MNT: replace file with importlib.resources (compat with pyOxidizer and zipapp) #20297

jackzyliu · 2021-06-17T23:08:16Z

Reference Issues/PRs

Closes #20081
helping #20150

What does this implement/fix? Explain your changes.

Replaces occurrences of __file__ with importlib.resources in sklearn/datasets/*
Compatibility with pyOxidizer and zipapp:
- change IO functions to use importlib.resources.{open, read}_{binary, text}
- a few path-based IO tests use importlib.resources.path to avoid making the assumption that resources already live in a filesystem

Any other comments?

What is the best way to test compatibility? Something like this?

sklearn/datasets:

- [x] sklearn/datasets/_base.py:357:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:484:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:601:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:750:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:857:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:956:    base_dir = join(dirname(__file__), "data/")
- [x] sklearn/datasets/_base.py:970:    with open(dirname(__file__) + "/descr/linnerud.rst") as f:
- [x] sklearn/datasets/_base.py:1052:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:1122:    module_path = join(dirname(__file__), "images")
- [x] sklearn/datasets/_kddcup99.py:205:    module_path = dirname(__file__)
- [x] sklearn/datasets/_lfw.py:338:    module_path = dirname(__file__)
- [x] sklearn/datasets/_lfw.py:528:    module_path = dirname(__file__)
- [x] sklearn/datasets/_rcv1.py:281:    module_path = dirname(__file__)
- [x] sklearn/datasets/_california_housing.py:176:    module_path = dirname(__file__)
- [x] sklearn/datasets/_olivetti_faces.py:140:    module_path = dirname(__file__)
- [x] sklearn/datasets/tests/test_svmlight_format.py:19:currdir = os.path.dirname(os.path.abspath(__file__))
- [x] sklearn/datasets/tests/test_openml.py:36:currdir = os.path.dirname(os.path.abspath(__file__))

Other (should probably be separate PR(s) since __file__ is used slightly differently in these occurrences):

- sklearn/__check_build/__init__.py:19:    local_dir = os.path.split(__file__)[0]
- sklearn/utils/__init__.py:1164:    root = str(Path(__file__).parent.parent)  # sklearn package
- sklearn/utils/_testing.py:733:        cwd = op.normpath(op.join(op.dirname(sklearn.__file__), ".."))

jackzyliu · 2021-06-17T23:28:48Z

sklearn/datasets/_base.py


+DATA_MODULE = "sklearn.datasets.data"
+DESCR_MODULE = "sklearn.datasets.descr"
+IMAGES_MODULE = "sklearn.datasets.images"


Both string and module work. Using string here so that it can be more conveniently serialized (e.g. pickle) as part of a Bunch.

jackzyliu · 2021-06-17T23:31:12Z

sklearn/datasets/_base.py

        feature_names=feature_names,
-        filename=csv_filename,
+        filename=data_file_name,
+        file_module=DATA_MODULE


To be compatible with pyOxidizer and zipapp, we can no longer pass around an absolute file path. Instead, we separately pass around filename (just the name of the file) and file_module that the file lives in.

jackzyliu · 2021-06-17T23:32:40Z

sklearn/datasets/tests/test_base.py

-        assert all([os.path.exists(bunch.get(f, False)) for f in filenames])
+        assert all([
+            f in bunch and 'file_module' in bunch and
+            importlib_resources.is_resource(bunch['file_module'], bunch[f])


Modifying the test here to remove checks on the absolute paths.

jackzyliu · 2021-06-17T23:33:50Z

sklearn/datasets/_base.py

+    descr = load_descr(IMAGES_MODULE, 'README.txt')
+
+    images_module_traversable = importlib_resources.files(IMAGES_MODULE)
+    with importlib_resources.as_file(images_module_traversable) as images_dir:


I think this is only available python>=3.9, so it could be problematic. So we probably need to find another (hopefully clean) way to do this.

We can use the backport library importlib_resources for python<=3.8 as noted here.

@jackzyliu Do you know if this is the reason for the failures in the PR? If so shall we try to make sure everything else is working and leave the images for after?

@jackzyliu Do you know if this is the reason for the failures in the PR checks? If so shall we try to make sure everything else is working and tackle the images after?

Yea, it is.

