Build and test PyWavelets Pyodide wheels in CI

[agriyakhetarpal]

This PR adds and configures a CI job to test wasm32 wheels for Pywavelets compiled via the Emscripten toolchain. Almost all tests pass, and some of them are skipped because of the current lack of threading for Pyodide (please refer to pyodide/pyodide#237).

It is likely that Pyodide has issues with file system access which was evident through OSErrors when trying to access .npy files, it's likely that this is a bug and I shall file it accordingly in the Pyodide issue tracker. The current workaround that has been adopted in this PR is to save the arrays loaded from these .npy files to compressed .npz file formats, which do not cause issues related to file permissions.

An IS_WASM check has been added to check for the platform and allow pytest to skip the concurrency-based tests after collection, as noted above.

[agriyakhetarpal]

Pyodide wheel builds are succeeding and tests are passing on my fork, see here: https://github.com/agriyakhetarpal/pywt/actions/runs/8003758318

[agriyakhetarpal]

Okay, so importlib.resources.files() is available on Python 3.9 and later we already support Python 3.9 and later, we might have to revert this change or look at other ways to access the data folder without the use of __file__, or reintroduce it and check if this was one of the issues that caused test failures earlier. Should be addressed by 7315b43.

[rgommers]

Thanks @agriyakhetarpal, this looks good! Nice to have a Pyodide CI job that works.

Beyond the one inline comment, I thought I'd ask if you want to rewrite the commit history (e.g., one commit for the CI job, one for the npy->npz, one for the test changes, and one for the other unrelated bits) or if you want me to squash-merge it. Either way is fine with me.

[agriyakhetarpal]

Thanks for the prompt review, @rgommers! It did take a bit of back and forth with the experiments – I don't have a strong opinion on rewriting history as such; it would be great if we can squash merge the PR, it's much easier

[rgommers]

Very nice!

 ============================= test session starts ==============================
platform emscripten -- Python 3.11.3, pytest-8.0.1, pluggy-1.4.0
...
================= 1030 passed, 6 skipped in 115.83s (0:01:55) ==================

Okay, in it goes.

[agriyakhetarpal]

This PR adds and configures a CI job to test wasm32 wheels for Pywavelets compiled via the Emscripten toolchain. Almost all tests pass, and some of them are skipped because of the current lack of threading for Pyodide (please refer to pyodide/pyodide#237).

It is likely that Pyodide has issues with file system access which was evident through OSErrors when trying to access .npy files, it's likely that this is a bug and I shall file it accordingly in the Pyodide issue tracker. The current workaround that has been adopted in this PR is to save the arrays loaded from these .npy files to compressed .npz file formats, which do not cause issues related to file permissions.

An IS_WASM check has been added to check for the platform and allow pytest to skip the concurrency-based tests after collection, as noted above.

[agriyakhetarpal]

Pyodide wheel builds are succeeding and tests are passing on my fork, see here: https://github.com/agriyakhetarpal/pywt/actions/runs/8003758318

[agriyakhetarpal]

Okay, so importlib.resources.files() is available on Python 3.9 and later we already support Python 3.9 and later, we might have to revert this change or look at other ways to access the data folder without the use of __file__, or reintroduce it and check if this was one of the issues that caused test failures earlier. Should be addressed by 7315b43.

[rgommers]

Thanks @agriyakhetarpal, this looks good! Nice to have a Pyodide CI job that works.

Beyond the one inline comment, I thought I'd ask if you want to rewrite the commit history (e.g., one commit for the CI job, one for the npy->npz, one for the test changes, and one for the other unrelated bits) or if you want me to squash-merge it. Either way is fine with me.

[agriyakhetarpal]

Thanks for the prompt review, @rgommers! It did take a bit of back and forth with the experiments – I don't have a strong opinion on rewriting history as such; it would be great if we can squash merge the PR, it's much easier

[rgommers]

Very nice!

 ============================= test session starts ==============================
platform emscripten -- Python 3.11.3, pytest-8.0.1, pluggy-1.4.0
...
================= 1030 passed, 6 skipped in 115.83s (0:01:55) ==================

Okay, in it goes.

