subheading Description
@subheading Examples

@lisp
CFFI> (with-foreign-string (ptr "Saluton")
        (setf (mem-ref ptr :char 3) (char-code #\a))
        (loop for i from 0 below 8
              collect (code-char (mem-ref ptr :char i))))
@result{} (#\S #\a #\l #\a #\t #\o #\n #\Null)
CFFI> (setq ptr-to-int (foreign-alloc :int))
@result{} #<A Mac Pointer #x1047D0>
CFFI> (mem-ref ptr-to-int :int)
@result{} 1054619
CFFI> (setf (mem-ref ptr-to-int :int) 1984)
@result{} 1984
CFFI> (mem-ref ptr-to-int :int)
@result{} 1984
@end lisp

@subheading See Also
@seealso{mem-aref}


@c ===================================================================
@c NULL-POINTER

@node null-pointer
@unnumberedsec null-pointer
@subheading Syntax
@Function{null-pointer @result{} pointer}

@subheading Arguments and Values

@table @var
@item pointer
A @code{NULL} pointer.
@end table

@subheading Description
The function @code{null-pointer} returns a null pointer.

@subheading Examples

@lisp
CFFI> (null-pointer)
@result{} #<A Null Mac Pointer>
CFFI> (pointerp *)
@result{} T
@end lisp

@subheading See Also
@seealso{null-pointer-p} @*
@seealso{make-pointer}


@c ===================================================================
@c NULL-POINTER-P

@node null-pointer-p
@unnumberedsec null-pointer-p
@subheading Syntax
@Function{null-pointer-p ptr @result{} boolean}

@subheading Arguments and Values

@table @var
@item ptr
A foreign pointer that may be a null pointer.

@item boolean
@code{T} or @code{NIL}.
@end table

@subheading Description
The function @code{null-pointer-p} returns true if @var{ptr} is a null
pointer and false otherwise.

@subheading Examples

@lisp
CFFI> (null-pointer-p (null-pointer))
@result{} T
@end lisp

@lisp
(defun contains-str-p (big little)
  (not (null-pointer-p
        (foreign-funcall "strstr" :string big :string little :pointer))))

CFFI> (contains-str-p "Popcorns" "corn")
@result{} T
CFFI> (contains-str-p "Popcorns" "salt")
@result{} NIL
@end lisp

@subheading See Also
@seealso{null-pointer} @*
@seealso{pointerp}


@c ===================================================================
@c POINTERP

@node pointerp
@unnumberedsec pointerp
@subheading Syntax
@Function{pointerp ptr @result{} boolean}

@subheading Arguments and Values

@table @var
@item ptr
An object that may be a foreign pointer.

@item boolean
@code{T} or @code{NIL}.
@end table

@subheading Description
The function @code{pointerp} returns true if @var{ptr} is a foreign
pointer and false otherwise.

@subheading Implementation-specific Notes
In Allegro CL, foreign pointers are integers thus in this
implementation @code{pointerp} will return true for any ordinary integer.

@subheading Examples

@lisp
CFFI> (foreign-alloc 32)
@result{} #<A Mac Pointer #x102D20>
CFFI> (pointerp *)
@result{} T
CFFI> (pointerp "this is not a pointer")
@result{} NIL
@end lisp

@subheading See Also
@seealso{make-pointer}
@seealso{null-pointer-p}


@c ===================================================================
@c POINTER-ADDRESS

@node pointer-address
@unnumberedsec pointer-address
@subheading Syntax
@Function{pointer-address ptr @result{} address}

@subheading Arguments and Values

@table @var
@item ptr
A foreign pointer.

@item address
An integer.
@end table

@subheading Description
The function @code{pointer-address} will return the @var{address} of
a foreign pointer @var{ptr}.

@subheading Examples

@lisp
CFFI> (pointer-address (null-pointer))
@result{} 0
CFFI> (pointer-address (make-pointer 123))
@result{} 123
@end lisp

@subheading See Also
@seealso{make-pointer} @*
@seealso{inc-pointer} @*
@seealso{null-pointer} @*
@seealso{null-pointer-p} @*
@seealso{pointerp} @*
@seealso{pointer-eq} @*
@seealso{mem-ref}


@c ===================================================================
@c POINTER-EQ

@node pointer-eq
@unnumberedsec pointer-eq
@subheading Syntax
@Function{pointer-eq ptr1 ptr2 @result{} boolean}

@subheading Arguments and Values

@table @var
@item ptr1
@itemx ptr2
A foreign pointer.

@item boolean
@code{T} or @code{NIL}.
@end table

@subheading Description
The function @code{pointer-eq} returns true if @var{ptr1} and
@var{ptr2} point to the same memory address and false otherwise.

@subheading Implementation-specific Notes
The representation of foreign pointers varies across the various Lisp
implementations as does the behaviour of the built-in Common Lisp
equality predicates. Comparing two pointers that point to the same
address with @code{EQ} Lisps will return true on some Lisps, others require
more general predicates like @code{EQL} or @code{EQUALP} and finally
some will return false using any of these predicates. Therefore, for
portability, you should use @code{POINTER-EQ}.

@subheading Examples
This is an example using @acronym{SBCL}, see the
implementation-specific notes above.

@lisp
CFFI> (eql (null-pointer) (null-pointer))
@result{} NIL
CFFI> (pointer-eq (null-pointer) (null-pointer))
@result{} T
@end lisp

@subheading See Also
@seealso{inc-pointer}


@c ===================================================================
@c WITH-FOREIGN-OBJECT

@node with-foreign-object
@unnumberedsec with-foreign-object
@subheading Syntax
@Macro{with-foreign-object (var type &optional count) &body body}

@Macro{with-foreign-objects (bindings) &body body}

bindings ::= @{(var type &optional count)@}*

@subheading Arguments and Values

@table @var
@item var
A symbol.

