* Add Python extension lowering pass * Add DocstringAttribute * Add extension module codegen * Handle different argument counts efficiently * Add warnings to extension lowering * Fix module name * Fix extension codegen * Fix argument check * Auto-convert Codon exceptions to Python exceptions * Fix #183 * Fix #162; Fix #135 * Fix #155 * Fix CPython interface in codegen * Fix #191 * Fix #187 * Fix #189 * Generate object file in pyext mode * Convert Codon exceptions to Python exceptions * Fix vtable init; Fix failing tests on Linux * Fix #190 * Fix #156 * Fix union routing * Remove need for import python * Automatic @export and wrapping for toplevel functions * Reorganize API * Add Python extension IR structs * Add special calls for no-suspend yield-expr * Add special calls for no-suspend yield-expr * pyextension.h support [wip] * pyextension.h support [wip] * pyextension.h support * pyextension.h support for toplevel functions * clang-format * Add PyFunction::nargs field * Update pyextension codegen (WIP) * SUpport nargs * Add support for @pycapture * PyType codegen (WIP) * Py method codegen (WIP) * Add type ptr hook * Add getset codegen * Add type alloc function * Add type pointer hook codegen * Re-organize codegen * Add member codegen * Update module init codegen * Update module init codegen * Add support for typePtrHook and new to/from_py hooks * Fix extension codegen * Fix init codegen * Fix init codegen; add "tp_new" slot * Fix type hook * Add extra flags * Specialized wrappers (PyType specs) * Add static Python link option * Fix C imports * Add guards * Remove unused field * Python mode only when pyExt set * Update python module * Fix assert * Update codegen/passes * Fix tuple parsing in index expression * Fix empty tuple unification * Do not Cythonize underscore fns * clang-format * Fix switch * Add Py support for cmp/setitem * Add Py support for cmp/setitem * Add type is support * GetSet support * clang-format * GetSet support (fixes) * Avoid useless vtable alloc * Add iter support * Fix size_t capture bug * clang-format * Fix POD type unification with tuples * Add __try_from_py__ API * Fix annotation * Add static reflection methods (setattr; internal.static.*); refactor PyExt to python.codon; handle errors and kwargs in PyExt * Python compat fixes * Update Python object conversions * Fix PyErrors * clang-format; add copyright * Add PyFunction::keywords field * Fix JIT MRO handling; Refactor out Jupyter support * Refactor out Jupyter support * Add support for custom linking args (link=[]) to TOML plugins * Fix tests * Use g++ instead of gcc * Fix Jupyter CMAKE * Fix Jupyter CMAKE * Add _PyArg_Parser definition * Add complex64 type * Add extra complex64 tests * Fix Python calls; add staticenumerate * Fix call * Fix calls * Update pyext wrappers * Fix staticenumerate; Support static calls in tuple() * Fix pyext routing * Add add/mul for tuples * clang-format * Fix pyext codegen * Fix wrap_multiple * Add seq_alloc_atomic_uncollectable * Fix default generics issue * Add binary/ternary ops * Fix missing generic issue * Fix number slots * Update pow * Remove unnecessary pyobj * Fix allocation * Refactor errors * Add test extension * Fix formatting * clang-format * Fix getitem/setitem/delitem in pyext * Fix pyext iterators * Add builtin pow() (fix #294) * Fix #244 * Fix #231 * Fix #229 * Fix #205 * Update docs * Fix error message * Add pyext tests * Add pyext support for @property * Add pyext support for toplevel fns and @tuple classes * More pyext tests * More pyext tests * Fix file error checking * More pyext tests * Update pyext tests * Update docs * Add pyext test to CI * Add pyext support for @tuple.__new__ * Add pyext support for @tuple.__new__ * Fix hetero-tuple issue with fn_overloads * More pyext tests * Bump versions * Fix del magic in pyext * Fix init magic for tuples in pyext * Have test-pypi only run on develop branch * Make exception type indices unnamed-addr * Fix #316; Fix #317 (slash issue) * Use uncollectible-alloc for vtable * Fix #249 * Add pyext docs * Fix #249; Fix clashing vtables; Fix super() and class_copy * Add content-atomic type property instruction * __contents_atomic__ support * Update internal functions * Use PIC when generating Python extension * Cleanup * Add Dockerfile & fix -fPIC * Cleanup * Fix setup.py * Fix pyext fn iteration * Fix CI * clang-format * Update long conversions in Py bridge * Support wide-int to str conversions * Fix test * Add pow for arbitrary-width ints * Fix Linux backtraces * Cleanup * Add more tests * Fix docs; Remove tuple.__add__ for scalars * Update docs --------- Co-authored-by: Ibrahim Numanagić <ibrahimpasa@gmail.com>
2.5 KiB
Calling Python from Codon is possible in two ways:
from python import
allows importing and calling Python functions from existing Python modules.@python
allows writing Python code directly in Codon.
In order to use these features, the CODON_PYTHON
environment variable
must be set to the appropriate Python shared library:
export CODON_PYTHON=/path/to/libpython.X.Y.so
For example, with a brew
-installed Python 3.9 on macOS, this might be
/usr/local/opt/python@3.9/Frameworks/Python.framework/Versions/3.9/lib/libpython3.9.dylib
Note that only Python versions 3.6 and later are supported.
from python import
Let's say we have a Python function defined in mymodule.py:
def multiply(a, b):
return a * b
We can call this function in Codon using from python import
and
indicating the appropriate call and return types:
from python import mymodule.multiply(int, int) -> int
print(multiply(3, 4)) # 12
(Be sure the PYTHONPATH
environment variable includes the path of
mymodule.py!)
from python import
does not need to specify explicit types, in which case
Codon will operate directly on the Python objects, and convert Codon types
to Python types as necessary:
from python import numpy as np # Codon will call NumPy through CPython's API
x = np.array([1, 2, 3, 4]) * 10
print(x) # [10 20 30 40]
@python
Codon programs can contain functions that will be executed by Python via
pydef
:
@python
def multiply(a: int, b: int) -> int:
return a * b
print(multiply(3, 4)) # 12
This makes calling Python modules like NumPy very easy:
@python
def myrange(n: int) -> List[int]:
from numpy import arange
return list(arange(n))
print(myrange(5)) # [0, 1, 2, 3, 4]
Data conversions
Codon uses two new magic methods to transfer data to and from Python:
__to_py__
: Produces a Python object (PyObject*
in C) given a Codon object.__from_py__
: Produces a Codon object given a Python object.
import python # needed to initialize the Python runtime
o = (42).__to_py__() # type of 'o' is 'cobj', equivalent to a pointer in C
print(o) # 0x100e00610
n = int.__from_py__(o)
print(n) # 42
Codon stores the results of __to_py__
calls by wrapping them in an instance of
a new class called pyobj
, which correctly handles the underlying Python object's
reference count. All operations on pyobj
s then go through CPython's API.