ENH: Meson build system support for F2PY's `f2py -c` command by NamamiShanker · Pull Request #22225 · numpy/numpy

NamamiShanker · 2022-09-07T18:49:05Z

This PR is a continuation of my older PR for implementing F2PY's fronted with argparse.

My blog explains the core ideas behind the changes.

This PR introduces a new flag for F2PY - --backend which can accept build system for building Python extension modules.

An example to produce a Python module out of the classic fibonacci sequence Fortran program will be:-

f2py -m fib --lower -c --backend=meson fibfortran.f

The output thus produced works correctly:

>>> import fib
>>> print(fib.fib.__doc__)
a = fib(n)

Wrapper for ``fib``.

Parameters
----------
n : input int

Returns
-------
a : rank-1 array('d') with bounds (n)

>>> print(fib.fib(8))
[ 0.  1.  1.  2.  3.  5.  8. 13.]
>>>

The default value of --backend flag is distutils making this feature completely backwards compatible.

This reverts commit b334020.

rgommers · 2022-09-09T08:01:18Z

numpy/f2py/backends/meson_backend.py

+from string import Template
+
+class MesonTemplate:
+	"""Template meson build file generation class."""


There's tabs here instead of spaces. @NamamiShanker can you please check your editor settings and ensure tabs are always converted to spaces?

Sure @rgommers. I corrected this file and the others.

rgommers

Looks like nice progress, thanks @NamamiShanker. I just had a quick look through the build system part of this and made a few comments for things that stood out to me.

rgommers · 2022-09-09T08:02:10Z

numpy/f2py/backends/src/meson.build.src

+py3 = py_mod.find_installation('python3')
+py3_dep = py3.dependency()
+message(py3.path())
+message(py3.get_install_dir())


This should have pure: false as argument.

rgommers · 2022-09-09T08:04:22Z

numpy/f2py/backends/src/meson.build.src

@@ -0,0 +1,22 @@
+project('${modulename}', 'c',
+  version : '0.1',


I think you want to set the minimum meson version here, and it should be very recent; lots of Python bugs were fixed recently. I recommend the most recent release, 0.63.2, and bumping it up to 0.63.3/0.64.0 when those become available.

Corrected it.

rgommers · 2022-09-09T08:06:07Z

numpy/f2py/setup.py

 def configuration(parent_package='', top_path=None):
    config = Configuration('f2py', parent_package, top_path)
    config.add_subpackage('tests')
+    config.add_subpackage('backends')


This is only needed if there's a backends/setup.py.

I am not much familiar with distutils configuration files. Would it be more apt to use the add_data_dir function instead? With the current configuration, the backend folder will not be included in the installed numpy/f2py folder.

add_data_dir is more appropriate. However, after looking at how it's done elsewhere, add_subpackage is used more often (even if it gives a warning). So you can leave it as is.

rgommers · 2022-09-09T08:06:37Z

numpy/f2py/setup.py

    config = Configuration('f2py', parent_package, top_path)
    config.add_subpackage('tests')
+    config.add_subpackage('backends')
+    config.add_data_dir('backends/src')


I don't think this does anything, the next line covers it already. You should need one or the other I believe.

rgommers · 2022-09-09T08:11:14Z

numpy/tests/test_public_api.py

    "distutils.unixccompiler",
    "dual",
    "f2py.auxfuncs",
+    "f2py.backends",


This does not look like it should be public, or should it? If not, can you please name it f2py._backends?

And the same for all the other things - nothing should be added to this list ideally.

rgommers · 2022-09-09T08:17:01Z

numpy/f2py/backends/meson_backend.py

+		if opt_flags is None:
+			opt_flags = []
+		if arch_flags is None:
+			arch_flags = []


These flags don't quite fit anymore (the difference between F77 and F90 is not needed, nor are the exact paths to the compiler; the --opt, --arch etc. are specified in a very non-standard way, Meson has a better way to do things like debug flags). Is there a plan to clean this up? Maybe they should be kept unchanged for now, and then when numpy.distutils is gone they can be updated?

jmrohwer · 2022-09-19T11:05:28Z

I am interested in this because I am porting the build system of another package (PySCeS) to try and get rid of the distutils and numpy.distutils dependencies. The package has two Fortran extension modules. Refer to #22211

This PR seems to provide just what I need so I tried only compiling the Fortran extension modules on the command line. This is with numpy installed from NamamiShanker:main. The relevant code is here: https://github.com/PySCeS/pysces/tree/main/pysces/nleq2

Using f2py with the default numpy.distutils backend, works fine. However, the meson backend fails.