@item type
A foreign type, evaluated.

@item count
An integer.
@end table

@subheading Description
The macros @code{with-foreign-object} and @code{with-foreign-objects}
bind @var{var} to a pointer to @var{count} newly allocated objects
of type @var{type} during @var{body}. The buffer has dynamic extent
and may be stack allocated if supported by the host Lisp.

@subheading Examples

@lisp
CFFI> (with-foreign-object (array :int 10)
        (dotimes (i 10)
          (setf (mem-aref array :int i) (random 100)))
        (loop for i below 10
              collect (mem-aref array :int i)))
@result{} (22 7 22 52 69 1 46 93 90 65)
@end lisp

@subheading See Also
@seealso{foreign-alloc}


@c ===================================================================
@c WITH-FOREIGN-POINTER

@node with-foreign-pointer
@unnumberedsec with-foreign-pointer
@subheading Syntax
@Macro{with-foreign-pointer (var size &optional size-var) &body body}

@subheading Arguments and Values

@table @var
@item var
@itemx size-var
A symbol.

@item size
An integer.

@item body
A list of forms to be executed.
@end table

@subheading Description
The @code{with-foreign-pointer} macro, binds @var{var} to @var{size}
bytes of foreign memory during @var{body}. The pointer in @var{var}
is invalid beyond the dynamic extend of @var{body} and may be
stack-allocated if supported by the implementation.

If @var{size-var} is supplied, it will be bound to @var{size} during
@var{body}.

@subheading Examples

@lisp
CFFI> (with-foreign-pointer (string 4 size)
        (setf (mem-ref string :char (1- size)) 0)
        (lisp-string-to-foreign "Popcorns" string size)
        (loop for i from 0 below size
              collect (code-char (mem-ref string :char i))))
@result{} (#\P #\o #\p #\Null)
@end lisp

@subheading See Also
@seealso{foreign-alloc} @*
@seealso{foreign-free}


@c ===================================================================
@c CHAPTER: Strings

@node Strings
@chapter Strings

As with many languages, Lisp and C have special support for logical
arrays of characters, going so far as to give them a special name,
``strings''.  In that spirit, @cffi{} provides special support for
translating between Lisp and C strings.

The @code{:string} type and the symbols related below also serve as an
example of what you can do portably with @cffi{}; were it not
included, you could write an equally functional @file{strings.lisp}
without referring to any implementation-specific symbols.

@menu
Dictionary

* foreign-string-alloc::        
* foreign-string-free::         
* foreign-string-to-lisp::      
* lisp-string-to-foreign::      
* with-foreign-string::         
* with-foreign-pointer-as-string::
@end menu


@c ===================================================================
@c FOREIGN-STRING-ALLOC

@node foreign-string-alloc
@unnumberedsec foreign-string-alloc
@subheading Syntax
@Function{foreign-string-alloc string-or-ub8-array @result{} pointer}

@subheading Arguments and Values

@table @var
@item string-or-ub8-array
A Lisp string or a Lisp array with element-type @code{(unsigned-byte 8)}.

@item pointer
A pointer to the newly allocated foreign string.
@end table

@subheading Description
The @code{foreign-string-alloc} function allocates a foreign string
containing a Lisp string or @code{(unsigned-byte 8)} array.

This string must be freed with @code{foreign-string-free}.

@subheading Examples

@lisp
CFFI> (setq str (foreign-string-alloc "Hello, foreign world!"))
@result{} #<FOREIGN-ADDRESS #x00400560>
CFFI> (foreign-funcall "strlen" :pointer str :int)
@result{} 21
@end lisp

@subheading See Also
@seealso{foreign-string-free} @*
@seealso{with-foreign-string}
@c @seealso{:string}


@c ===================================================================
@c FOREIGN-STRING-FREE

@node foreign-string-free
@unnumberedsec foreign-string-free
@subheading Syntax
@Function{foreign-string-free pointer}

@subheading Arguments and Values

@table @var
@item pointer
A pointer to a string allocated by @code{foreign-string-alloc}.
@end table

@subheading Description
The @code{foreign-string-free} function frees a foreign string
allocated by @code{foreign-string-alloc}.

@subheading Examples

@subheading See Also
@seealso{foreign-string-alloc}


@c ===================================================================
@c FOREIGN-STRING-TO-LISP

@node foreign-string-to-lisp
@unnumberedsec foreign-string-to-lisp
@subheading Syntax
@Function{foreign-string-to-lisp ptr &optional size null-terminated-p @
                                 @result{} string}

@subheading Arguments and Values

@table @var
@item ptr
A pointer.

@item size
The maximum string size. @code{array-total-size-limit}, by default.

@item null-terminated-p
Specifies if the string @var{ptr} points to is null terminated. True,
by default.
@end table

@subheading Description
The @code{foreign-string-to-lisp} function copies at most @var{size}
characters from @var{ptr} into a Lisp string.

When @var{null-terminated-p} is true (the default), characters are
copied until @var{size} is reached or a @code{NULL} character is
found.

If @var{ptr} is a null pointer, returns nil.

Note that the @code{:string} type will automatically convert between
Lisp strings and foreign strings.

@subheading Examples

@lisp
CFFI> (foreign-funcall "getenv" :string "HOME" :pointer)
@result{} #<FOREIGN-ADDRESS #xBFFFFFD5>
CFFI> (foreign-string-to-lisp *)
@result{} "/Users/luis"
@end lisp

@subheading See Also
@seealso{lisp-string-to-foreign} @*
@seealso{foreign-string-alloc}
@c @seealso{:string}


@c ===================================================================
@c LISP-STRING-TO-FOREIGN

@node lisp-string-to-foreign
@unnumberedsec lisp-string-to-foreign
@subheading Syntax
@Function{lisp-string-to-foreign string-or-ub8-array ptr size}

@subheading Arguments and Values

@table @var
@item string-or-ub8-array
A Lisp string or a Lisp @code{(unsigned-byte 8)} array.