[agriyakhetarpal]

This PR adds and configures a CI job to test wasm32 wheels for Pywavelets compiled via the Emscripten toolchain. Almost all tests pass, and some of them are skipped because of the current lack of threading for Pyodide (please refer to pyodide/pyodide#237).

It is likely that Pyodide has issues with file system access which was evident through OSErrors when trying to access .npy files, it's likely that this is a bug and I shall file it accordingly in the Pyodide issue tracker. The current workaround that has been adopted in this PR is to save the arrays loaded from these .npy files to compressed .npz file formats, which do not cause issues related to file permissions.

An IS_WASM check has been added to check for the platform and allow pytest to skip the concurrency-based tests after collection, as noted above.

[agriyakhetarpal]

Pyodide wheel builds are succeeding and tests are passing on my fork, see here: https://github.com/agriyakhetarpal/pywt/actions/runs/8003758318

[agriyakhetarpal]

Okay, so importlib.resources.files() is available on Python 3.9 and later we already support Python 3.9 and later, we might have to revert this change or look at other ways to access the data folder without the use of __file__, or reintroduce it and check if this was one of the issues that caused test failures earlier. Should be addressed by 7315b43.

[rgommers]

Thanks @agriyakhetarpal, this looks good! Nice to have a Pyodide CI job that works.

Beyond the one inline comment, I thought I'd ask if you want to rewrite the commit history (e.g., one commit for the CI job, one for the npy->npz, one for the test changes, and one for the other unrelated bits) or if you want me to squash-merge it. Either way is fine with me.

[agriyakhetarpal]

Thanks for the prompt review, @rgommers! It did take a bit of back and forth with the experiments – I don't have a strong opinion on rewriting history as such; it would be great if we can squash merge the PR, it's much easier

[rgommers]

Very nice!

 ============================= test session starts ==============================
platform emscripten -- Python 3.11.3, pytest-8.0.1, pluggy-1.4.0
...
================= 1030 passed, 6 skipped in 115.83s (0:01:55) ==================

Okay, in it goes.

[agriyakhetarpal]

This PR adds and configures a CI job to test wasm32 wheels for Pywavelets compiled via the Emscripten toolchain. Almost all tests pass, and some of them are skipped because of the current lack of threading for Pyodide (please refer to pyodide/pyodide#237).

It is likely that Pyodide has issues with file system access which was evident through OSErrors when trying to access .npy files, it's likely that this is a bug and I shall file it accordingly in the Pyodide issue tracker. The current workaround that has been adopted in this PR is to save the arrays loaded from these .npy files to compressed .npz file formats, which do not cause issues related to file permissions.

An IS_WASM check has been added to check for the platform and allow pytest to skip the concurrency-based tests after collection, as noted above.

[agriyakhetarpal]

Pyodide wheel builds are succeeding and tests are passing on my fork, see here: https://github.com/agriyakhetarpal/pywt/actions/runs/8003758318

[agriyakhetarpal]

Okay, so importlib.resources.files() is available on Python 3.9 and later we already support Python 3.9 and later, we might have to revert this change or look at other ways to access the data folder without the use of __file__, or reintroduce it and check if this was one of the issues that caused test failures earlier. Should be addressed by 7315b43.

[rgommers]

Thanks @agriyakhetarpal, this looks good! Nice to have a Pyodide CI job that works.

Beyond the one inline comment, I thought I'd ask if you want to rewrite the commit history (e.g., one commit for the CI job, one for the npy->npz, one for the test changes, and one for the other unrelated bits) or if you want me to squash-merge it. Either way is fine with me.

[agriyakhetarpal]

Thanks for the prompt review, @rgommers! It did take a bit of back and forth with the experiments – I don't have a strong opinion on rewriting history as such; it would be great if we can squash merge the PR, it's much easier

[rgommers]

Very nice!

 ============================= test session starts ==============================
platform emscripten -- Python 3.11.3, pytest-8.0.1, pluggy-1.4.0
...
================= 1030 passed, 6 skipped in 115.83s (0:01:55) ==================

Okay, in it goes.

