blob: a9d6ea823f445d7e3fd30b8cf82e578dbb8a1b31 [file] [log] [blame]
Georg Brandl8ec7f652007-08-15 14:28:01 +00001
2:mod:`marshal` --- Internal Python object serialization
3=======================================================
4
5.. module:: marshal
6 :synopsis: Convert Python objects to streams of bytes and back (with different
7 constraints).
8
9
10This module contains functions that can read and write Python values in a binary
11format. The format is specific to Python, but independent of machine
12architecture issues (e.g., you can write a Python value to a file on a PC,
13transport the file to a Sun, and read it back there). Details of the format are
14undocumented on purpose; it may change between Python versions (although it
15rarely does). [#]_
16
17.. index::
18 module: pickle
19 module: shelve
20 object: code
21
22This is not a general "persistence" module. For general persistence and
23transfer of Python objects through RPC calls, see the modules :mod:`pickle` and
24:mod:`shelve`. The :mod:`marshal` module exists mainly to support reading and
25writing the "pseudo-compiled" code for Python modules of :file:`.pyc` files.
26Therefore, the Python maintainers reserve the right to modify the marshal format
27in backward incompatible ways should the need arise. If you're serializing and
Raymond Hettinger84e26b62007-10-31 21:57:58 +000028de-serializing Python objects, use the :mod:`pickle` module instead -- the
29performance is comparable, version independence is guaranteed, and pickle
30supports a substantially wider range of objects than marshal.
Georg Brandl8ec7f652007-08-15 14:28:01 +000031
32.. warning::
33
34 The :mod:`marshal` module is not intended to be secure against erroneous or
35 maliciously constructed data. Never unmarshal data received from an
36 untrusted or unauthenticated source.
37
38Not all Python object types are supported; in general, only objects whose value
39is independent from a particular invocation of Python can be written and read by
40this module. The following types are supported: ``None``, integers, long
Raymond Hettinger405884c2007-10-31 22:15:49 +000041integers, floating point numbers, strings, Unicode objects, tuples, lists, sets,
Georg Brandl8ec7f652007-08-15 14:28:01 +000042dictionaries, and code objects, where it should be understood that tuples, lists
43and dictionaries are only supported as long as the values contained therein are
44themselves supported; and recursive lists and dictionaries should not be written
45(they will cause infinite loops).
46
Georg Brandlbf863b12007-08-15 19:06:04 +000047.. warning::
Georg Brandlc62ef8b2009-01-03 20:55:06 +000048
Georg Brandlbf863b12007-08-15 19:06:04 +000049 On machines where C's ``long int`` type has more than 32 bits (such as the
50 DEC Alpha), it is possible to create plain Python integers that are longer
51 than 32 bits. If such an integer is marshaled and read back in on a machine
52 where C's ``long int`` type has only 32 bits, a Python long integer object
53 is returned instead. While of a different type, the numeric value is the
54 same. (This behavior is new in Python 2.2. In earlier versions, all but the
55 least-significant 32 bits of the value were lost, and a warning message was
56 printed.)
Georg Brandl8ec7f652007-08-15 14:28:01 +000057
58There are functions that read/write files as well as functions operating on
59strings.
60
61The module defines these functions:
62
63
64.. function:: dump(value, file[, version])
65
66 Write the value on the open file. The value must be a supported type. The
67 file must be an open file object such as ``sys.stdout`` or returned by
68 :func:`open` or :func:`os.popen`. It must be opened in binary mode (``'wb'``
69 or ``'w+b'``).
70
71 If the value has (or contains an object that has) an unsupported type, a
72 :exc:`ValueError` exception is raised --- but garbage data will also be written
73 to the file. The object will not be properly read back by :func:`load`.
74
75 .. versionadded:: 2.4
76 The *version* argument indicates the data format that ``dump`` should use
77 (see below).
78
79
80.. function:: load(file)
81
82 Read one value from the open file and return it. If no valid value is read
83 (e.g. because the data has a different Python version's incompatible marshal
84 format), raise :exc:`EOFError`, :exc:`ValueError` or :exc:`TypeError`. The
85 file must be an open file object opened in binary mode (``'rb'`` or
86 ``'r+b'``).
87
88 .. warning::
89
90 If an object containing an unsupported type was marshalled with :func:`dump`,
91 :func:`load` will substitute ``None`` for the unmarshallable type.
92
93
94.. function:: dumps(value[, version])
95
96 Return the string that would be written to a file by ``dump(value, file)``. The
97 value must be a supported type. Raise a :exc:`ValueError` exception if value
98 has (or contains an object that has) an unsupported type.
99
100 .. versionadded:: 2.4
101 The *version* argument indicates the data format that ``dumps`` should use
102 (see below).
103
104
105.. function:: loads(string)
106
107 Convert the string to a value. If no valid value is found, raise
108 :exc:`EOFError`, :exc:`ValueError` or :exc:`TypeError`. Extra characters in the
109 string are ignored.
110
111
112In addition, the following constants are defined:
113
114.. data:: version
115
116 Indicates the format that the module uses. Version 0 is the historical format,
117 version 1 (added in Python 2.4) shares interned strings and version 2 (added in
118 Python 2.5) uses a binary format for floating point numbers. The current version
119 is 2.
120
121 .. versionadded:: 2.4
122
123
124.. rubric:: Footnotes
125
126.. [#] The name of this module stems from a bit of terminology used by the designers of
127 Modula-3 (amongst others), who use the term "marshalling" for shipping of data
128 around in a self-contained form. Strictly speaking, "to marshal" means to
129 convert some data from internal to external form (in an RPC buffer for instance)
130 and "unmarshalling" for the reverse process.
131