I will make the images portion backward-compatible.

ogrisel

The point I a expressed in #20150 (comment) is to avoid use the open() Python builtin anywhere and instead using the functions from importlib.resources.open_binary/text or importlib.resources.read_binary/text documented here:

https://docs.python.org/3/library/importlib.html#module-importlib.resources

The goal is to avoid assuming those resource names are actual filenames of resource stored on a filesystem at any point.

sklearn/datasets/_base.py

jackzyliu · 2021-06-18T16:08:53Z

The goal is to avoid assuming those resource names are actual filenames of resource stored on a filesystem at any point.

Gotcha. I will make these changes accordingly and make the .files portion backward-compatible.

jackzyliu · 2021-06-23T01:17:58Z

sklearn/datasets/_base.py

        target_names[0] is the name of the target[0] class.
    """
-    with open(join(module_path, "data", data_file_name)) as csv_file:
+    with importlib_resources.open_text(data_module, data_file_name) as csv_file:


.open_text has encoding="utf-8" by default (reference)

Happy to make it explicit though.

I think this is fine to rely on the default utf-8 encoding assumption. I am pretty sure that all our CSV file only need the ascii encoding but this is a valid subset of utf-8 as far as I know.

We just needed to ensure that the loader would not implicitly try to use a platform dependent encoding.

jackzyliu · 2021-06-23T01:21:52Z

sklearn/datasets/_base.py

    return data, target, target_names


+def load_gzip_compressed_csv_data(data_module, data_file_name, **kwargs):


I don't think we can ask np.loadtxt to decompress if we are passing in file object (as opposed to a file name ending with '.gz'). So for the few .csv.gz files, we first manually decompress the file object and the use np.loadtxt to load data as before.

jackzyliu · 2021-06-23T01:22:50Z

sklearn/datasets/tests/test_base.py

+        ("breast_cancer.csv", 569, 30, ["malignant", "benign"]),
+    ],
+)
+def test_load_csv_data(


Adding standalone tests for load_csv_data and load_gzip_compressed_csv_data.

jackzyliu · 2021-06-23T01:26:53Z

sklearn/datasets/_base.py

+
+    filenames = list()
+    images = list()
+    for filename in sorted(importlib_resources.contents(IMAGES_MODULE)):


We can directly use .contents() (which returns a list of file names and is available since python 3.7) to avoid using .files(). It was my mistake not remembering this option.

jackzyliu · 2021-06-23T01:33:43Z

@ogrisel

Just updated according to your comments. Is it more along the line of what you were thinking?

I think the failing tests come from main. Will rebase after those are fixed.

jackzyliu · 2021-06-23T23:20:24Z

@ogrisel @quendee

At this point, all the conversions are done for sklearn/datasets. I think we should probably have separate PR(s) for the other 3 occurrences (see below: e.g. sklearn/__check_build/__init__.py:19) since __file__ is used slightly differently in these cases.

For ease of review, I have one separate, mostly self-contained commit for each of the following component:

sklearn/datasets/_base.py
test adjustment for test_common.py
sklearn/datasets/{_californa_housing, _covtype, _kddcup99, ...}.py: just the same changes applied to _base.py
sklearn/datasets/tests/test_openml.py: This involves changing data directory names from digits to id_* to make them valid module names (thus the large number of file changes), but otherwise straightforward.
sklearn/datasets/tests/test_svmlight_format.py: A few tests explicitly test path-based input for load_svmlight_file and load_svmlight_files. So for those tests, I used importlib.resources.path (to generate a temporary path within context manager if necessary), but otherwise used importlib.resources.open_binary as much as possible. Hope this is ok.

Let me know of comments.

sklearn/datasets:

- [x] sklearn/datasets/_base.py:357:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:484:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:601:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:750:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:857:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:956:    base_dir = join(dirname(__file__), "data/")
- [x] sklearn/datasets/_base.py:970:    with open(dirname(__file__) + "/descr/linnerud.rst") as f:
- [x] sklearn/datasets/_base.py:1052:    module_path = dirname(__file__)
- [x] sklearn/datasets/_base.py:1122:    module_path = join(dirname(__file__), "images")
- [x] sklearn/datasets/_kddcup99.py:205:    module_path = dirname(__file__)
- [x] sklearn/datasets/_lfw.py:338:    module_path = dirname(__file__)
- [x] sklearn/datasets/_lfw.py:528:    module_path = dirname(__file__)
- [x] sklearn/datasets/_rcv1.py:281:    module_path = dirname(__file__)
- [x] sklearn/datasets/_california_housing.py:176:    module_path = dirname(__file__)
- [x] sklearn/datasets/_olivetti_faces.py:140:    module_path = dirname(__file__)
- [x] sklearn/datasets/tests/test_svmlight_format.py:19:currdir = os.path.dirname(os.path.abspath(__file__))
- [x] sklearn/datasets/tests/test_openml.py:36:currdir = os.path.dirname(os.path.abspath(__file__))