Pertinent versions:

Python 3.10 on linux
numpy 1.23.0.dev0+2144.gcf21444b6
meson 0.63.2

Navigating to the nleq2 source directory, and executing

f2py -c nleq2.pyf nleq2.f wnorm.f linalg_nleq2.f zibconst.f zibmon.f zibsec.f

works fine and the *.so binary module file is created, and can be imported from Python.

However, executing the following fails:

f2py -c --backend meson nleq2.pyf nleq2.f wnorm.f linalg_nleq2.f zibconst.f zibmon.f zibsec.f

The traceback is too big to include inline, I attach it as a file.
traceback.txt

We use a custom *.pyf file because there is only a very small subset of the Fortran routines in the code that we want to expose in Python.

Trying to debug this, I used a custom --build-dir argument to the f2py command above, so that the build files would be left for inspection. The generated nleq2module.c file for the meson build was much bigger (271kb) than that for the numpy.distutils build (40.4kb). This makes me think that the meson build is trying to generate interfaces for all the routines in the Fortran code, and that the custom *.pyf file is not fully honored.

rgommers · 2023-08-18T09:44:03Z

@HaoZeke @NamamiShanker we're now close to the 1.26.0 release, and this is still pending. I think we do need this for Python 3.12 support of using f2py -c, right?

mattip · 2023-08-18T10:28:23Z

I overhauled the random example cython compilation to use meson in #24153, maybe worth seeing if anything there is applicable here.

HaoZeke · 2023-08-22T10:54:42Z

I'll aim to have this and the previous dependent PR in before the weekend. Should be followed up to ensure most projects start using it (and submit bug reports if needed).

Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com>

HaoZeke · 2023-08-25T04:22:39Z

I am interested in this because I am porting the build system of another package (PySCeS) to try and get rid of the distutils and numpy.distutils dependencies. The package has two Fortran extension modules. Refer to #22211

This PR seems to provide just what I need so I tried only compiling the Fortran extension modules on the command line. This is with numpy installed from NamamiShanker:main. The relevant code is here: https://github.com/PySCeS/pysces/tree/main/pysces/nleq2

Using f2py with the default numpy.distutils backend, works fine. However, the meson backend fails.

Pertinent versions:
* Python 3.10 on linux

* numpy 1.23.0.dev0+2144.gcf21444b6

* meson 0.63.2
Navigating to the nleq2 source directory, and executing
f2py -c nleq2.pyf nleq2.f wnorm.f linalg_nleq2.f zibconst.f zibmon.f zibsec.f
works fine and the *.so binary module file is created, and can be imported from Python.

However, executing the following fails:
f2py -c --backend meson nleq2.pyf nleq2.f wnorm.f linalg_nleq2.f zibconst.f zibmon.f zibsec.f
The traceback is too big to include inline, I attach it as a file. traceback.txt

We use a custom *.pyf file because there is only a very small subset of the Fortran routines in the code that we want to expose in Python.

Trying to debug this, I used a custom --build-dir argument to the f2py command above, so that the build files would be left for inspection. The generated nleq2module.c file for the meson build was much bigger (271kb) than that for the numpy.distutils build (40.4kb). This makes me think that the meson build is trying to generate interfaces for all the routines in the Fortran code, and that the custom *.pyf file is not fully honored.

This sounds like a frontend (f2pyarg) issue. Could you check with #24532?

HaoZeke

There's a lot of excellent stuff here, but perhaps changes might be best done on #24532..