@item ptr
A foreign pointer.

@item size
An integer.
@end table

@subheading Description
The @code{lisp-string-to-foreign} function copies at most @var{size}-1
characters from a Lisp string or @code{(unsigned-byte 8)} arrayto
@var{ptr}. The foreign string will be null-terminated.

@subheading Examples

@lisp
CFFI> (with-foreign-pointer-as-string (str 255)
        (lisp-string-to-foreign "Hello, foreign world!" str 6))
@result{} "Hello"

CFFI> (with-foreign-pointer-as-string (str 255)
        (lisp-string-to-foreign
         (make-array 6 :element-type '(unsigned-byte 8)
                     :initial-contents '(65 66 67 68 69 60))
         str 4))
@result{} "ABC"
@end lisp

@subheading See Also
@seealso{foreign-string-alloc} @*
@seealso{foreign-string-to-lisp} @*
@seealso{with-foreign-pointer-as-string}


@c ===================================================================
@c WITH-FOREIGN-STRING

@node with-foreign-string
@unnumberedsec with-foreign-string
@subheading Syntax
@Macro{with-foreign-string (var lisp-string-or-ub8-array) &body body}

@subheading Arguments and Values

@table @var
@item var
A symbol.

@item lisp-string-or-ub8-array
A Lisp string or a Lisp array with element type @code{(unsigned-byte 8)}.

@item body
A list of forms to be executed.
@end table

@subheading Description
The @code{with-foreign-string} macro will bind @var{var} to a newly
allocated foreign string containing @var{lisp-string-or-ub8-array}.

@subheading Examples

@lisp
CFFI> (with-foreign-string (foo "12345")
        (foreign-funcall "strlen" :pointer foo :int))
@result{} 5

CFFI> (let ((array (coerce #(84 117 114 97 110 103 97)
                           '(array (unsigned-byte 8)))))
        (with-foreign-string (foreign-string array)
          (foreign-string-to-lisp foreign-string)))
@result{} "Turanga"
@end lisp

@subheading See Also
@seealso{foreign-string-alloc} @*
@seealso{with-foreign-pointer-as-string}


@c ===================================================================
@c WITH-FOREIGN-POINTER-AS-STRING

@node with-foreign-pointer-as-string
@unnumberedsec with-foreign-pointer-as-string
@subheading Syntax
@Macro{with-foreign-pointer-as-string (var size &optional size-var) &body body}

@subheading Arguments and Values

@table @var
@item var
A symbol.

@item lisp-string
A Lisp string.

@item body
List of forms to be executed.
@end table

@subheading Description
The @code{with-foreign-pointer-as-string} macro is similar to
@code{with-foreign-pointer} except that @var{var}, as a Lisp string, is
used as the returned value of an implicit @code{progn} around @var{body}.

@subheading Examples

@lisp
CFFI> (with-foreign-pointer-as-string (str 6 str-size)
        (lisp-string-to-foreign "Hello, foreign world!" str str-size))
@result{} "Hello"
@end lisp

@subheading See Also
@seealso{foreign-string-alloc} @*
@seealso{with-foreign-string}


@c ===================================================================
@c CHAPTER: Variables

@node Variables
@chapter Variables

@menu
Dictionary

* defcvar::                     
* get-var-pointer::             
@end menu


@c ===================================================================
@c DEFCVAR

@node defcvar
@unnumberedsec defcvar
@subheading Syntax
@Macro{defcvar name-and-options type documentation @result{} lisp-name}

name-and-options ::= name | (name &key read-only (library :default))
name ::= lisp-name [foreign-name] | foreign-name [lisp-name]

@subheading Arguments and Values

@table @var
@item foreign-name
A string denoting a foreign function.

@item lisp-name
A symbol naming the Lisp function to be created.

@item type
A foreign type.

@item read-only
A boolean.

@item documentation
A Lisp string; not evaluated.
@end table

@subheading Description
The @code{defcvar} macro defines a symbol macro @var{lisp-name} that looks
up @var{foreign-name} and dereferences it acording to @var{type}.  It
can also be @code{setf}ed, unless @var{read-only} is true, in which
case an error will be signaled.

When one of @var{lisp-name} or @var{foreign-name} is omitted, the
other is automatically derived using the following rules:

@itemize
@item
Foreign names are converted to Lisp names by uppercasing, replacing
underscores with hyphens, and wrapping around asterisks.
@item
Lisp names are converted to foreign names by lowercasing, replacing
hyphens with underscores, and removing asterisks, if any.
@end itemize

@subheading Examples

@lisp
CFFI> (defcvar "errno" :int)
@result{} *ERRNO*
CFFI> (foreign-funcall "strerror" :int *errno* :string)
@result{} "Inappropriate ioctl for device"
CFFI> (setf *errno* 1)
@result{} 1
CFFI> (foreign-funcall "strerror" :int *errno* :string)
@result{} "Operation not permitted"
@end lisp

Trying to modify a read-only foreign variable:

@lisp
CFFI> (defcvar ("errno" +error-number+) :int :read-only t)
@result{} +ERROR-NUMBER+
CFFI> (setf +error-number+ 12)
;; @lispcmt{@error{} Trying to modify read-only foreign var: +ERROR-NUMBER+.}
@end lisp

@emph{Note that accessing @code{errno} this way won't work with every
C standard library.}

@subheading See Also
@seealso{get-var-pointer}


@c ===================================================================
@c GET-VAR-POINTER

@node get-var-pointer
@unnumberedsec get-var-pointer
@subheading Syntax
@Function{get-var-pointer symbol @result{} pointer}

@subheading Arguments and Values

@table @var
@item symbol
A symbol denoting a foreign variable defined with @code{defcvar}.