Others (should probably be separate PR(s) since __file__ is used slightly differently in these occurrences):

- sklearn/__check_build/__init__.py:19:    local_dir = os.path.split(__file__)[0]
- sklearn/utils/__init__.py:1164:    root = str(Path(__file__).parent.parent)  # sklearn package
- sklearn/utils/_testing.py:733:        cwd = op.normpath(op.join(op.dirname(sklearn.__file__), ".."))

jackzyliu · 2021-06-29T22:04:10Z

Updated whats new, and added a few .DESCR checks for the few modified data fetch functions in sklearn/datasets (e.g. fetch_olivetti_faces).

ogrisel

Thanks @jackzyliu. This look good overall. Here are a few suggestion for further improvement. Once done I think that would be ready for merge on my end.

ogrisel · 2021-06-30T08:27:06Z

sklearn/datasets/_base.py

+    fdescr = importlib_resources.read_text(descr_module, descr_file_name)
+
+    return fdescr


As far as I can see most of the descriptions are loaded from DESCR_MODULE so to avoid redundancy we could have:

load_descr(descr_file_name, descr_module=DESCR_MODULE):

that would make it possible to only import load_descr without having to import the DESCR_MODULE constant most of the time.

For convenience you could add an optional descr_file_name=None / descr_module=DESCR_MODULE to load_csv_data and load_gzip_compressed_csv_data and when descr_file_name is not None, make those functions return the loaded description as as additional return value.

As far as I can see most of the descriptions are loaded from DESCR_MODULE so to avoid redundancy we could have:

load_descr(descr_file_name, descr_module=DESCR_MODULE):

that would make it possible to only import load_descr without having to import the DESCR_MODULE constant most of the time.

By extension, I am applying the same changes to load_csv_data and load_gzip_compressed_csv_data to avoid the same redundancy since these data are always loaded from DATA_MODULE.

load_csv_data(data_file_name, data_module=DATA_MODULE)

And since as positional arguments to importlib.resources functions, the resource module always comes before the resource name, I am enforcing named arguments to avoid usage error.

load_csv_data(data_file_name, *, data_module=DATA_MODULE) load_descr(descr_file_name, *, descr_module=DESCR_MODULE)

Lastly, combined with the suggestion to add optional descr_file_name=None, descr_module=DESCR_MODULE, we have the following function signature:

load_csv_data(data_file_name, *, data_module=DATA_MODULE, descr_file_name=None, descr_module=DESCR_MODULE)

ogrisel · 2021-06-30T08:27:33Z

sklearn/datasets/_base.py

+            gzip_encoding = kwargs["encoding"]
+
+        compressed_file = gzip.open(compressed_file, mode="rt", encoding=gzip_encoding)
+        data = np.loadtxt(compressed_file, **kwargs)


I don't understand why we would pass the encoding kwarg to np.loadtxt if it has arleady been used to decode the text of gzip.open.

Either call gzip.open with mode="rb" and only pass encoding to np.loadtxt so that it handles the text decoding correctly itself or pass it only to gzip.open with model="rt" (in which case it would make sense to have load_gzip_compressed_csv_data(data_module, data_file_name, encoding="utf-8", **kwargs) to simplify the code to only forward kwargs to np.loadtxt.

Good point. Although encoding is passed around a bunch of times in np.loadtxt, it's not really used when the file is already in text.

Going with the latter option.

ogrisel · 2021-06-30T08:32:46Z

sklearn/datasets/_base.py

+    data_file_name = "iris.csv"
+    data, target, target_names = load_csv_data(DATA_MODULE, data_file_name)


I think we can avoid the extra indirection in a local variable that is used only once.

Suggested change

data_file_name = "iris.csv"

data, target, target_names = load_csv_data(DATA_MODULE, data_file_name)

data, target, target_names = load_csv_data(DATA_MODULE, "iris.csv")

data_file_name is actually used again in constructing the Bunch below, which is why I used a variable. Do you still want me to make the change?