[agriyakhetarpal]

This PR adds and configures a CI job to test wasm32 wheels for Pywavelets compiled via the Emscripten toolchain. Almost all tests pass, and some of them are skipped because of the current lack of threading for Pyodide (please refer to pyodide/pyodide#237).

It is likely that Pyodide has issues with file system access which was evident through OSErrors when trying to access .npy files, it's likely that this is a bug and I shall file it accordingly in the Pyodide issue tracker. The current workaround that has been adopted in this PR is to save the arrays loaded from these .npy files to compressed .npz file formats, which do not cause issues related to file permissions.

An IS_WASM check has been added to check for the platform and allow pytest to skip the concurrency-based tests after collection, as noted above.

[agriyakhetarpal]

Pyodide wheel builds are succeeding and tests are passing on my fork, see here: https://github.com/agriyakhetarpal/pywt/actions/runs/8003758318

[agriyakhetarpal]

Okay, so importlib.resources.files() is available on Python 3.9 and later we already support Python 3.9 and later, we might have to revert this change or look at other ways to access the data folder without the use of __file__, or reintroduce it and check if this was one of the issues that caused test failures earlier. Should be addressed by 7315b43.

[rgommers]

Thanks @agriyakhetarpal, this looks good! Nice to have a Pyodide CI job that works.

Beyond the one inline comment, I thought I'd ask if you want to rewrite the commit history (e.g., one commit for the CI job, one for the npy->npz, one for the test changes, and one for the other unrelated bits) or if you want me to squash-merge it. Either way is fine with me.

[agriyakhetarpal]

Thanks for the prompt review, @rgommers! It did take a bit of back and forth with the experiments – I don't have a strong opinion on rewriting history as such; it would be great if we can squash merge the PR, it's much easier

[rgommers]

Very nice!

 ============================= test session starts ==============================
platform emscripten -- Python 3.11.3, pytest-8.0.1, pluggy-1.4.0
...
================= 1030 passed, 6 skipped in 115.83s (0:01:55) ==================

Okay, in it goes.

[agriyakhetarpal]

This PR adds and configures a CI job to test wasm32 wheels for Pywavelets compiled via the Emscripten toolchain. Almost all tests pass, and some of them are skipped because of the current lack of threading for Pyodide (please refer to pyodide/pyodide#237).

It is likely that Pyodide has issues with file system access which was evident through OSErrors when trying to access .npy files, it's likely that this is a bug and I shall file it accordingly in the Pyodide issue tracker. The current workaround that has been adopted in this PR is to save the arrays loaded from these .npy files to compressed .npz file formats, which do not cause issues related to file permissions.

An IS_WASM check has been added to check for the platform and allow pytest to skip the concurrency-based tests after collection, as noted above.

[agriyakhetarpal]

Pyodide wheel builds are succeeding and tests are passing on my fork, see here: https://github.com/agriyakhetarpal/pywt/actions/runs/8003758318

[agriyakhetarpal]

Okay, so importlib.resources.files() is available on Python 3.9 and later we already support Python 3.9 and later, we might have to revert this change or look at other ways to access the data folder without the use of __file__, or reintroduce it and check if this was one of the issues that caused test failures earlier. Should be addressed by 7315b43.

[rgommers]

Thanks @agriyakhetarpal, this looks good! Nice to have a Pyodide CI job that works.

Beyond the one inline comment, I thought I'd ask if you want to rewrite the commit history (e.g., one commit for the CI job, one for the npy->npz, one for the test changes, and one for the other unrelated bits) or if you want me to squash-merge it. Either way is fine with me.

[agriyakhetarpal]

Thanks for the prompt review, @rgommers! It did take a bit of back and forth with the experiments – I don't have a strong opinion on rewriting history as such; it would be great if we can squash merge the PR, it's much easier

[rgommers]

Very nice!

 ============================= test session starts ==============================
platform emscripten -- Python 3.11.3, pytest-8.0.1, pluggy-1.4.0
...
================= 1030 passed, 6 skipped in 115.83s (0:01:55) ==================

Okay, in it goes.