@item pointer
A foreign pointer.
@end table

@subheading Description
The function @code{get-var-pointer} will return a @var{pointer} to the
foreign global variable @var{symbol} previously defined with
@code{defcvar}.

@subheading Examples

@lisp
CFFI> (defcvar "errno" :int :read-only t)
@result{} *ERRNO*
CFFI> *errno*
@result{} 25
CFFI> (get-var-pointer '*errno*)
@result{} #<A Mac Pointer #xA0008130>
CFFI> (mem-ref * :int)
@result{} 25
@end lisp

@subheading See Also
@seealso{defcvar}


@c ===================================================================
@c CHAPTER: Functions

@node Functions
@chapter Functions

@menu
* Calling Foreign Functions::   
* Defining Foreign Functions::  

Dictionary

* defcfun::                     
* foreign-funcall::             
* foreign-funcall-pointer::     
@end menu

@node Calling Foreign Functions
@section Calling Foreign Functions

@node Defining Foreign Functions
@section Defining Foreign Functions


@c ===================================================================
@c DEFCFUN

@node defcfun
@unnumberedsec defcfun
@subheading Syntax
@Macro{defcfun name-and-options return-type &body arguments [&rest] @
               @result{} lisp-name}

@table @asis
@item @var{name-and-options} name | (name &key library calling-convention cconv)
@item @var{name} ::= @var{lisp-name} [@var{foreign-name}] | @var{foreign-name} [@var{lisp-name}]
@item @var{arguments} ::= @{ (arg-name arg-type) @}*
@end table

@subheading Arguments and Values

@table @var
@item foreign-name
A string denoting a foreign function.

@item lisp-name
A symbol naming the Lisp function to be created.

@item arg-name
A symbol.

@item return-type
@itemx arg-type
A foreign type.

@item calling-convention
@itemx cconv
One of @code{:cdecl} (default) or @code{stdcall}.

@item library
A symbol designating a foreign library.
@end table

@subheading Description
The @code{defcfun} macro provides a declarative interface for defining
Lisp functions that call foreign functions.

When one of @var{lisp-name} or @var{foreign-name} is omitted, the
other is automatically derived using the following rules:

@itemize
@item
Foreign names are converted to Lisp names by uppercasing and replacing
underscores with hyphens.
@item
Lisp names are converted to foreign names by lowercasing and replacing
hyphens with underscores.
@end itemize

If you place the symbol @code{&rest} in the end of the argument list
after the fixed arguments, @code{defcfun} will treat the foreign
function as a @strong{variadic function}. The variadic arguments
should be passed in a way similar to what @code{foreign-funcall} would
expect. Unlike @code{foreign-funcall} though, @code{defcfun} will take
care of doing argument promotion. Note that in this case
@code{defcfun} will generate a Lisp @emph{macro} instead of a
function and will only work for Lisps that support
@code{foreign-funcall.}


@subheading Examples

@lisp
(defcfun "strlen" :int (n :string))

CFFI> (strlen "123")
@result{} 3
@end lisp

@lisp
(defcfun ("abs" c-abs) :int (n :int))

CFFI> (c-abs -42)
@result{} 42
@end lisp

Variadic function example:

@lisp
(defcfun "sprintf" :int
  (str :pointer)
  (control :string)
  &rest)

CFFI> (with-foreign-pointer-as-string (s 100)
        (sprintf s "%c %d %.2f %s" :char 90 :short 42 :float pi
                 :string "super-locrian"))
@result{} "A 42 3.14 super-locrian"
@end lisp

@subheading See Also
@seealso{foreign-funcall} @*
@seealso{foreign-funcall-pointer}


@c ===================================================================
@c FOREIGN-FUNCALL

@node foreign-funcall
@unnumberedsec foreign-funcall
@subheading Syntax
@Macro{foreign-funcall name-and-options &rest arguments @result{} return-value}

arguments ::= @{ arg-type arg @}* [return-type]
name-and-options ::= name | ( name &key library calling-convention cconv)

@subheading Arguments and Values

@table @var
@item name
A Lisp string.

@item arg-type
A foreign type.

@item arg
An argument of type @var{arg-type}.

@item return-type
A foreign type, @code{:void} by default.

@item return-value
A lisp object.

@item library
A lisp symbol; not evaluated.

@item calling-convention
@itemx cconv
One of @code{:cdecl} (default) or @code{:stdcall}.
@end table

@subheading Description
The @code{foreign-funcall} macro is the main primitive for calling
foreign functions.

@emph{Note: The return value of foreign-funcall on functions with a
:void return type is still undefined.}

@subheading Implementation-specific Notes
@itemize
@item
Corman Lisp does not support @code{foreign-funcall}. On
implementations that @strong{don't} support @code{foreign-funcall}
@code{cffi-features:no-foreign-funcall} will be present in
@code{*features*}. Note: in these Lisps you can still use the
@code{defcfun} interface.
@end itemize

@subheading Examples

@lisp
CFFI> (foreign-funcall "strlen" :string "foo" :int)
@result{} 3
@end lisp

Given the C code:

@example
void print_number(int n)
@{
    printf("N: %d\n", n);
@}
@end example

@lisp
CFFI> (foreign-funcall "print_number" :int 123456)
@print{} N: 123456
@result{} NIL
@end lisp

@noindent
Or, equivalently:

@lisp
CFFI> (foreign-funcall "print_number" :int 123456 :void)
@print{} N: 123456
@result{} NIL
@end lisp