Edit: I think the same goes to breast_cancer.csv below.

ogrisel · 2021-06-30T08:33:08Z

sklearn/datasets/_base.py

+    data_file_name = "breast_cancer.csv"
+    data, target, target_names = load_csv_data(DATA_MODULE, data_file_name)


Suggested change

data_file_name = "breast_cancer.csv"

data, target, target_names = load_csv_data(DATA_MODULE, data_file_name)

data, target, target_names = load_csv_data(DATA_MODULE, "breast_cancer.csv")

ogrisel · 2021-06-30T08:44:08Z

sklearn/datasets/tests/test_base.py

        assert len(bunch.target_names) == n_target
    if has_descr:
        assert bunch.DESCR
    if filenames:


If we expect a non-empty list of filenames, I think we should require the data module to be present in the bunch.

ogrisel · 2021-06-30T08:44:53Z

sklearn/datasets/tests/test_base.py

+        assert all(
+            [
+                f in bunch
+                and "file_module" in bunch


I would rather use the data_module field in all those Bunch instances:

Suggested change

and "file_module" in bunch

and "data_module" in bunch

ogrisel · 2021-06-30T08:48:37Z

sklearn/datasets/tests/test_openml.py

-    ) as modified_gzip:
+    original_data_module = OPENML_TEST_DATA_MODULE + "." + f"id_{data_id}"
+    original_data_file_name = "data-v1-dl-1666876.arff.gz"
+    corrupt_copy_path = os.path.join(tmpdir, "test_invalid_checksum.arff")


nitpick: I am pretty sure this can be simplified as:

Suggested change

corrupt_copy_path = os.path.join(tmpdir, "test_invalid_checksum.arff")

corrupt_copy_path = tmpdir / "test_invalid_checksum.arff"

ogrisel · 2021-06-30T08:51:37Z

sklearn/datasets/tests/test_svmlight_format.py

 import scipy.sparse as sp
 import os
 import shutil
+from importlib import resources as importlib_resources


I think we could just use:

Suggested change

from importlib import resources as importlib_resources

from importlib import resources

and then call resources.open_binary / open_binary.open_text instead of renaming this module every where.

jackzyliu · 2021-06-30T21:45:26Z

Hi @ogrisel, just updated according to your comments:

change signatures of underlying load functions to reduce various redundancies
allow load_descr within load_csv_data and load_gzip_compressed_csv_data
directly use from importlib import resources instead of renaming the module
rename module field in returning Bunch from file_module to data_module
style improvement (replace os.path.join with a /)

Build failure: codecov/patch shows failed, but the marked statements are all either 1) actually tested in 9cd899a (descr checks) or 2) themselves test statements.

Please let me know of further comments. Thanks!

glemaitre · 2021-07-22T15:31:38Z

Regarding the failure, I don't think this is big deal. We need to be sure to run the test locally (they are passing on my side) but they will be skipped anyway on the CIs.

glemaitre

LGTM. This only suggestions for the documentation to follow as much as possible numpydoc style

glemaitre · 2021-07-22T15:33:26Z

sklearn/datasets/_base.py

    data_file_name : string
        Name of csv file to be loaded from
-        module_path/data/data_file_name. For example 'wine_data.csv'.
+        data_module/data_file_name. For example 'wine_data.csv'.


Suggested change

data_module/data_file_name. For example 'wine_data.csv'.

`data_module/data_file_name`. For example `'wine_data.csv'`.

glemaitre · 2021-07-22T15:34:29Z

sklearn/datasets/_base.py

+    data_module : string or module, optional
+        module where data lives;
+        default "sklearn.datasets.data" (DATA_MODULE).


Suggested change

data_module : string or module, optional

module where data lives;

default "sklearn.datasets.data" (DATA_MODULE).

data_module : str or module, default="sklearn.datasets.data"

Module where data lives. The default is `"sklearn.datasets.data"`.

glemaitre · 2021-07-22T15:35:18Z

sklearn/datasets/_base.py

+        module where data lives;
+        default "sklearn.datasets.data" (DATA_MODULE).
+
+    descr_file_name : string, optional


Suggested change

descr_file_name : string, optional

descr_file_name : str, default=None

glemaitre · 2021-07-22T15:37:48Z

sklearn/datasets/_base.py

+        (See `load_descr`) Name of rst file to be loaded from
+        descr_module/descr_file_name. For example 'wine_data.rst';
+        If not None, also returns the corresponding description of
+        the dataset.


Suggested change

(See `load_descr`) Name of rst file to be loaded from

descr_module/descr_file_name. For example 'wine_data.rst';

If not None, also returns the corresponding description of

the dataset.

Name of rst file to be loaded from `descr_module/descr_file_name`.

