Doc/lib/libmarshal.tex

   1 \section{\module{marshal} ---
   2          Internal Python object serialization}
   3
   4 \declaremodule{builtin}{marshal}
   5 \modulesynopsis{Convert Python objects to streams of bytes and back
   6                 (with different constraints).}
   7
   8
   9 This module contains functions that can read and write Python
  10 values in a binary format.  The format is specific to Python, but
  11 independent of machine architecture issues (e.g., you can write a
  12 Python value to a file on a PC, transport the file to a Sun, and read
  13 it back there).  Details of the format are undocumented on purpose;
  14 it may change between Python versions (although it rarely
  15 does).\footnote{The name of this module stems from a bit of
  16   terminology used by the designers of Modula-3 (amongst others), who
  17   use the term ``marshalling'' for shipping of data around in a
  18   self-contained form. Strictly speaking, ``to marshal'' means to
  19   convert some data from internal to external form (in an RPC buffer for
  20   instance) and ``unmarshalling'' for the reverse process.}
  21
  22 This is not a general ``persistence'' module.  For general persistence
  23 and transfer of Python objects through RPC calls, see the modules
  24 \refmodule{pickle} and \refmodule{shelve}.  The \module{marshal} module exists
  25 mainly to support reading and writing the ``pseudo-compiled'' code for
  26 Python modules of \file{.pyc} files.  Therefore, the Python
  27 maintainers reserve the right to modify the marshal format in backward
  28 incompatible ways should the need arise.  If you're serializing and
  29 de-serializing Python objects, use the \module{pickle} module instead.
  30 \refstmodindex{pickle}
  31 \refstmodindex{shelve}
  32 \obindex{code}
  33
  34 \begin{notice}[warning]
  35 The \module{marshal} module is not intended to be secure against
  36 erroneous or maliciously constructed data.  Never unmarshal data
  37 received from an untrusted or unauthenticated source.
  38 \end{notice}
  39
  40 Not all Python object types are supported; in general, only objects
  41 whose value is independent from a particular invocation of Python can
  42 be written and read by this module.  The following types are supported:
  43 \code{None}, integers, long integers, floating point numbers,
  44 strings, Unicode objects, tuples, lists, dictionaries, and code
  45 objects, where it should be understood that tuples, lists and
  46 dictionaries are only supported as long as the values contained
  47 therein are themselves supported; and recursive lists and dictionaries
  48 should not be written (they will cause infinite loops).
  49
  50 \strong{Caveat:} On machines where C's \code{long int} type has more than
  51 32 bits (such as the DEC Alpha), it is possible to create plain Python
  52 integers that are longer than 32 bits.
  53 If such an integer is marshaled and read back in on a machine where
  54 C's \code{long int} type has only 32 bits, a Python long integer object
  55 is returned instead.  While of a different type, the numeric value is
  56 the same.  (This behavior is new in Python 2.2.  In earlier versions,
  57 all but the least-significant 32 bits of the value were lost, and a
  58 warning message was printed.)
  59
  60 There are functions that read/write files as well as functions
  61 operating on strings.
  62
  63 The module defines these functions:
  64
  65 \begin{funcdesc}{dump}{value, file\optional{, version}}
  66   Write the value on the open file.  The value must be a supported
  67   type.  The file must be an open file object such as
  68   \code{sys.stdout} or returned by \function{open()} or
  69   \function{posix.popen()}.  It must be opened in binary mode
  70   (\code{'wb'} or \code{'w+b'}).
  71
  72   If the value has (or contains an object that has) an unsupported type,
  73   a \exception{ValueError} exception is raised --- but garbage data
  74   will also be written to the file.  The object will not be properly
  75   read back by \function{load()}.
  76
  77   \versionadded[The \var{version} argument indicates the data
  78   format that \code{dump} should use (see below)]{2.4}
  79 \end{funcdesc}
  80
  81 \begin{funcdesc}{load}{file}
  82   Read one value from the open file and return it.  If no valid value
  83   is read, raise \exception{EOFError}, \exception{ValueError} or
  84   \exception{TypeError}.  The file must be an open file object opened
  85   in binary mode (\code{'rb'} or \code{'r+b'}).
  86
  87   \warning{If an object containing an unsupported type was
  88   marshalled with \function{dump()}, \function{load()} will substitute
  89   \code{None} for the unmarshallable type.}
  90 \end{funcdesc}
  91
  92 \begin{funcdesc}{dumps}{value\optional{, version}}
  93   Return the string that would be written to a file by
  94   \code{dump(\var{value}, \var{file})}.  The value must be a supported
  95   type.  Raise a \exception{ValueError} exception if value has (or
  96   contains an object that has) an unsupported type.
  97
  98   \versionadded[The \var{version} argument indicates the data
  99   format that \code{dumps} should use (see below)]{2.4}
 100 \end{funcdesc}
 101
 102 \begin{funcdesc}{loads}{string}
 103   Convert the string to a value.  If no valid value is found, raise
 104   \exception{EOFError}, \exception{ValueError} or
 105   \exception{TypeError}.  Extra characters in the string are ignored.
 106 \end{funcdesc}
 107
 108 In addition, the following constants are defined:
 109
 110 \begin{datadesc}{version}
 111   Indicates the format that the module uses. Version 0 is the
 112   historical format, version 1 (added in Python 2.4) shares interned
 113   strings and version 2 (added in Python 2.5) uses a binary format for
 114   floating point numbers. The current version is 2.
 115
 116   \versionadded{2.4}
 117 \end{datadesc}