HaoZeke · 2023-08-25T04:24:36Z

numpy/f2py/backends/meson_backend.py

+	def macros_substitution(self) -> None:
+		self.substitutions["macros"] = ""
+		if self.define_macros:
+			self.substitutions["macros"] = ",".join(f"\'-D{macro[0]}={macro[1]}\'" if macro[1] else f"-D{macro[0]}" for macro in self.define_macros)
+		if self.undef_macros:
+			self.substitutions["macros"] += "," + ",".join(f"\'-U{macro}\'" for macro in self.undef_macros)


These don't pass -D flags to the underlying preprocessor, that would have to be via Dc_args=-D{macro[0]}={macro[1]}... (e.g. seen here)

Similar issue with unsetting the defines.

HaoZeke · 2023-08-25T04:25:16Z

numpy/f2py/backends/meson_backend.py

+	def global_flags_substitution(self) -> None:
+		fortran_compiler_flags = self.fortran_flags + self.optimization_flags + self.architecture_flags
+		c_compiler_flags = self.optimization_flags + self.architecture_flags
+		self.substitutions["fortran_global_args"] = fortran_compiler_flags
+		self.substitutions["c_global_args"] = c_compiler_flags


Global flags are generally a bad idea, especially since a lot of these might be used as subprojects, it makes more sense to populate the _project variants.

HaoZeke · 2023-08-25T04:26:21Z

numpy/f2py/backends/meson_backend.py

+		self.substitutions['numpy_get_include'] = self.numpy_get_include.absolute()
+		self.substitutions['f2py_get_include'] = self.f2py_get_include.absolute()


Are these necessary? The documentation includes the option to get these dynamically from meson...

HaoZeke · 2023-08-25T04:34:05Z

numpy/f2py/backends/meson_backend.py

+	def linker_substitution(self) -> None:
+		self.substitutions["linker_args"] = ""
+		if self.linker_libpath:
+			linker_libpath_subs = ",".join(f"-L{libpath}" for libpath in self.linker_libpath)
+			self.substitutions["linker_args"] += linker_libpath_subs
+		if self.linker_libname:
+			linker_libname_subs = ",".join(f"-l{libname}" for libname in self.linker_libname)
+			self.substitutions["linker_args"] += f",{linker_libname_subs}"


Same issue as above.

jmrohwer · 2023-08-29T10:51:49Z

I am interested in this because I am porting the build system of another package (PySCeS) to try and get rid of the distutils and numpy.distutils dependencies. The package has two Fortran extension modules. Refer to #22211
This PR seems to provide just what I need so I tried only compiling the Fortran extension modules on the command line. This is with numpy installed from NamamiShanker:main. The relevant code is here: https://github.com/PySCeS/pysces/tree/main/pysces/nleq2
Using f2py with the default numpy.distutils backend, works fine. However, the meson backend fails.
Pertinent versions:
* Python 3.10 on linux

* numpy 1.23.0.dev0+2144.gcf21444b6

* meson 0.63.2
Navigating to the nleq2 source directory, and executing
f2py -c nleq2.pyf nleq2.f wnorm.f linalg_nleq2.f zibconst.f zibmon.f zibsec.f
works fine and the *.so binary module file is created, and can be imported from Python.
However, executing the following fails:
f2py -c --backend meson nleq2.pyf nleq2.f wnorm.f linalg_nleq2.f zibconst.f zibmon.f zibsec.f
The traceback is too big to include inline, I attach it as a file. traceback.txt
We use a custom *.pyf file because there is only a very small subset of the Fortran routines in the code that we want to expose in Python.
Trying to debug this, I used a custom --build-dir argument to the f2py command above, so that the build files would be left for inspection. The generated nleq2module.c file for the meson build was much bigger (271kb) than that for the numpy.distutils build (40.4kb). This makes me think that the meson build is trying to generate interfaces for all the routines in the Fortran code, and that the custom *.pyf file is not fully honored.
This sounds like a frontend (f2pyarg) issue. Could you check with #24532?