For example 'wine_data.rst'. See also :func:`load_descr`.

If not None, also returns the corresponding description of

the dataset.

glemaitre · 2021-07-22T15:38:13Z

sklearn/datasets/_base.py

+        If not None, also returns the corresponding description of
+        the dataset.
+
+    descr_module : string or module, optional


Suggested change

descr_module : string or module, optional

descr_module : str or module, default="sklearn.datasets.descr"

glemaitre · 2021-07-22T15:44:35Z

sklearn/datasets/_base.py

+    descr : string, optional
+        Description of the dataset (content of descr_file_name). Only returned
+        if descr_file_name is not None.


Can you format it as in the previous function.

glemaitre · 2021-07-22T15:45:11Z

sklearn/datasets/_base.py

+    descr_file_name : string
+        Name of rst file to be loaded from
+        descr_module/descr_file_name. For example 'wine_data.rst'.
+
+    descr_module : string or module, optional
+        module where descr_file_name lives;
+        default "sklearn.datasets.descr" (DESCR_MODULE).


If you use the docstring of the previous function as well.

glemaitre · 2021-07-22T15:45:33Z

sklearn/datasets/_base.py

+
+
+def load_descr(descr_file_name, *, descr_module=DESCR_MODULE):
+    """Loads descr_file_name from descr_module with importlib.resources.


Suggested change

"""Loads descr_file_name from descr_module with importlib.resources.

"""Load `descr_file_name` from `descr_module` with `importlib.resources`.

glemaitre · 2021-07-22T15:46:21Z

sklearn/datasets/_base.py

+
+    Returns
+    -------
+    string; content of descr_file_name


Suggested change

string; content of descr_file_name

fdescr : str

Content of `descr_file_name`.

glemaitre · 2021-07-22T15:47:56Z

sklearn/datasets/_base.py

+    filenames = list()
+    images = list()


Suggested change

filenames = list()

images = list()

filenames, images = [], []

jackzyliu · 2021-07-22T23:34:49Z

@glemaitre Thanks a lot for the review and suggestions! I have modified the documentations accordingly. Let me know if anything else needs changing.

Weirdly, the doc checks both seem have failed at build_doc.sh and specifically pip install "$(get_dep sphinxext-opengraph $SPHINXEXT_OPENGRAPH_VERSION)" where $SPHINXEXT_OPENGRAPH_VERSION didn't read in the values specified in config.yml. I wonder what could be the issue here.

I will wait for further comments before I push to force-run checks again.

glemaitre · 2021-07-23T07:51:33Z

I think that you only need to merge main into your branch and it should be fine.

…r and zipapp) sklearn/datasets/_base.py

…GH20081)

… (GH20081)

…20081)

- change signatures of underlying load functions to reduce various redundancies - allow `load_descr` within `load_csv_data` and `load_gzip_compressed_csv_data` - undo renaming of `importlib.resources` in import - rename module field in returning `Bunch` - other style improvements

…(GH20081)

jackzyliu · 2021-07-23T19:27:11Z

I think that you only need to merge main into your branch and it should be fine.

Thanks! All green now.

rth

Thanks @jackzyliu ! LGTM as well. Merging with two approvals above.

nithish08

LGTM

(compat with pyOxidizer and zipapp)

github-actions bot added the module:datasets label Jun 17, 2021

jackzyliu mentioned this pull request Jun 17, 2021

MNT remove __file__ and use importlib.resource instead #20150

Closed

jackzyliu commented Jun 17, 2021

View reviewed changes

jackzyliu force-pushed the replace_file branch from 503f769 to bef5821 Compare June 17, 2021 23:47

ogrisel reviewed Jun 18, 2021

View reviewed changes

sklearn/datasets/_base.py Outdated Show resolved Hide resolved

sklearn/datasets/_base.py Outdated Show resolved Hide resolved

jackzyliu force-pushed the replace_file branch from 8bf9b0c to 5ba134f Compare June 23, 2021 00:45

jackzyliu commented Jun 23, 2021

View reviewed changes

ogrisel reviewed Jun 30, 2021

View reviewed changes

cmarmo mentioned this pull request Jul 9, 2021

__file__ should be avoided in library code #20081

Closed

glemaitre self-requested a review July 22, 2021 14:32

glemaitre approved these changes Jul 22, 2021

View reviewed changes

glemaitre mentioned this pull request Jul 22, 2021

[WIP] replace __file__ to importlib resources #20284

Closed

jackzyliu added 4 commits July 23, 2021 11:50

MNT: replace __file__ with importlib.resources (compat with pyOxidize…

6ba9382

…r and zipapp) sklearn/datasets/_base.py

Add resource modules as exceptions to test_all_tests_are_importable (…

500a0fa

…GH20081)

replace __file__ with importlib.resources for sklearn/datasets/*.py (…

ed730ef

…GH20081)

replace __file__with importlib.resources for test_openml.py (GH20081)

