Clearify code a little

[elinks.git] / doc / lua-scripting.txt

[[lua-scripting]]
Scripting ELinks with Lua
-------------------------

This file documents the Lua scripting interface of the ELinks web browser.

Introduction
~~~~~~~~~~~~

What is it
^^^^^^^^^^

Lua scripting capabilities permit users to customize the ELinks behaviour to
unusual degree - they allow automatic rewriting of HTML code of the received
documents, rewriting of the URLs entered by user etc. You can even write your
own bookmarks system with Lua. See also contrib/lua/ for some examples of the
possibilities of ELinks Lua support.

Please do not confuse Lua scripting with JavaScript, EcmaScript, VBScript and
similiar. Those are embedded in page, allowing per-document scripting related
to its presentation and providing some degree of interactivity etc. On the
contrary, the current Lua support permits scripts to be embedded to the
browser directly, changing the behaviour of the browser, not the document.

The original Lua support (in the form of Links-Lua fork of original Links) was
written by Peter Wang and Cliff Cunnington.  There are some rough edges
remaining, but is suitable for everyday use (I have been using it every day
for a year).

Where to get it
^^^^^^^^^^^^^^^

The Lua scripting support comes with the stock ELinks distribution, no
additional patches and tweaks should be needed.

The web site of the original Links-Lua is at
link:http://links.sourceforge.net/links-lua/[]. Some older patches against
regular Links are available at
link:http://www.sourceforge.net/projects/links/[], but they are not being
maintained.

Lua can be found at link:http://www.lua.org/[].

What it runs on
^^^^^^^^^^^^^^^

The Lua support has only been tested under Linux, although it *should*
work under other platforms that ELinks and Lua support (perhaps with some
changes to source code?).

Also, note that many of the scripts given here assume a Unix system.
Your mileage will definitely vary on other platforms.


Installing
~~~~~~~~~~

Installing Lua
^^^^^^^^^^^^^^

Before you can compile ELinks with Lua support, you must compile and install
Lua.  The following instructions are for a Linux system.  People on other
systems should try to enable `popen` support, but this is not necessary
(you will lose a bit of functionality though).

1. Download and unpack the Lua `tar.gz` or `zip` somewhere.
2. `cd` into the `lua` directory.
3. Open `config` in a text editor and uncomment the `POPEN` line.
4. Optionally, change the `INSTALL_ROOT line.
5. Run `make; make so; make sobin; make install`.  

On systems without shared object support, simply run `make; make install`
instead.


Installing ELinks
^^^^^^^^^^^^^^^^^

Follow the instructions for building ELinks (it is the standard
`./configure; make; make install` procedure).  During the configure
step make sure that Lua has been detected on your system.


Running ELinks with Lua
^^^^^^^^^^^^^^^^^^^^^^^

Simply start ELinks as you normally would.  To check you have Lua support
compiled in, open up the "Help | About" dialog box.  It should list "Lua"
under "Features".  If not, make sure you do not have other copies of ELinks
running, or start ELinks again with the "-no-connect" option on the
command-line.