Works with #24532 , thanks!
One thing that I did pick up though is that it only worked in a develop install. When I installed numpy in a new venv using python setup.py build -j 4 install, I got the following error:

Traceback (most recent call last):
  File "/home/jr/.virtualenvs/numpy-test/bin/f2py", line 33, in <module>
    sys.exit(load_entry_point('numpy==2.0.0.dev0', 'console_scripts', 'f2py')())
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jr/.virtualenvs/numpy-test/lib/python3.11/site-packages/numpy/f2py/f2py2e.py", line 732, in main
    run_compile()
  File "/home/jr/.virtualenvs/numpy-test/lib/python3.11/site-packages/numpy/f2py/f2py2e.py", line 705, in run_compile
    builder.compile()
  File "/home/jr/.virtualenvs/numpy-test/lib/python3.11/site-packages/numpy/f2py/_backends/_meson.py", line 130, in compile
    self.write_meson_build(self.build_dir)
  File "/home/jr/.virtualenvs/numpy-test/lib/python3.11/site-packages/numpy/f2py/_backends/_meson.py", line 108, in write_meson_build
    src = meson_template.generate_meson_build()
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jr/.virtualenvs/numpy-test/lib/python3.11/site-packages/numpy/f2py/_backends/_meson.py", line 71, in generate_meson_build
    template = Template(self.meson_build_template())
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jr/.virtualenvs/numpy-test/lib/python3.11/site-packages/numpy/f2py/_backends/_meson.py", line 44, in meson_build_template
    raise FileNotFoundError(
FileNotFoundError: [Errno 2] Meson build template /home/jr/.virtualenvs/numpy-test/lib/python3.11/site-packages/numpy/f2py/_backends/meson.build.template does not exist.

Looking in the site-packages directory, I noticed that the file numpy/f2py/_backends/meson.build.template was missing. I guess it needs to be specified in MANIFEST.in or whatever mechanism the numpy build uses to specify additional package data files.

Should address numpy#22225 (comment)

Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com>

Should address numpy#22225 (comment)

* FIX: Import f2py2e rather than f2py for run_main * FIX: Import f2py2e instead of f2py * ENH: Add F2PY back-end work from gh-22225 Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com> * ENH: Add meson skeleton from gh-2225 Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com> * MAINT: Trim backend.py down to f2py2e flags * ENH: Add a factory function for backends * ENH: Add a distutils backend * ENH: Handle --backends in f2py Defaults to distutils for now * DOC: Add some minor comments in f2py2e * MAINT: Refactor and rework meson.build.src * MAINT: Add objects * MAINT: Cleanup distutils backend * MAINT: Refactor to add everything back to backend Necessary for the meson.build for now. Refactors / cleanup needs better argument handling in f2py2e * MAINT: Fix overly long line * BUG: Construct wrappers for meson backend * MAINT: Rework, simplify template massively * ENH: Truncate meson.build to skeleton only * MAINT: Minor backend housekeeping, name changes * MAINT: Less absolute paths, update setup.py [f2py] * MAINT: Move f2py module name functionality Previously part of np.distutils * ENH: Handle .pyf files * TST: Fix typo in isoFortranEnvMap.f90 * MAINT: Typo in f2py2e support for pyf files * DOC: Add release note for --backend * MAINT: Conditional switch for Python 3.12 [f2py] * MAINT: No absolute paths in backend [f2py-meson] The files are copied over anyway, this makes it easier to extend the generated skeleton * MAINT: Prettier generated meson.build files [f2py] * ENH: Add meson's dependency(blah) to f2py * DOC: Document the new flag * MAINT: Simplify and rename backend template [f2py] Co-authored-by: rgommers <rgommers@users.noreply.github.com> * ENH: Support build_type via --debug [f2py-meson] * MAINT,DOC: Reduce warn,rework doc [f2py-meson] Co-authored-by: rgommers <rgommers@users.noreply.github.com> * ENH: Rework deps: to --dep calls [f2py-meson] Also shows how incremental updates to the parser can be done. * MAINT,DOC: Add --backend to argparse, add docs * MAINT: Rename meson template [f2py-meson] * MAINT: Add meson.build for f2py Should address #22225 (comment) * BLD: remove duplicate f2py handling in meson.build files --------- Co-authored-by: Namami Shanker <namami2011@gmail.com> Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com> Co-authored-by: rgommers <rgommers@users.noreply.github.com> Co-authored-by: Ralf Gommers <ralf.gommers@gmail.com>