1a6a2c6

jackzyliu added 5 commits July 23, 2021 11:50

replace __file__ with importlib.resources for test_svmlight_format.py…

6ceabf5

… (GH20081)

DOC add whats new (GH20081)

442ef7e

TST add descr checks for data fetch functions in sklearn/datasets (GH…

5e9ffe2

…20081)

Modify function documentation to better follow Numpy docstring style …

7977915

…(GH20081)

jackzyliu force-pushed the replace_file branch from 563bdfe to 7977915 Compare July 23, 2021 18:50

rth approved these changes Jul 27, 2021

View reviewed changes

rth merged commit 5562cc5 into scikit-learn:main Jul 27, 2021

nithish08 reviewed Jul 28, 2021

View reviewed changes

TomDLT pushed a commit to TomDLT/scikit-learn that referenced this pull request Jul 29, 2021

MNT: replace __file__ with importlib.resources (scikit-learn#20297)

49c0dae

(compat with pyOxidizer and zipapp)

samronsin pushed a commit to samronsin/scikit-learn that referenced this pull request Nov 30, 2021

MNT: replace __file__ with importlib.resources (scikit-learn#20297)

7633c1f

(compat with pyOxidizer and zipapp)

		return data, target, target_names


		def load_gzip_compressed_csv_data(data_module, data_file_name, **kwargs):

		fdescr = importlib_resources.read_text(descr_module, descr_file_name)

		return fdescr

		data_file_name = "iris.csv"
		data, target, target_names = load_csv_data(DATA_MODULE, data_file_name)

		data_file_name = "breast_cancer.csv"
		data, target, target_names = load_csv_data(DATA_MODULE, data_file_name)

	corrupt_copy_path = os.path.join(tmpdir, "test_invalid_checksum.arff")
	corrupt_copy_path = tmpdir / "test_invalid_checksum.arff"

	from importlib import resources as importlib_resources
	from importlib import resources

	data_module/data_file_name. For example 'wine_data.csv'.
	`data_module/data_file_name`. For example `'wine_data.csv'`.

	descr_file_name : string, optional
	descr_file_name : str, default=None

	descr_module : string or module, optional
	descr_module : str or module, default="sklearn.datasets.descr"



		def load_descr(descr_file_name, *, descr_module=DESCR_MODULE):
		"""Loads descr_file_name from descr_module with importlib.resources.

	"""Loads descr_file_name from descr_module with importlib.resources.
	"""Load `descr_file_name` from `descr_module` with `importlib.resources`.

	string; content of descr_file_name
	fdescr : str
	Content of `descr_file_name`.

Uh oh!

MNT: replace __file__ with importlib.resources (compat with pyOxidizer and zipapp) #20297

MNT: replace __file__ with importlib.resources (compat with pyOxidizer and zipapp) #20297

Uh oh!

Conversation

jackzyliu commented Jun 17, 2021 • edited by rth Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Reference Issues/PRs

What does this implement/fix? Explain your changes.

Any other comments?

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!

quendee Jun 18, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ogrisel left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

jackzyliu commented Jun 18, 2021

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!

jackzyliu Jun 23, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jackzyliu Jun 23, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jackzyliu commented Jun 23, 2021

Uh oh!

jackzyliu commented Jun 23, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jackzyliu commented Jun 29, 2021

Uh oh!

ogrisel 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!

jackzyliu Jun 30, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

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

MNT: replace file with importlib.resources (compat with pyOxidizer and zipapp) #20297

MNT: replace file with importlib.resources (compat with pyOxidizer and zipapp) #20297

jackzyliu commented Jun 17, 2021 •

edited by rth

Loading

quendee Jun 18, 2021 •

edited

Loading

ogrisel left a comment •

edited

Loading

jackzyliu Jun 23, 2021 •

edited

Loading

jackzyliu Jun 23, 2021 •

edited

Loading

jackzyliu commented Jun 23, 2021 •

edited

Loading

jackzyliu Jun 30, 2021 •

edited

Loading

jackzyliu commented Jun 30, 2021 •

edited

Loading