Using ELinks with Lua
~~~~~~~~~~~~~~~~~~~~~

Out of the box, ELinks with Lua will do nothing different from regular ELinks.
You need to write some scripts.

ELinks Lua additions
^^^^^^^^^^^^^^^^^^^^

The Lua support is based on the idea of @dfn{hooks}.  A hook is a function that
gets called at a particular point during the execution of ELinks.  To make
ELinks do what you want, you can add and edit such hooks.

The Lua support also adds an extra dialog box, which you can open while in
ELinks with the comma (`,`) key.  Here you can enter Lua expressions for
evaluation, or override it to do something different.

And finally, you can bind keystrokes to Lua functions.  These keystrokes
won't let you do any more than is possible with the Lua Console, but
they're more convenient.

Note that this document assumes you have some knowledge of programming in Lua.
For that, you should refer to the Lua reference manual
(link:http://www.lua.org/docs.html[]). In fact, the language is relatively
trivial, though. You could already do wonders with simply refactoring the
example scripts.

Config file
^^^^^^^^^^^

On startup, ELinks reads in two Lua scripts.  Firstly, a system-wide
configuration file called `/etc/elinks/hooks.lua`, then a file in your home
directory called `~/.elinks/hooks.lua`.  From these files, you can include
other Lua files with `dofile`, if necessary.

To see what kind of things you should put in here, look at
`contrib/lua/hooks.lua`.

Hooks
^^^^^

The following hooks are available.

goto_url_hook (url, current_url)::
        This hook is called when the user enters a string into the "Go to URL"
        dialog box.  It is given the string entered, and the current URL
        (which may be `nil`).  It should return a string, which is the URL
        that ELinks should follow, or `nil` to cancel the operation.

follow_url_hook (url)::
        This hook is passed the URL that Links is about to follow.  It should
        return a string (the URL modified or unmodified), or `nil` to stop
        ELinks following the URL

pre_format_html_hook (url, html)::
        This hook gets called just before the final time an HTML document is
        formatted, i.e. it only gets called once, after the entire document is
        downloaded.  It will be passed the URL and HTML text as strings, and
        should return the modified HTML text, or `nil` if there were no
        modifications.

lua_console_hook (string)::
        This hook is passed the string that the user entered into the "Lua
        Console" dialog box.  It should return two values: the type of action
        to take (`run`, `eval`, `goto-url` or `nil`), and
        a second argument, which is the shell command to run or the Lua
        expression to evaluate. Examples:
         - `return "run", "someprogram"` will attempt to run the program
           `someprogram`.
         - `return "eval", "somefunction(1+2)"` will attempt to call the Lua
           function `somefunction` with an argument, 3.
         - `return "goto-url", "http://www.bogus.com"` will ask Links to visit
           the URL "http://www.bogus.com".
         - `return nil` will do nothing.

quit_hook ()::
        This hook is run just before ELinks quits.  It is useful for cleaning
        up things, such as temporary files you have created.


Functions
^^^^^^^^^

As well as providing hooks, ELinks provides some functions in addition to the
standard Lua functions.

enable_systems_functions ()::
Enable some potentially dangerous functions, as well as some other
functions which were unfortunate enough to be lumped in the same group.

The functions are: `openfile`, `closefile`, `readfrom`, `writeto`, `appendto`,
`pipe_read`, `remove`, `rename`, `flush`, `seek`, `tmpname`, `read`, `write`
`execute`, `exit`, `clock`, `date`, `getenv`, `setlocale`.

Note: `setlocale` is a standard Lua function and will not affect
the current ELinks locale.
@end deffn

current_url ()::
        Returns the URL of the current page being shown (in the ELinks session
        that invoked the function).

current_link ()::
        Returns the URL of the currently selected link, or `nil` if none is
        selected.

current_title ()::
        Returns the title of the current page, or `nil` if none.

current_document ()::
        Returns the current document as a string, unformatted.

current_document_formatted ([width])::
        Returns the current document, formatted for the specified screen
        width.  If the width is not specified, then the document is formatted
        for the current screen width (i.e. what you see on screen).  Note that
        this function does *not* guarantee all lines will be shorter than
        `width`, just as some lines may be wider than the screen when
        viewing documents online.

pipe_read (command)::
        Executes `command` and reads in all the data from stdout, until there
        is no more.  This is a hack, because for some reason the standard Lua
        function `read` seems to crash ELinks when used in pipe-reading mode.

execute (string)::
        Executes shell commands `string` and returns the exit code.  Beware
        that you must not read or write to stdin and stdout.  And unlike the
        standard Lua function of the same name, the return value is
        meaningless.

bind_key (keymap, keystroke, function)::
        Currently, `keymap` must be the string `"main"`.  Keystroke is a
        keystroke as you would write it in the ELinks config file
        `~/.elinks/elinks.conf`.  The function `function` should take no
        arguments, and should return the same values as `lua_console_hook`.


User protocol
^^^^^^^^^^^^^

There is one more little thing which Links-Lua adds, which will not be
described in detail here.  It is the fake "user:" protocol, which can be used
when writing your own addons.  It allows you to generate web pages containing
links to "user://blahblah", which can be intercepted by the `follow_url_hook`
(among other things) to perform unusual actions.  For a concrete example, see
the bookmark addon.


Example recipes
~~~~~~~~~~~~~~~

This chapter contains some example scripts that you can use.  All of them come
from `contrib/lua/hooks.lua`.  I really recommend you to see it directly
instead of copying code out of this document.  Also, not everything in there is
covered here.

If you would like to contribute scripts, that would be great!  Please send
them to me at mailto:tjaden@@users.sourceforge.net[].  Cliff and I plan to
start a script repository, provided we get some contributions.  As for script
ideas, you'll just have to be a little creative :-)

Also take a look at the `contrib/lua/` directory in the ELinks distribution.
Note that Peter and Cliff don't maintain the Lua support intensively anymore,
thus it would be probably nice to Cc me (mailto:pasky@@ucw.cz[]) if you want
to contribute some patch, so that I would be able to add it to the ELinks
distribution.


Go to URL on steroids
^^^^^^^^^^^^^^^^^^^^^

There are some web sites that I visit often.  Bookmarks are okay, but they are
separate from the "Go to URL" dialog box, so I keep forgetting to use them.
Also, when I visit a search engine home page, all I really want to do is enter
a search term.