* FIX: Import f2py2e rather than f2py for run_main * FIX: Import f2py2e instead of f2py * ENH: Add F2PY back-end work from numpygh-22225 Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com> * ENH: Add meson skeleton from numpygh-2225 Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com> * MAINT: Trim backend.py down to f2py2e flags * ENH: Add a factory function for backends * ENH: Add a distutils backend * ENH: Handle --backends in f2py Defaults to distutils for now * DOC: Add some minor comments in f2py2e * MAINT: Refactor and rework meson.build.src * MAINT: Add objects * MAINT: Cleanup distutils backend * MAINT: Refactor to add everything back to backend Necessary for the meson.build for now. Refactors / cleanup needs better argument handling in f2py2e * MAINT: Fix overly long line * BUG: Construct wrappers for meson backend * MAINT: Rework, simplify template massively * ENH: Truncate meson.build to skeleton only * MAINT: Minor backend housekeeping, name changes * MAINT: Less absolute paths, update setup.py [f2py] * MAINT: Move f2py module name functionality Previously part of np.distutils * ENH: Handle .pyf files * TST: Fix typo in isoFortranEnvMap.f90 * MAINT: Typo in f2py2e support for pyf files * DOC: Add release note for --backend * MAINT: Conditional switch for Python 3.12 [f2py] * MAINT: No absolute paths in backend [f2py-meson] The files are copied over anyway, this makes it easier to extend the generated skeleton * MAINT: Prettier generated meson.build files [f2py] * ENH: Add meson's dependency(blah) to f2py * DOC: Document the new flag * MAINT: Simplify and rename backend template [f2py] Co-authored-by: rgommers <rgommers@users.noreply.github.com> * ENH: Support build_type via --debug [f2py-meson] * MAINT,DOC: Reduce warn,rework doc [f2py-meson] Co-authored-by: rgommers <rgommers@users.noreply.github.com> * ENH: Rework deps: to --dep calls [f2py-meson] Also shows how incremental updates to the parser can be done. * MAINT,DOC: Add --backend to argparse, add docs * MAINT: Rename meson template [f2py-meson] * MAINT: Add meson.build for f2py Should address numpy#22225 (comment) * BLD: remove duplicate f2py handling in meson.build files --------- Co-authored-by: Namami Shanker <namami2011@gmail.com> Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com> Co-authored-by: rgommers <rgommers@users.noreply.github.com> Co-authored-by: Ralf Gommers <ralf.gommers@gmail.com>

* FIX: Import f2py2e rather than f2py for run_main * FIX: Import f2py2e instead of f2py * ENH: Add F2PY back-end work from gh-22225 * ENH: Add meson skeleton from gh-2225 * MAINT: Trim backend.py down to f2py2e flags * ENH: Add a factory function for backends * ENH: Add a distutils backend * ENH: Handle --backends in f2py Defaults to distutils for now * DOC: Add some minor comments in f2py2e * MAINT: Refactor and rework meson.build.src * MAINT: Add objects * MAINT: Cleanup distutils backend * MAINT: Refactor to add everything back to backend Necessary for the meson.build for now. Refactors / cleanup needs better argument handling in f2py2e * MAINT: Fix overly long line * BUG: Construct wrappers for meson backend * MAINT: Rework, simplify template massively * ENH: Truncate meson.build to skeleton only * MAINT: Minor backend housekeeping, name changes * MAINT: Less absolute paths, update setup.py [f2py] * MAINT: Move f2py module name functionality Previously part of np.distutils * ENH: Handle .pyf files * TST: Fix typo in isoFortranEnvMap.f90 * MAINT: Typo in f2py2e support for pyf files * DOC: Add release note for --backend * MAINT: Conditional switch for Python 3.12 [f2py] * MAINT: No absolute paths in backend [f2py-meson] The files are copied over anyway, this makes it easier to extend the generated skeleton * MAINT: Prettier generated meson.build files [f2py] * ENH: Add meson's dependency(blah) to f2py * DOC: Document the new flag * MAINT: Simplify and rename backend template [f2py] * ENH: Support build_type via --debug [f2py-meson] * MAINT,DOC: Reduce warn,rework doc [f2py-meson] * ENH: Rework deps: to --dep calls [f2py-meson] Also shows how incremental updates to the parser can be done. * MAINT,DOC: Add --backend to argparse, add docs * MAINT: Rename meson template [f2py-meson] * MAINT: Add meson.build for f2py Should address #22225 (comment) * BLD: remove duplicate f2py handling in meson.build files --------- Co-authored-by: Rohit Goswami <rgoswami@quansight.com> Co-authored-by: Namami Shanker <namami2011@gmail.com> Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com> Co-authored-by: rgommers <rgommers@users.noreply.github.com> Co-authored-by: Ralf Gommers <ralf.gommers@gmail.com>

