- C++ 84.5%
- Python 14.9%
- CMake 0.5%
7ff27dfec5
pycdc did not reconstruct `match` statements: class patterns hit the
unhandled MATCH_CLASS opcode and bailed to "# WARNING: Decompyle incomplete",
and value patterns were mis-rendered as an if/elif chain.
This adds reconstruction of the common, statically-recognizable shapes:
* Class patterns - `case Cls(a, b):` - including positional captures and
`_` wildcard sub-patterns (`case Cls(_):`).
* Value patterns - `case 0:` / `case 'x':` - compiled as a COPY-threaded
COMPARE_OP chain rather than MATCH_CLASS.
* The `case _:` wildcard (which carries no test opcode).
A pre-scan over the code object recognizes the simple, guard-free shape of
each case and records it; the MATCH_CLASS handler and a small guarded hook in
COMPARE_OP open ASTMatchBlock/ASTCaseBlock and skip the pattern-test
machinery, and the block-close logic closes each case at its fail-target and
the match at its merge. Anything outside the recognized shape (kwarg patterns,
guards, sequence/mapping patterns) is left unregistered and still bails
honestly, so no incorrect output is produced.
New AST nodes: ASTMatchBlock (subject) and ASTCaseBlock (pattern), rendered
via the existing block-header machinery (type_str + print_src). PycBuffer
gains setPos() so the reconstructor can skip to a case body.
All hooks are guarded by per-offset match tables that are empty for code
without a match statement, so non-match input is unaffected.
Test: tests/input/match_statement.py exercises class patterns, a wildcard
sub-pattern, value patterns and `case _`. Full existing suite still passes.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
||
|---|---|---|
| .github/workflows | ||
| bytes | ||
| scripts | ||
| tests | ||
| .gitignore | ||
| ASTNode.cpp | ||
| ASTNode.h | ||
| ASTree.cpp | ||
| ASTree.h | ||
| bytecode.cpp | ||
| bytecode.h | ||
| bytecode_ops.inl | ||
| CMakeLists.txt | ||
| data.cpp | ||
| data.h | ||
| FastStack.h | ||
| LICENSE | ||
| pyc_code.cpp | ||
| pyc_code.h | ||
| pyc_module.cpp | ||
| pyc_module.h | ||
| pyc_numeric.cpp | ||
| pyc_numeric.h | ||
| pyc_object.cpp | ||
| pyc_object.h | ||
| pyc_sequence.cpp | ||
| pyc_sequence.h | ||
| pyc_string.cpp | ||
| pyc_string.h | ||
| pycdas.cpp | ||
| pycdc.cpp | ||
| README.markdown | ||
Decompyle++
A Python Byte-code Disassembler/Decompiler
Decompyle++ aims to translate compiled Python byte-code back into valid and human-readable Python source code. While other projects have achieved this with varied success, Decompyle++ is unique in that it seeks to support byte-code from any version of Python.
Decompyle++ includes both a byte-code disassembler (pycdas) and a decompiler (pycdc).
As the name implies, Decompyle++ is written in C++. If you wish to contribute, please fork us on github at https://github.com/zrax/pycdc
Building Decompyle++
-
Generate a project or makefile with CMake (See CMake's documentation for details)
- The following options can be passed to CMake to control debug features:
Option Description -DCMAKE_BUILD_TYPE=DebugProduce debugging symbols -DENABLE_BLOCK_DEBUG=ONEnable block debugging output -DENABLE_STACK_DEBUG=ONEnable stack debugging output
- The following options can be passed to CMake to control debug features:
-
Build the generated project or makefile
- For projects (e.g. MSVC), open the generated project file and build it
- For makefiles, just run
make - To run tests (on *nix or MSYS), run
make check JOBS=4(optionalFILTER=xxxxto run only certain tests)
Usage
To run pycdas, the PYC Disassembler:
./pycdas [PATH TO PYC FILE]
The byte-code disassembly is printed to stdout.
To run pycdc, the PYC Decompiler:
./pycdc [PATH TO PYC FILE]
The decompiled Python source is printed to stdout.
Any errors are printed to stderr.
Marshalled code objects:
Both tools support Python marshalled code objects, as output from marshal.dumps(compile(...)).
To use this feature, specify -c -v <version> on the command line - the version must be specified as the objects themselves do not contain version metadata.
Authors, Licence, Credits
Decompyle++ is the work of Michael Hansen and Darryl Pogue.
Additional contributions from:
- charlietang98
- Kunal Parmar
- Olivier Iffrig
- Zlodiy
It is released under the terms of the GNU General Public License, version 3; See LICENSE file for details.