@lisp
CFFI> (foreign-funcall "printf" :string (format nil "%s: %d.~%")
                       :string "So long and thanks for all the fish"
                       :int 42 :int)
@print{} So long and thanks for all the fish: 42.
@result{} 41
@end lisp

@subheading See Also
@seealso{defcfun} @*
@seealso{foreign-funcall-pointer}


@c ===================================================================
@c FOREIGN-FUNCALL-POINTER

@node foreign-funcall-pointer
@unnumberedsec foreign-funcall-pointer
@subheading Syntax
@Macro{foreign-funcall pointer options &rest arguments @result{} return-value}

arguments ::= @{ arg-type arg @}* [return-type]
options ::= ( &key calling-convention cconv )

@subheading Arguments and Values

@table @var
@item pointer
A foreign pointer.

@item arg-type
A foreign type.

@item arg
An argument of type @var{arg-type}.

@item return-type
A foreign type, @code{:void} by default.

@item return-value
A lisp object.

@item calling-convention
@itemx cconv
One of @code{:cdecl} (default) or @code{:stdcall}.
@end table

@subheading Description
The @code{foreign-funcall} macro is the main primitive for calling
foreign functions.

@emph{Note: The return value of foreign-funcall on functions with a
:void return type is still undefined.}

@subheading Implementation-specific Notes
@itemize
@item
Corman Lisp does not support @code{foreign-funcall}. On
implementations that @strong{don't} support @code{foreign-funcall}
@code{cffi-features:no-foreign-funcall} will be present in
@code{*features*}. Note: in these Lisps you can still use the
@code{defcfun} interface. 
@end itemize

@subheading Examples

@lisp
CFFI> (foreign-funcall-pointer (foreign-symbol-pointer "abs")
                               :int -42 :int)
@result{} 42
@end lisp

@subheading See Also
@seealso{defcfun} @*
@seealso{foreign-funcall}


@c ===================================================================
@c CHAPTER: Libraries

@node Libraries
@chapter Libraries

@menu
* Defining a library::          
* Library definition style::    

Dictionary

* close-foreign-library::       Close a foreign library.
* *darwin-framework-directories*::  Search path for Darwin frameworks.
* define-foreign-library::      Explain how to load a foreign library.
* *foreign-library-directories*::  Search path for shared libraries.
* load-foreign-library::        Load a foreign library.
* load-foreign-library-error::  Signalled on failure of its namesake.
* use-foreign-library::         Load a foreign library when needed.
@end menu


@node Defining a library
@section Defining a library

Almost all foreign code you might want to access exists in some kind
of shared library.  The meaning of @dfn{shared library} varies among
platforms, but for our purposes, we will consider it to include
@file{.so} files on @sc{unix}, frameworks on Darwin (and derivatives
like Mac @acronym{OS X}), and @file{.dll} files on Windows.

Bringing one of these libraries into the Lisp image is normally a
two-step process.

@enumerate
@item
Describe to @cffi{} how to load the library at some future point,
depending on platform and other factors, with a
@code{define-foreign-library} top-level form.

@item
Load the library so defined with either a top-level
@code{use-foreign-library} form or by calling the function
@code{load-foreign-library}.
@end enumerate

@xref{Tutorial-Loading,, Loading foreign libraries}, for a working
example of the above two steps.


@node Library definition style
@section Library definition style

Looking at the @code{libcurl} library definition presented earlier,
you may ask why we did not simply do this:

@lisp
(define-foreign-library libcurl
  (t (:default "libcurl")))
@end lisp

@noindent
Indeed, this would work just as well on the computer on which I tested
the tutorial.  There are a couple of good reasons to provide the
@file{.so}'s current version number, however.  Namely, the versionless
@file{.so} is not packaged on most @sc{unix} systems along with the
actual, fully-versioned library; instead, it is included in the
``development'' package along with C headers and static @file{.a}
libraries.

The reason @cffi{} does not try to account for this lies in the
meaning of the version numbers.  A full treatment of shared library
versions is beyond this manual's scope; see @ref{Versioning,, Library
interface versions, libtool, @acronym{GNU} Libtool}, for helpful
information for the unfamiliar.  For our purposes, consider that a
mismatch between the library version with which you tested and the
installed library version may cause undefined
behavior.@footnote{Windows programmers may chafe at adding a
@sc{unix}-specific clause to @code{define-foreign-library}.  Instead,
ask why the Windows solution to library incompatibility is ``include
your own version of every library you use with every program''.}

@impnote{Maybe some notes should go here about OS X, which I know
little about.  --stephen}


@c ===================================================================
@c CLOSE-FOREIGN-LIBRARY

@node close-foreign-library
@unnumberedsec close-foreign-library
@subheading Syntax
@Function{close-foreign-library library @result{} success}

@subheading Arguments and Values

@table @var
@item library
A symbol or an instance of @code{foreign-library}.

@item success
A Lisp boolean.
@end table

@subheading Description

Closes @var{library} which can be a symbol designating a library
define through @code{define-foreign-library} or an instance of
@code{foreign-library} as returned by @code{load-foreign-library}.

@c @subheading Examples
@c @xref{Tutorial-Loading,, Loading foreign libraries}.

@subheading See Also

@seealso{define-foreign-library} @*
@seealso{load-foreign-library} @*
@seealso{use-foreign-library}


@c ===================================================================
@c *DARWIN-FRAMEWORK-DIRECTORIES*

@node *darwin-framework-directories*
@unnumberedsec *darwin-framework-directories*
@subheading Syntax

@Variable{*darwin-framework-directories*}

@subheading Value type

A list, in which each element is a string, a pathname, or a simple
Lisp expression.

