ENH: pathlib support for fromfile(), .tofile() and .dump() #12915
Conversation
|
Needs an entry in the 1.17.0 release notes. |
|
I must admit, I have no clue why this commit (3c060d5) would cause one particular build (Windows Python37-32bit-fast) to fail, while all the others pass.. As far as I can tell, the failing tests should be unaffected by the changes. I don't have access to a Windows machine, and I cannot reproduce on Linux Python37 32bit. |
|
Those matmul test failures are likely unrelated since we see them from time to time. Although they are only sporadic, it would be good to have them fixed. |
|
reopen to trigger re-test |
|
Needs a release note. |
Hmm, they seem to occur for complex64, same as for the wheels test failures. The wheels tests are checking |
Is this something that I need to do? |
|
You need to add an appropriate text description of the change to one of the sections in |
|
@sorenrasmussenai Ping. Take a look at |
| import pathlib | ||
| except ImportError: | ||
| try: | ||
| import pathlib2 as pathlib |
There was a problem hiding this comment.
I'm not actually sure our stuff works on pathlib2, or whether our CI exercises this code path.
Either way, if it turns out this test fails on systems using pathlib2, at least with the test present we'll likely get a bug report.
numpy/core/tests/test_multiarray.py
Outdated
| def test_roundtrip_dump_pathlib(self): | ||
| p = pathlib.Path(self.filename) | ||
| self.x.dump(p) | ||
| if sys.version_info[:2] >= (3, 6): |
There was a problem hiding this comment.
This shouldn't be necessary, numpy emulates the 3.6 behavior internally?
|
Needed a fix now that |
|
The windows failure looks like the occasional heisenbug in |
|
What version of python is the failing pypy? Failure looks like we're failing to detect a pathlike object in np.compat. |
|
3.6.3, latest nightly build, which recently fixed some pathlib issues but maybe not all of them |
| self.x.dump(p) | ||
| y = np.load(p, allow_pickle=True) | ||
| assert_array_equal(y, self.x) | ||
|
|
There was a problem hiding this comment.
self.x.dump(p) is failing to write to the file on PyPy, while self.x.tofile(p) is working fine. Is there a call to flush missing somewhere, or a missing close to trigger the write?
|
#13684 has been merged, now this has conflicts. Rebasing this off the new master HEAD (and resolving the conflicst) will give a linear history, merging master into the branch may work too. |
308940d to
3c618a3
Compare
|
@mattip: Merged master into this branch, but probably worth squash-merging the PR as a whole |
|
Thanks @sorenrasmussenai and the reviewers |
See #8576
This is my first time messing with the Python C API, so I hope I didn't mess up too bad. I appreciate any feedback!