The following script allows me to type certain strings into the "Go to URL"
dialog box, and it will convert them to the URL I actually want to visit.  As
a bonus, it allows me perform some searches on sites like Google without
loading up the front page first.

-------------------------------------------------------------------------------
function match (prefix, url)
    return strsub (url, 1, strlen (prefix)) == prefix
end

function strip (str)
    return gsub (str, "^%s*(.-)%s*$", "%1")
end

function plusify (str)
    return gsub (str, "%s", "+")
end

function goto_url_hook (url, current_url)
    -- Google search (e.g. ,gg unix browsers).
    if match (",gg", url) then
        url = plusify (strip (strsub (url, 4)))
        return "http://www.google.com/search?q="..url.."&btnG=Google+Search"
 
    -- Freshmeat search.
    elseif match (",fm", url) then
        url = plusify (strip (strsub (url, 4)))
        return "http://www.freshmeat.net/search/?q="..url

    -- Appwatch search (e.g. ,aw lynx).
    elseif match (",aw", url) then
        url = plusify (strip (strsub (url, 4)))
        return "http://www.appwatch.com/Linux/Users/find?q="..url
    
    -- Dictionary.com search (e.g. ,dict congenial).
    elseif match (",dict", url) then
        url = plusify (strip (strsub (url, 6)))
        return "http://www.dictionary.com/cgi-bin/dict.pl?db=%2A&term="..url

    -- RPM search (e.g. ,rpm links).
    elseif match (",rpm", url) then
        url = plusify (strip (strsub (url, 5)))
        return "http://www.rpmfind.net/linux/rpm2html/search.php?query="
                ..url.."&submit=Search+..."
        
    -- Netcraft.com search (e.g. ,whatis www.google.com).
    elseif match (",whatis", url) then
        url = plusify (strip (strsub (url, 8)))
        return "http://uptime.netcraft.com/up/graph/?host="..url

    -- LinuxToday home page.
    elseif match (",lt", url) then
        return "http://linuxtoday.com/"

    -- Weather forecast for Melbourne, Australia.
    elseif match (",forecast", url) then
        return "http://www.bom.gov.au/cgi-bin/wrap_fwo.pl?IDF02V00.txt"

    -- Unmatched
    else
        return url
    end
end
-------------------------------------------------------------------------------

(Note that this was noticably enhanced and rewritten in the ELinks standart
hooks.)


Expanding ~ (tilde)
^^^^^^^^^^^^^^^^^^^

By adding an extra snippet of code to the previous example, we can make ELinks
expand pathnames such as `~/foo/bar`
and `~user/zappo`, like in the shell
and other Unix programs.

-------------------------------------------------------------------------------
-- Home directory: If you do not enable system functions, you will need
-- to set the following to your home directory.

home_dir = (getenv and getenv ("HOME")) or "/home/MYSELF"

function goto_url_hook (url, current_url)
                .
                .

    -- Expand ~ to home directories.
    elseif match ("~", url) then
        if strsub(url, 2, 2) == "/" then    -- ~/foo
            return home_dir..strsub(url, 2)
        else                                -- ~foo/bar
            return "/home/"..strsub(url, 2)
        end

                .
                .
-------------------------------------------------------------------------------


Filtering crap
^^^^^^^^^^^^^^

Many web pages nowadays have columns to the left and right of the text, which
are utterly useless.  If you happen to be viewing the page in a 80x25 screen,
the text you want to read ends up crammed into a tiny space in the centre.  We
use ELinks Lua support to manipulate the HTML before it reaches the parser.


linuxtoday.com
++++++++++++++

Note: This recipe is out of date.

Linux Today has two problems when viewed in ELinks: the useless columns on the
left and the right and all the text appears in cyan.  Here is a quick recipe
to fix that:

-------------------------------------------------------------------------------
-- Plain strfind (no metacharacters)
function sstrfind (s, pattern)
    return strfind (s, pattern, 1, 1)
end

function pre_format_html_hook (url, html)
    -- Strip the left and right columns from Linux Today pages
    -- and change the font colour to white.
    if sstrfind (url, "linuxtoday.com") then
        if sstrfind (url, "news_story") then
            html = gsub (html, '<TABLE CELLSPACING="0".-</TABLE>', '', 1)
            html = gsub (html, '<TR BGCOLOR="#FFF.-</TR></TABLE>', '', 1)
        else
            html = gsub (html, 'WIDTH="120">\n<TR.+</TABLE></TD>', '>', 1)
        end
        html = gsub (html, '<A HREF="http://www.internet.com.-</A>', '')
        html = gsub (html, "<IFRAME.-</IFRAME>", "")
        -- emphasis in text is lost
        return gsub (html, 'text="#002244"', 'text="#001133"', 1)
    end

    return nil
end
-------------------------------------------------------------------------------

linuxgames.com
++++++++++++++

Here is a simpler example, for link:http://www.linuxgames.com/[].