@subheading Initial value

A list containing the following, in order: an expression corresponding
to Darwin path @file{~/Library/Frameworks/},
@code{#P"/Library/Frameworks/"}, and
@code{#P"/System/Library/Frameworks/"}.

@subheading Description

The meaning of ``simple Lisp expression'' is explained in
@ref{*foreign-library-directories*}.  In contrast to that variable,
this is not a fallback search path; the default value described above
is intended to be a reasonably complete search path on Darwin systems.

@subheading Examples

@lisp
CFFI> (load-foreign-library '(:framework "OpenGL"))
@result{} #P"/System/Library/Frameworks/OpenGL.framework/OpenGL"
@end lisp

@subheading See also

@seealso{*foreign-library-directories*} @*
@seealso{define-foreign-library}


@c ===================================================================
@c DEFINE-FOREIGN-LIBRARY

@node define-foreign-library
@unnumberedsec define-foreign-library

@subheading Syntax

@Macro{define-foreign-library name-and-options @{ load-clause @}* @result{} name}

name-and-options ::= name | (name &key calling-convention cconv)
load-clause ::= (feature library &key calling-convention cconv)

@subheading Arguments and Values

@table @var
@item name
A symbol.

@item feature
A feature expression.

@item library
A library designator.

@item calling-convention
@itemx cconv
One of @code{:cdecl} (default) or @code{:stdcall}
@end table

@subheading Description

Creates a new library designator called @var{name}.  The
@var{load-clause}s describe how to load that designator when passed to
@code{load-foreign-library} or @code{use-foreign-library}.

When trying to load the library @var{name}, the relevant function
searches the @var{load-clause}s in order for the first one where
@var{feature} evaluates to true.  That happens for any of the
following situations:@footnote{This is described in
@code{cffi-feature-p} in @file{libraries.lisp}.}

@enumerate 1
@item
If @var{feature} is a symbol (idiomatically a keyword), a symbol with
the same name, but interned into the @code{cffi-features} package, is
present in @code{common-lisp:*features*}.

@item
If @var{feature} is a list, depending on @code{(first @var{feature})},
a keyword:

@table @code
@item :and
All of the feature expressions in @code{(rest @var{feature})} are
true.

@item :or
At least one of the feature expressions in @code{(rest @var{feature})}
is true.

@item :not
The feature expression @code{(second @var{feature})} is not true.
@end table

@item
Finally, if @var{feature} is @code{t}, this @var{load-clause} is
picked unconditionally.
@end enumerate

Upon finding the first true @var{feature}, the library loader then
loads the @var{library}.  The meaning of ``library designator'' is
described in @ref{load-foreign-library}.

Functions associated to a library defined by
@code{define-foreign-library} (e.g. through @code{defcfun}'s
@code{:library} option, will inherit the library's options.  The
precedence is as follows:

@enumerate 1
@item
@code{defcfun}/@code{foreign-funcall} specific options;

@item
@var{load-clause} options;

@item
global library options (the @var{name-and-options} argument)
@end enumerate


@subheading Examples

@xref{Tutorial-Loading,, Loading foreign libraries}.


@subheading See Also

@seealso{close-foreign-library} @*
@seealso{load-foreign-library}


@c ===================================================================
@c *FOREIGN-LIBRARY-DIRECTORIES*

@node *foreign-library-directories*
@unnumberedsec *foreign-library-directories*
@subheading Syntax

@Variable{*foreign-library-directories*}

@subheading Value type

A list, in which each element is a string, a pathname, or a simple
Lisp expression.

@subheading Initial value

The empty list.

@subheading Description

You should not have to use this variable.

Most, if not all, Lisps supported by @cffi{} have a reasonable default
search algorithm for foreign libraries.  For example, Lisps for
@sc{unix} usually call
@uref{http://www.opengroup.org/onlinepubs/009695399/functions/dlopen.html,,
@code{dlopen(3)}}, which in turn looks in the system library
directories.  Only if that fails does @cffi{} look for the named
library file in these directories, and load it from there if found.

Thus, this is intended to be a @cffi{}-only fallback to the library
search configuration provided by your operating system.  For example,
if you distribute a foreign library with your Lisp package, you can
add the library's containing directory to this list and portably
expect @cffi{} to find it.

A @dfn{simple Lisp expression} is intended to provide functionality
commonly used in search paths such as
@acronym{ASDF}'s@footnote{@xref{Using asdf to load systems,,, asdf,
asdf: another system definition facility}, for information on
@code{asdf:*central-registry*}.}, and is defined recursively as
follows:@footnote{See @code{mini-eval} in @file{libraries.lisp} for
the source of this definition.  As is always the case with a Lisp
@code{eval}, it's easier to understand the Lisp definition than the
english.}

@enumerate
@item
A list, whose @samp{first} is a function designator, and whose
@samp{rest} is a list of simple Lisp expressions to be evaluated and
passed to the so-designated function.  The result is the result of the
function call.

@item
A symbol, whose result is its symbol value.

@item
Anything else evaluates to itself.
@end enumerate


@subheading Examples

@example
$ ls
@print{} liblibli.so    libli.lisp
@end example

@noindent
In @file{libli.lisp}:

@lisp
(pushnew #P"/home/sirian/lisp/libli/" *foreign-library-directories*
         :test #'equal)

(load-foreign-library '(:default "liblibli"))
@end lisp

@noindent
The following example would achieve the same effect:

@lisp
(pushnew '(merge-pathnames #p"lisp/libli/" (user-homedir-pathname))
          *foreign-library-directories*
          :test #'equal)
@result{} ((MERGE-PATHNAMES #P"lisp/libli/" (USER-HOMEDIR-PATHNAME)))

(load-foreign-library '(:default "liblibli"))
@end lisp

@subheading See also

@seealso{*darwin-framework-directories*} @*
@seealso{define-foreign-library}


@c ===================================================================
@c LOAD-FOREIGN-LIBRARY

@node load-foreign-library
@unnumberedsec load-foreign-library
@subheading Syntax
@Function{load-foreign-library library @result{} handler}

@subheading Arguments and Values

@table @var
@item library
A library designator.

