Skip to main content

Pythonic C-style structs for parsing binary data

Project description


deconstruct provides a Pythonic analogue of C's struct, primarily for the purpose of interpreting (i.e. deconstructing) contiguous binary data.

Internally, deconstruct uses Python's struct module and can be considered an abstraction of sorts. struct (the module) can be frustrating to use: its format strings appear arcane and furthermore separate the description of the data from its representation, a definite strength of C's struct.

In contrast, deconstruct allows structs to be defined and used using a syntax that is Pythonic while maintaining close correspondence to C.


With deconstruct, the struct definition (adapted from input.h):

#include <stdint.h>

struct input_event {  
    uint64_t time[2];
    int16_t type;
    int16_t code;
    int32_t value;

can be represented as:

import deconstruct as c

class InputEvent(c.Struct):
    time: c.uint64[2]
    type: c.int16
    code: c.int16
    value: c.int32

This definition can be used to interpret and then access binary data:

>>> buffer = b'Some  arbitrary  buffer!'
>>> event = InputEvent(buffer)
>>> event.code
>>> event.time
(8241904116577431379, 2340027244253309282)
>>> print(event)
struct InputEvent [ByteOrder.NATIVE, TypeWidth.STANDARD] {
    time: uint64 = (8241904116577431379, 2340027244253309282)
    type: int16 = 25120
    code: int16 = 26229
    value: int32 = 561145190

Of course, in reality the buffer passed in is more likely to come from something more useful, like a file. Notice that fixed-size, n-dimensional arrays can be specified using the syntax type[length], a further improvement on Python's struct.


deconstruct is now on PyPI:

pip install deconstruct

Alternatively you can install straight from this repository:

pip install

Built wheels are also available under Releases, as is a changelog. The latest release is version 0.5.

deconstruct has no dependencies but requires Python >= 3.6 as it makes use of the class annotation syntax added in that release (see PEP 526).

API listing

Struct(buffer: bytes)

Subclass this to define your own structs. Subclasses should only declare fields of C types defined in this package.

When you instantiate your Struct with a bytes-like object, deconstruct creates a format string and uses it to unpack that buffer. In the instance, C types will be replaced with their equivalent Python types for use (e.g. bytes for char, int for schar and float for double). All types available for use in struct field definitions and their details are documented in the table below.


Name Type Description
__byte_order__ ByteOrder Set this in your subclass definition to define the byte order used when unpacking the struct. One of:
  • ByteOrder.NATIVE (default value)
  • ByteOrder.BIG_ENDIAN
__type_width__ TypeWidth Set this in your subclass definition to define the type width and padding used for the struct. One of:
  • TypeWidth.NATIVE
  • TypeWidth.STANDARD (default value)
When TypeWidth.NATIVE is set, the struct will use native type widths and alignment. When TypeWidth.STANDARD is used, the struct will use Python's struct's "standard" widths1 and no padding.

Note that TypeWidth.NATIVE can only be used with ByteOrder.NATIVE. This is a limitation of Python's struct.


Name Type Description
format_string str The format string for this struct
sizeof int The total size in bytes of the struct. Equivalent to C's sizeof


Signature Return type Description
|`to_bytes()`    |`bytes`    |Returns the in-memory ("packed") representation of this struct instance|

You can also print Struct instances for easier debugging and compare them using the == operator.

C types

deconstruct defines the following special types for use in Struct definitions:2

deconstruct type C99 type Python format character "Standard" width (bytes)1 Resolves to Python type
char char c 1 bytes of length 1
schar signed char b 1 int
uchar unsigned char B 1 int
short short h 2 int
ushort unsigned short H 2 int
int int i 2 int
uint unsigned int I 2 int
long long l 4 int
ulong unsigned long L 4 int
longlong long long q 8 int
ulonglong unsigned long long Q 8 int
bool bool (_Bool) ? 1 bool
float float f 4 float
double double d 8 float
int8 int8_t b* 1 int
uint8 uint8_t B* 1 int
int16 int16_t h* 2 int
uint16 uint16_t H* 2 int
int32 int32_t l* 4 int
uint32 uint32_t L* 4 int
int64 int64_t q* 8 int
uint64 uint64_t Q* 8 int
ptr void*/intptr_t/uintptr_t P N/A** int
size size_t n N/A** int
ssize ssize_t N N/A** int
* format character with __type_width__ = TypeWidth.STANDARD - platform specific otherwise.
** only available with __type_width__ = TypeWidth.NATIVE.


As mentioned earlier, all the types above support a type[length] syntax to define arrays. Multidimensional arrays work as you would expect, with int[2][2] declaring a 2-D array of type int and total length 4. When a Struct is used to unpack a buffer, each array will resolve to a tuple (or in the case of a multidimensional array, a nested tuple) of their equivalent Python types, as documented in the table above. The only exception to this is char, an array of which will be automatically concatenated to a single bytes object (if this behaviour is undesirable, use schar or uchar instead).

1. Python's struct has the concept of "standard" type sizes. This is somewhat confusing coming from C as its standards go to some length not to define a standard ABI. However, as this terminology is so fundamental to the documentation of Python's struct it is replicated here for simplicity's sake. These sizes correspond with the minimum sizes implied for C's types.

2. Because some of these conflict with Python's primitives, it is not recommended to import * from deconstruct as this will severely pollute your namespace (in fact this is a bad idea in general). I like to import deconstruct as c as shown above.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for deconstruct, version 0.5
Filename, size File type Python version Upload date Hashes
Filename, size deconstruct-0.5-py3-none-any.whl (14.7 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size deconstruct-0.5.tar.gz (7.9 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page