Georg Brandl | 79e3d55 | 2008-01-19 22:14:27 +0000 | [diff] [blame] | 1 | .. highlightlang:: c |
| 2 | |
| 3 | .. _marshalling-utils: |
| 4 | |
| 5 | Data marshalling support |
| 6 | ======================== |
| 7 | |
Jeroen Ruigrok van der Werven | a4c03ab | 2009-04-25 20:40:10 +0000 | [diff] [blame] | 8 | These routines allow C code to work with serialized objects using the same |
| 9 | data format as the :mod:`marshal` module. There are functions to write data |
| 10 | into the serialization format, and additional functions that can be used to |
| 11 | read the data back. Files used to store marshalled data must be opened in |
| 12 | binary mode. |
Georg Brandl | 79e3d55 | 2008-01-19 22:14:27 +0000 | [diff] [blame] | 13 | |
| 14 | Numeric values are stored with the least significant byte first. |
| 15 | |
Jeroen Ruigrok van der Werven | a4c03ab | 2009-04-25 20:40:10 +0000 | [diff] [blame] | 16 | The module supports two versions of the data format: version 0 is the |
| 17 | historical version, version 1 (new in Python 2.4) shares interned strings in |
| 18 | the file, and upon unmarshalling. Version 2 (new in Python 2.5) uses a binary |
| 19 | format for floating point numbers. *Py_MARSHAL_VERSION* indicates the current |
| 20 | file format (currently 2). |
Georg Brandl | 79e3d55 | 2008-01-19 22:14:27 +0000 | [diff] [blame] | 21 | |
| 22 | |
| 23 | .. cfunction:: void PyMarshal_WriteLongToFile(long value, FILE *file, int version) |
| 24 | |
Jeroen Ruigrok van der Werven | a4c03ab | 2009-04-25 20:40:10 +0000 | [diff] [blame] | 25 | Marshal a :ctype:`long` integer, *value*, to *file*. This will only write |
| 26 | the least-significant 32 bits of *value*; regardless of the size of the |
| 27 | native :ctype:`long` type. |
Georg Brandl | 79e3d55 | 2008-01-19 22:14:27 +0000 | [diff] [blame] | 28 | |
| 29 | .. versionchanged:: 2.4 |
| 30 | *version* indicates the file format. |
| 31 | |
| 32 | |
| 33 | .. cfunction:: void PyMarshal_WriteObjectToFile(PyObject *value, FILE *file, int version) |
| 34 | |
| 35 | Marshal a Python object, *value*, to *file*. |
| 36 | |
| 37 | .. versionchanged:: 2.4 |
| 38 | *version* indicates the file format. |
| 39 | |
| 40 | |
| 41 | .. cfunction:: PyObject* PyMarshal_WriteObjectToString(PyObject *value, int version) |
| 42 | |
| 43 | Return a string object containing the marshalled representation of *value*. |
| 44 | |
| 45 | .. versionchanged:: 2.4 |
| 46 | *version* indicates the file format. |
| 47 | |
| 48 | |
| 49 | The following functions allow marshalled values to be read back in. |
| 50 | |
| 51 | XXX What about error detection? It appears that reading past the end of the |
Jeroen Ruigrok van der Werven | a4c03ab | 2009-04-25 20:40:10 +0000 | [diff] [blame] | 52 | file will always result in a negative numeric value (where that's relevant), |
| 53 | but it's not clear that negative values won't be handled properly when there's |
| 54 | no error. What's the right way to tell? Should only non-negative values be |
| 55 | written using these routines? |
Georg Brandl | 79e3d55 | 2008-01-19 22:14:27 +0000 | [diff] [blame] | 56 | |
| 57 | |
| 58 | .. cfunction:: long PyMarshal_ReadLongFromFile(FILE *file) |
| 59 | |
Jeroen Ruigrok van der Werven | a4c03ab | 2009-04-25 20:40:10 +0000 | [diff] [blame] | 60 | Return a C :ctype:`long` from the data stream in a :ctype:`FILE\*` opened |
| 61 | for reading. Only a 32-bit value can be read in using this function, |
| 62 | regardless of the native size of :ctype:`long`. |
Georg Brandl | 79e3d55 | 2008-01-19 22:14:27 +0000 | [diff] [blame] | 63 | |
| 64 | |
| 65 | .. cfunction:: int PyMarshal_ReadShortFromFile(FILE *file) |
| 66 | |
Jeroen Ruigrok van der Werven | a4c03ab | 2009-04-25 20:40:10 +0000 | [diff] [blame] | 67 | Return a C :ctype:`short` from the data stream in a :ctype:`FILE\*` opened |
| 68 | for reading. Only a 16-bit value can be read in using this function, |
| 69 | regardless of the native size of :ctype:`short`. |
Georg Brandl | 79e3d55 | 2008-01-19 22:14:27 +0000 | [diff] [blame] | 70 | |
| 71 | |
| 72 | .. cfunction:: PyObject* PyMarshal_ReadObjectFromFile(FILE *file) |
| 73 | |
| 74 | Return a Python object from the data stream in a :ctype:`FILE\*` opened for |
| 75 | reading. On error, sets the appropriate exception (:exc:`EOFError` or |
| 76 | :exc:`TypeError`) and returns *NULL*. |
| 77 | |
| 78 | |
| 79 | .. cfunction:: PyObject* PyMarshal_ReadLastObjectFromFile(FILE *file) |
| 80 | |
| 81 | Return a Python object from the data stream in a :ctype:`FILE\*` opened for |
Jeroen Ruigrok van der Werven | a4c03ab | 2009-04-25 20:40:10 +0000 | [diff] [blame] | 82 | reading. Unlike :cfunc:`PyMarshal_ReadObjectFromFile`, this function |
| 83 | assumes that no further objects will be read from the file, allowing it to |
| 84 | aggressively load file data into memory so that the de-serialization can |
| 85 | operate from data in memory rather than reading a byte at a time from the |
| 86 | file. Only use these variant if you are certain that you won't be reading |
| 87 | anything else from the file. On error, sets the appropriate exception |
| 88 | (:exc:`EOFError` or :exc:`TypeError`) and returns *NULL*. |
Georg Brandl | 79e3d55 | 2008-01-19 22:14:27 +0000 | [diff] [blame] | 89 | |
| 90 | |
| 91 | .. cfunction:: PyObject* PyMarshal_ReadObjectFromString(char *string, Py_ssize_t len) |
| 92 | |
Jeroen Ruigrok van der Werven | a4c03ab | 2009-04-25 20:40:10 +0000 | [diff] [blame] | 93 | Return a Python object from the data stream in a character buffer |
| 94 | containing *len* bytes pointed to by *string*. On error, sets the |
| 95 | appropriate exception (:exc:`EOFError` or :exc:`TypeError`) and returns |
| 96 | *NULL*. |
Jeroen Ruigrok van der Werven | be82d2f | 2009-04-25 20:41:40 +0000 | [diff] [blame] | 97 | |
| 98 | .. versionchanged:: 2.5 |
| 99 | This function used an :ctype:`int` type for *len*. This might require |
| 100 | changes in your code for properly supporting 64-bit systems. |