@item handler
An instance of @code{foreign-library}.
@end table

@subheading Description

Load the library indicated by @var{library}.  A @dfn{library
designator} is defined as follows:

@enumerate
@item
If a symbol, is considered a name previously defined with
@code{define-foreign-library}.

@item
If a string or pathname, passed as a namestring directly to the
implementation's foreign library loader.  If that fails, search the
directories in @code{*foreign-library-directories*} with
@code{cl:probe-file}; if found, the absolute path is passed to the
implementation's loader.

@item
If a list, the meaning depends on @code{(first @var{library})}:

@table @code
@item :framework
The second list element is taken to be a Darwin framework name, which
is then searched in @code{*darwin-framework-directories*}, and loaded
when found.

@item :or
Each remaining list element, itself a library designator, is loaded in
order, until one succeeds.

@item :default
The name is transformed according to the platform's naming convention
to shared libraries, and the resultant string is loaded as a library
designator.  For example, on @sc{unix}, the name is suffixed with
@file{.so}.
@end table
@end enumerate

If the load fails, signal a @code{load-foreign-library-error}.

@strong{Please note:} For system libraries, you should not need to
specify the directory containing the library.  Each operating system
has its own idea of a default search path, and you should rely on it
when it is reasonable.

@subheading Implementation-specific Notes
On ECL platforms where its dynamic FFI is not supported (ie. when
@code{:dffi} is not present in @code{*features*}),
@code{cffi:load-foreign-library} does not work and you must use ECL's
own @code{ffi:load-foreign-library} with a constant string argument.

@subheading Examples

@xref{Tutorial-Loading,, Loading foreign libraries}.

@subheading See Also

@seealso{close-foreign-library} @*
@seealso{*darwin-framework-directories*} @*
@seealso{define-foreign-library} @*
@seealso{*foreign-library-directories*} @*
@seealso{load-foreign-library-error} @*
@seealso{use-foreign-library}


@c ===================================================================
@c LOAD-FOREIGN-LIBRARY-ERROR

@node load-foreign-library-error
@unnumberedsec load-foreign-library-error

@subheading Syntax

@Condition{load-foreign-library-error}

@subheading Class precedence list

@code{load-foreign-library-error}, @code{error},
@code{serious-condition}, @code{condition}, @code{t}

@subheading Description

Signalled when a foreign library load completely fails.  The exact
meaning of this varies depending on the real conditions at work, but
almost universally, the implementation's error message is useless.
However, @cffi{} does provide the useful restarts @code{retry} and
@code{use-value}; invoke the @code{retry} restart to try loading the
foreign library again, or the @code{use-value} restart to try loading
a different foreign library designator.

@subheading See also

@seealso{load-foreign-library}


@c ===================================================================
@c USE-FOREIGN-LIBRARY

@node use-foreign-library
@unnumberedsec use-foreign-library

@subheading Syntax

@Macro{use-foreign-library name}

@subheading Arguments and values

@table @var
@item name
A library designator; unevaluated.
@end table


@subheading Description

@xref{load-foreign-library}, for the meaning of ``library
designator''.  This is intended to be the top-level form used
idiomatically after a @code{define-foreign-library} form to go ahead
and load the library. @c ; it also sets the ``current foreign library''.
Finally, on implementations where the regular evaluation rule is
insufficient for foreign library loading, it loads it at the required
time.@footnote{Namely, @acronym{CMUCL}.  See
@code{use-foreign-library} in @file{libraries.lisp} for details.}

@c current foreign library is a concept created a few hours ago as of
@c this writing.  It is not actually used yet, but probably will be.

@subheading Examples

@xref{Tutorial-Loading,, Loading foreign libraries}.


@subheading See also

@seealso{load-foreign-library}


@c ===================================================================
@c CHAPTER: Callbacks

@node Callbacks
@chapter Callbacks

@menu
Dictionary

* callback::                    
* defcallback::                 
* get-callback::                
@end menu


@c ===================================================================
@c CALLBACK

@node callback
@unnumberedsec callback
@subheading Syntax
@Macro{callback symbol @result{} pointer}

@subheading Arguments and Values

@table @var
@item symbol
A symbol denoting a callback.

@item pointer
@itemx new-value
A pointer.
@end table

@subheading Description
The @code{callback} macro is analogous to the standard CL special
operator @code{function} and will return a pointer to the callback
denoted by the symbol @var{name}.

@subheading Examples

@lisp
CFFI> (defcallback sum :int ((a :int) (b :int))
        (+ a b))
@result{} SUM
CFFI> (callback sum)
@result{} #<A Mac Pointer #x102350>
@end lisp

@subheading See Also
@seealso{get-callback} @*
@seealso{defcallback}


@c ===================================================================
@c DEFCALLBACK

@node defcallback
@unnumberedsec defcallback
@subheading Syntax
@Macro{defcallback name-and-options return-type arguments &body body @result{} name}

name-and-options ::= name | (name &key calling-convention cconv)
arguments ::= (@{ (arg-name arg-type) @}*)

@subheading Arguments and Values

@table @var
@item name
A symbol naming the callback created.

@item return-type
The foreign type for the callback's return value.

@item arg-name
A symbol.

@item arg-type
A foreign type.