-------------------------------------------------------------------------------
function pre_format_html_hook (url, html)
                .
                .

    elseif strfind (url, "linuxgames.com", 1, 1) then
        return gsub (html, "<CENTER>.-</center>", "", 1)

                .
                .
-------------------------------------------------------------------------------


Reading gzipped files
^^^^^^^^^^^^^^^^^^^^^

Note: ELinks already supports gzipped files natively.

Sometimes documents come gzipped in order to save space, but then you need to
uncompress them to read them with ELinks.  Here is a recipe to handle gzipped
files on a Unix system.

-------------------------------------------------------------------------------
-- This script requires system functions.

function pre_format_html_hook (url, html)
                .
                .

    -- Handle gzip'd files within reasonable size.
    if strfind (url, "%.gz$") and strlen (html) < 65536 then
        local tmp = tmpname ()
        writeto (tmp) write (html) writeto ()
        html = pipe_read ("(gzip -dc "..tmp.." || cat "..tmp..") 2>/dev/null")
        remove (tmp)
        return html
    end

                .
                .
-------------------------------------------------------------------------------


Printing
^^^^^^^^

Printing a web page with ELinks usually involves quite a few steps: Save the
current document onto disk.  Run it through ELinks on the command-line (so it
fits into 80 columns) to generate a plain text version.  Remove the 80th
column from the text version, as it will make printers wrap down to the next
line.  Finally, run the processed file through `lpr', then delete it.

The following functions allow you to print web pages directly from ELinks,
using `lpr' or `enscript'.  Type `lpr()` or `enscript()` in the Lua Console to
run them.  (In the `hooks.lua`, I have also made it so you can just type `lpr`
or `enscript`.)

-------------------------------------------------------------------------------
-- This script requires system functions.

function catto (output)
    writeto (output)
    write (current_document_formatted (79))
    writeto ()
end

-- Send the current document to `lpr'.
function lpr ()
    -- You must compile Lua with `popen' support for pipes to work.
    -- See `config' in the Lua distribution.
    catto ("|lpr")
end

-- Send the current document to `enscript'.
function enscript ()
    catto ("|enscript -fCourier8")
end
-------------------------------------------------------------------------------


Deferring to Netscape
^^^^^^^^^^^^^^^^^^^^^

If you come across a brain-dead web page that is totally unreadable with
ELinks, you'd probably want to open it with a graphical browser.  The
following function opens the current document in Netscape.

-------------------------------------------------------------------------------
-- This function requires `execute', a system function.

-- When starting Netscape: Set to `nil' if you do not want
-- to open a new window for each document.
netscape_new_window = 1

-- Open current document in Netscape.
function netscape ()
    local new = netscape_new_window and ",new_window" or ""
    execute ("( netscape -remote 'openURL("..current_url ()..new..")'"
             .." || netscape '"..current_url ().."' ) 2>/dev/null &")
end
-------------------------------------------------------------------------------


Alternative bookmark system
^^^^^^^^^^^^^^^^^^^^^^^^^^^

Many people would like to have a bookmark system with categories (note that
ELinks already supports that, marketing name Hiearchical bookmarks), and also
to be able to view them and search for them in an HTML page.  I have written
an alternative bookmark system (for ELinks), which some people may like better
than the standard bookmark system.  



More ideas
^^^^^^^^^^

 - The Lua interface needs to be redesigned to provide more flexible, coherent
   and usable interface to the scripts.

 - Cliff Cunnington had a neat idea of clipping text that you see in web pages
   (you enter a regexp that will match the start and end of the text you want
   to clip), and saving the text to disk, along with the URL and timestamp.
   This would help if you find that you can't ever remember where you had seen
   a piece of text, or if you want to keep a piece of information but don't
   need to save the entire page.

 - People who use download management programs could write a function to send
   the current link to their favourite downloading program.

 - If you wrote a small C program to put text into the X11 selection
   clipboard, you could pass the current link or URL to that program, to make
   it easier to paste URLs into other windows.  It might be possible to do the
   same with GPM, or the KDE/GNOME equivalents.

 - Send the current page to Babelfish for translation.

 - Look for stupid JavaScript URLs and convert them to something usable.

 - More things are possible, I'm sure.  If you have an idea that requires
   another hook or function, contact me (Peter Wang) and I'll see what I can
   do.