[agriyakhetarpal]

This PR adds and configures a CI job to test wasm32 wheels for Pywavelets compiled via the Emscripten toolchain. Almost all tests pass, and some of them are skipped because of the current lack of threading for Pyodide (please refer to pyodide/pyodide#237).

It is likely that Pyodide has issues with file system access which was evident through OSErrors when trying to access .npy files, it's likely that this is a bug and I shall file it accordingly in the Pyodide issue tracker. The current workaround that has been adopted in this PR is to save the arrays loaded from these .npy files to compressed .npz file formats, which do not cause issues related to file permissions.

An IS_WASM check has been added to check for the platform and allow pytest to skip the concurrency-based tests after collection, as noted above.

[agriyakhetarpal]

Pyodide wheel builds are succeeding and tests are passing on my fork, see here: https://github.com/agriyakhetarpal/pywt/actions/runs/8003758318

[agriyakhetarpal]

Okay, so importlib.resources.files() is available on Python 3.9 and later we already support Python 3.9 and later, we might have to revert this change or look at other ways to access the data folder without the use of __file__, or reintroduce it and check if this was one of the issues that caused test failures earlier. Should be addressed by 7315b43.

[rgommers]

Thanks @agriyakhetarpal, this looks good! Nice to have a Pyodide CI job that works.

Beyond the one inline comment, I thought I'd ask if you want to rewrite the commit history (e.g., one commit for the CI job, one for the npy->npz, one for the test changes, and one for the other unrelated bits) or if you want me to squash-merge it. Either way is fine with me.

[agriyakhetarpal]

Thanks for the prompt review, @rgommers! It did take a bit of back and forth with the experiments – I don't have a strong opinion on rewriting history as such; it would be great if we can squash merge the PR, it's much easier

[rgommers]

Very nice!

 ============================= test session starts ==============================
platform emscripten -- Python 3.11.3, pytest-8.0.1, pluggy-1.4.0
...
================= 1030 passed, 6 skipped in 115.83s (0:01:55) ==================

Okay, in it goes.

[agriyakhetarpal]

This PR adds and configures a CI job to test wasm32 wheels for Pywavelets compiled via the Emscripten toolchain. Almost all tests pass, and some of them are skipped because of the current lack of threading for Pyodide (please refer to pyodide/pyodide#237).

It is likely that Pyodide has issues with file system access which was evident through OSErrors when trying to access .npy files, it's likely that this is a bug and I shall file it accordingly in the Pyodide issue tracker. The current workaround that has been adopted in this PR is to save the arrays loaded from these .npy files to compressed .npz file formats, which do not cause issues related to file permissions.

An IS_WASM check has been added to check for the platform and allow pytest to skip the concurrency-based tests after collection, as noted above.

[agriyakhetarpal]

Pyodide wheel builds are succeeding and tests are passing on my fork, see here: https://github.com/agriyakhetarpal/pywt/actions/runs/8003758318

[agriyakhetarpal]

Okay, so importlib.resources.files() is available on Python 3.9 and later we already support Python 3.9 and later, we might have to revert this change or look at other ways to access the data folder without the use of __file__, or reintroduce it and check if this was one of the issues that caused test failures earlier. Should be addressed by 7315b43.

[rgommers]

Thanks @agriyakhetarpal, this looks good! Nice to have a Pyodide CI job that works.

Beyond the one inline comment, I thought I'd ask if you want to rewrite the commit history (e.g., one commit for the CI job, one for the npy->npz, one for the test changes, and one for the other unrelated bits) or if you want me to squash-merge it. Either way is fine with me.

[agriyakhetarpal]

Thanks for the prompt review, @rgommers! It did take a bit of back and forth with the experiments – I don't have a strong opinion on rewriting history as such; it would be great if we can squash merge the PR, it's much easier

[rgommers]

Very nice!

 ============================= test session starts ==============================
platform emscripten -- Python 3.11.3, pytest-8.0.1, pluggy-1.4.0
...
================= 1030 passed, 6 skipped in 115.83s (0:01:55) ==================

Okay, in it goes.