@item calling-convention
@itemx cconv
One of @code{:cdecl} (default) or @code{:stdcall}.
@end table

@subheading Description
The macro @code{defcallback} defines a Lisp function the can be called
from C (but not from Lisp). The arguments passed to this function will
be converted to the appropriate Lisp representation and its return
value will be converted to its C representation.

This Lisp function can be accessed by the @code{callback} macro or the
@code{get-callback} function.

@strong{Portability note:} @code{defcallback} will not work correctly
on some Lisps if it's not a top-level form.

@subheading Examples

@lisp
(defcfun "qsort" :void
  (base :pointer)
  (nmemb :int)
  (size :int)
  (fun-compar :pointer))

(defcallback < :int ((a :pointer) (b :pointer))
  (let ((x (mem-ref a :int))
        (y (mem-ref b :int)))
    (cond ((> x y) 1)
          ((< x y) -1)
          (t 0))))

CFFI> (with-foreign-object (array :int 10)
        ;; @lispcmt{Initialize array.}
        (loop for i from 0 and n in '(7 2 10 4 3 5 1 6 9 8)
              do (setf (mem-aref array :int i) n))
        ;; @lispcmt{Sort it.}
        (qsort array 10 (foreign-type-size :int) (callback <))
        ;; @lispcmt{Return it as a list.}
        (loop for i from 0 below 10
              collect (mem-aref array :int i)))
@result{} (1 2 3 4 5 6 7 8 9 10)
@end lisp

@subheading See Also
@seealso{callback} @*
@seealso{get-callback}


@c ===================================================================
@c GET-CALLBACK

@node get-callback
@unnumberedsec get-callback
@subheading Syntax
@Accessor{get-callback symbol @result{} pointer}

@subheading Arguments and Values

@table @var
@item symbol
A symbol denoting a callback.

@item pointer
A pointer.
@end table

@subheading Description
This is the functional version of the @code{callback} macro. It
returns a pointer to the callback named by @var{symbol} suitable, for
example, to pass as arguments to foreign functions.

@subheading Examples

@lisp
CFFI> (defcallback sum :int ((a :int) (b :int))
        (+ a b))
@result{} SUM
CFFI> (get-callback 'sum)
@result{} #<A Mac Pointer #x102350>
@end lisp

@subheading See Also
@seealso{callback} @*
@seealso{defcallback}

@c ===================================================================
@c CHAPTER: Limitations

@node Limitations
@chapter Limitations

These are @cffi{}'s limitations across all platforms; for information
on the warts on particular Lisp implementations, see
@ref{Implementation Support}.

@itemize @bullet
@item
The tutorial includes a treatment of the primary, intractable
limitation of @cffi{}, or any @acronym{FFI}: that the abstractions
commonly used by C are insufficiently expressive.
@xref{Tutorial-Abstraction,, Breaking the abstraction}, for more
details.

@item
C @code{struct}s cannot be passed by value.
@end itemize

@c more?


@node Platform-specific features
@appendix Platform-specific features

@cffi{} does some platform tests on loading.  The details vary between
Lisps; in fact, the purpose is to unify the list of available platform
features for use elsewhere in the @cffi{} code.  These features are
also part of the public interface; see @ref{define-foreign-library}.

The exact meanings of the features follow.  Though you will usually
refer to these symbols as keywords, @cffi{} internally views them in
the package @code{cffi-features}.

@table @code
@item flat-namespace
This Lisp has a flat namespace for foreign symbols meaning that you
won't be able to load two different libraries with homograph functions
and successfully differentiate them through the @code{:library}
option to @code{defcfun}, @code{defcvar}, etc@dots{}

@item darwin
This operating system is Darwin or a derivative thereof, such as
Mac @acronym{OS X}.

@item no-foreign-funcall
The macro @code{foreign-funcall} is @strong{not} available.  On such
platforms, the only way to call a foreign function is through
@code{defcfun}.  @xref{foreign-funcall}, and @ref{defcfun}.

@item no-long-long
The C @code{long long} type is @strong{not} available as a foreign
type.

@item no-stdcall
This Lisp doesn't support the @code{stdcall} calling convention.  Note
that it only makes sense to support @code{stdcall} on (32-bit) x86
platforms.

@item ppc32
The underlying @acronym{CPU} architecture is 32-bit PowerPC.

@item unix
This operating system is a @sc{unix}-like, such as
@acronym{GNU}/Linux, Darwin, or even Cygwin on Lisps that show the
@sc{unix}-like interface provided by Cygwin to Lisp code.

@item windows
This operating system is Windows.

@item x86
The underlying @acronym{CPU} architecture is x86, such as on
processors from Intel or @acronym{AMD}.
@end table


@node Glossary
@appendix Glossary

@table @dfn
@item aggregate type
A @cffi{} type for C data defined as an organization of data of simple
type; in structures and unions, which are themselves aggregate types,
they are represented by value.

@item foreign value
This has two meanings; in any context, only one makes sense.

When using type translators, the foreign value is the lower-level Lisp
value derived from the object passed to @code{translate-to-foreign}
(@pxref{translate-to-foreign}).  This value should be a Lisp number or
a pointer (satisfies @code{pointerp}), and it can be treated like any
general Lisp object; it only completes the transformation to a true
foreign value when passed through low-level code in the Lisp
implementation, such as the foreign function caller or indirect memory
addressing combined with a data move.

In other contexts, this refers to a value accessible by C, but which
may only be accessed through @cffi{} functions.  The closest you can
get to such a foreign value is through a pointer Lisp object, which
itself counts as a foreign value in only the previous sense.

@item simple type
A @cffi{} type that is ultimately represented as a builtin type;
@cffi{} only provides extra semantics for Lisp that are invisible to C
code or data.
@end table

@node Comprehensive Index
@unnumbered Index
@printindex cp

@bye