charris · 2023-09-05T18:35:00Z

This is updated in #24532, so closing this.

* FIX: Import f2py2e rather than f2py for run_main * FIX: Import f2py2e instead of f2py * ENH: Add F2PY back-end work from numpygh-22225 Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com> * ENH: Add meson skeleton from numpygh-2225 Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com> * MAINT: Trim backend.py down to f2py2e flags * ENH: Add a factory function for backends * ENH: Add a distutils backend * ENH: Handle --backends in f2py Defaults to distutils for now * DOC: Add some minor comments in f2py2e * MAINT: Refactor and rework meson.build.src * MAINT: Add objects * MAINT: Cleanup distutils backend * MAINT: Refactor to add everything back to backend Necessary for the meson.build for now. Refactors / cleanup needs better argument handling in f2py2e * MAINT: Fix overly long line * BUG: Construct wrappers for meson backend * MAINT: Rework, simplify template massively * ENH: Truncate meson.build to skeleton only * MAINT: Minor backend housekeeping, name changes * MAINT: Less absolute paths, update setup.py [f2py] * MAINT: Move f2py module name functionality Previously part of np.distutils * ENH: Handle .pyf files * TST: Fix typo in isoFortranEnvMap.f90 * MAINT: Typo in f2py2e support for pyf files * DOC: Add release note for --backend * MAINT: Conditional switch for Python 3.12 [f2py] * MAINT: No absolute paths in backend [f2py-meson] The files are copied over anyway, this makes it easier to extend the generated skeleton * MAINT: Prettier generated meson.build files [f2py] * ENH: Add meson's dependency(blah) to f2py * DOC: Document the new flag * MAINT: Simplify and rename backend template [f2py] Co-authored-by: rgommers <rgommers@users.noreply.github.com> * ENH: Support build_type via --debug [f2py-meson] * MAINT,DOC: Reduce warn,rework doc [f2py-meson] Co-authored-by: rgommers <rgommers@users.noreply.github.com> * ENH: Rework deps: to --dep calls [f2py-meson] Also shows how incremental updates to the parser can be done. * MAINT,DOC: Add --backend to argparse, add docs * MAINT: Rename meson template [f2py-meson] * MAINT: Add meson.build for f2py Should address numpy#22225 (comment) * BLD: remove duplicate f2py handling in meson.build files --------- Co-authored-by: Namami Shanker <namami2011@gmail.com> Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com> Co-authored-by: rgommers <rgommers@users.noreply.github.com> Co-authored-by: Ralf Gommers <ralf.gommers@gmail.com>

HaoZeke and others added 30 commits September 7, 2022 15:57

ENH: Start refactoring f2py front-end

a5dfd68

ENH: Start using logging

1eacfb0

ENH: Add more options for the new front-end

173a7bb

ENH: Half the f2py options

52762a9

ENH: Switch to using f2pyarg

c0689cb

ENH: Add f2py frontend stuff

44bec87

FIX: f2pyarg --build-dir create directory if not present

81b14f4

MAINT: Move f2pyarg helpers to service

7c68149

FIX: Fix incorrect imports in f2pyarg service

9ab806a

MAINT: f2py check if input f2cmap file exists

c70272c

MAINT: Fix --wrap-functions docs

9fde953

MAINT: f2pyarg Remove unnecessary flags

4c4b99a

MAINT: f2pyarg remove unnecessary flags

bbb44a7

MAINT: f2pyarg fix default values

d26bd6a

FIX: f2pyarg transfer -h flag to generate_wrappers parser

9ce2140

ENH: f2pyarg add file segregation method

bcf7902

ENH: Create temp dir for null builddir value

568b94d

ENH: Chop callcrackfortran and move it to service. Add other helpers.

b3d8a2b

ENH: Initial f2pyarg main arg process function

4180fca

FIX: fix imports

a0eab32

FIX: Set default verbosity to true like f2py2e

528dc4f

MAINT: Remove unneccessary imports

d2a72b1

ENH: Handle overwrite-signature flag

475f4df

ENH: update auxfuncs

aa3d300

ENH: Segregate pyf files

885c772

MAINT: Refactoring

f39eb6d

MAINT: Rename _parse_postlist to _check_postlist

1df8ee2

Revert "ENH: update auxfuncs"

15be187

This reverts commit b334020.

FIX: Accept array of debug options rather that boolean flag

e13faa1

FIX: Default module name and default build directories

28e5beb

rgommers added the component: numpy.f2py label Sep 9, 2022

rgommers reviewed Sep 9, 2022

View reviewed changes

HaoZeke self-requested a review September 11, 2022 05:38

HaoZeke changed the title ~~ENH: Meson buiid system support for F2PY's f2py -c command~~ ENH: Meson build system support for F2PY's f2py -c command Oct 10, 2022

rgommers mentioned this pull request Aug 18, 2023

BUG: f2py doesn't work on CPython 3.12 #24444

Closed

HaoZeke added a commit to HaoZeke/numpy that referenced this pull request Aug 25, 2023

MAINT: Import f2py.utils from numpygh-22225

cc651f1

Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com>

HaoZeke added a commit to HaoZeke/numpy that referenced this pull request Aug 25, 2023

ENH: Add F2PY back-end work from numpygh-22225

7d9e94d

Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com>

HaoZeke added a commit to HaoZeke/numpy that referenced this pull request Aug 25, 2023

ENH: Add F2PY back-end work from numpygh-22225

92c4774

Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com>

HaoZeke mentioned this pull request Aug 25, 2023

ENH: meson backend for f2py #24532

Merged

3 tasks

HaoZeke reviewed Aug 25, 2023

View reviewed changes

charris added this to the 1.26.0 release milestone Aug 25, 2023

HaoZeke added a commit to HaoZeke/numpy that referenced this pull request Aug 29, 2023

MAINT: Add meson.build for f2py

81fd728

Should address numpy#22225 (comment)

HaoZeke added a commit to HaoZeke/numpy that referenced this pull request Sep 2, 2023

ENH: Add F2PY back-end work from numpygh-22225

91f3825

Co-authored-by: NamamiShanker <NamamiShanker@users.noreply.github.com>

HaoZeke added a commit to HaoZeke/numpy that referenced this pull request Sep 2, 2023

MAINT: Add meson.build for f2py

518074e

Should address numpy#22225 (comment)

charris mentioned this pull request Sep 5, 2023

ENH: meson backend for f2py #24647

Merged

charris closed this Sep 5, 2023

rgommers mentioned this pull request Sep 5, 2023

BUG: f2py compile error with full path of signature file on Windows #22211

Closed

jmrohwer mentioned this pull request Oct 25, 2023

BUG: f2py -c does not work on Windows with meson (Python 3.12) #25000

Closed

		@@ -0,0 +1,22 @@
		project('${modulename}', 'c',
		version : '0.1',

		self.substitutions['numpy_get_include'] = self.numpy_get_include.absolute()
		self.substitutions['f2py_get_include'] = self.f2py_get_include.absolute()

Uh oh!

Conversation

NamamiShanker commented Sep 7, 2022

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rgommers left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jmrohwer commented Sep 19, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rgommers commented Aug 18, 2023

Uh oh!

mattip commented Aug 18, 2023

Uh oh!

HaoZeke commented Aug 22, 2023

Uh oh!

HaoZeke commented Aug 25, 2023

Uh oh!

HaoZeke left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jmrohwer commented Aug 29, 2023

Uh oh!

charris commented Sep 5, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

jmrohwer commented Sep 19, 2022 •

edited

Loading