GH-43868: [CI][Python] Skip test that requires PARQUET_TEST_DATA env on emscripten

[raulcd]

Rationale for this change

The following PR:

• GH-41579: [C++][Python][Parquet] Support reading/writing key-value metadata from/to ColumnChunkMetaData #41580

Made mandatory for a test the requirement to have PARQUET_TEST_DATA env defined.

This is currently not available from python_test_emscripten.sh as we require to mount the filesystem for both Node and ChromeDriver.

What changes are included in this PR?

Skip the test that requires PARQUET_TEST_DATA for emscripten.

Are these changes tested?

Via archery

Are there any user-facing changes?

No

• GitHub Issue: [CI][Python] python_test_emscripten fails with missing PARQUET_TEST_DATA environment variable #43905
• GitHub Issue: [Python][Parquet][Emscripten] PARQUET_TEST_DATA isn't set in test-conda-python-emscripten #43868

[raulcd]

@github-actions crossbow submit test-conda-python-emscripten

[github-actions]

⚠️ GitHub issue #43905 has been automatically assigned in GitHub to PR creator.

[github-actions]

Revision: 3091b519c41ce62e702fbee8f4dc7692a5bc72ff

Submitted crossbow builds: ursacomputing/crossbow @ actions-d04eba703d

Task	Status
test-conda-python-emscripten

[raulcd]

ok, this does not work on emscripten. I have no idea how environment variables work there and / or if they work. We probably should skip this test on that environment. I am also not sure if we should just delete this check, my expectation is that if a test requires this to be set will already fail if it does not find the parquet test data files. @jorisvandenbossche should we just remove this check?

arrow/python/pyarrow/tests/parquet/conftest.py

Lines 34 to 36 in 2767dc5

	if not result:
	raise RuntimeError('Please point the PARQUET_TEST_DATA environment '
	'variable to the test data directory')

edit: added link to check

[pitrou]

cc @kou for insight.

[kou]

I think that we can't inject Parquet test data directory path by environment variable.
Because environment variable doesn't exist on Web browser.

But... we may be able to use https://github.com/pyodide/pyodide/blob/a448c045cf5640c5050e0c3722bafc0d58659fcc/src/js/pyodide.ts#L151-L160 :

diff --git a/python/scripts/run_emscripten_tests.py b/python/scripts/run_emscripten_tests.py
index 1a4b4a4e05..79996b1f92 100644
--- a/python/scripts/run_emscripten_tests.py
+++ b/python/scripts/run_emscripten_tests.py
@@ -19,6 +19,7 @@
 import argparse
 import contextlib
 import http.server
+import json
 import os
 import queue
 import shutil
@@ -153,7 +154,7 @@ class NodeDriver:
         self.execute_js(
             f"""
         const {{ loadPyodide }} = require('{dist_dir}/pyodide.js');
-        let pyodide = await loadPyodide();
+        let pyodide = await loadPyodide({json.dumps({"env": {"PARQUET_TEST_DATA": os.environ.get("PARQUET_TEST_DATA")}})});
         """
         )
 

[raulcd]

@github-actions crossbow submit test-conda-python-emscripten

[github-actions]

⚠️ GitHub issue #43868 has been automatically assigned in GitHub to PR creator.

[github-actions]

Revision: 3d3708aed26eae449356f00dde909863e2c1b12a

Submitted crossbow builds: ursacomputing/crossbow @ actions-8e719208e8

Task	Status
test-conda-python-emscripten

[raulcd]

I was testing locally and it seems that even though the file exists at the location inside the container:

root@4084a63715db:/# ls -lrt /arrow/cpp/submodules/parquet-testing/data/column_chunk_key_value_metadata.parquet 
-rw-rw-r-- 1 1000 1000 400 Sep  2 11:59 /arrow/cpp/submodules/parquet-testing/data/column_chunk_key_value_metadata.parquet

The test fails:

_____________________ test_column_chunk_key_value_metadata _____________________

parquet_test_datadir = PosixPath('/arrow/cpp/submodules/parquet-testing/data')

    def test_column_chunk_key_value_metadata(parquet_test_datadir):
>       metadata = pq.read_metadata(parquet_test_datadir /
                                    'column_chunk_key_value_metadata.parquet')

/lib/python3.12/site-packages/pyarrow/tests/parquet/test_metadata.py:788: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
/lib/python3.12/site-packages/pyarrow/parquet/core.py:2297: in read_metadata
    file_ctx = where = filesystem.open_input_file(where)
pyarrow/_fs.pyx:789: in pyarrow._fs.FileSystem.open_input_file
    ???
pyarrow/error.pxi:155: in pyarrow.lib.pyarrow_internal_check_status
    ???
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

>   ???
E   FileNotFoundError: [Errno 44] Failed to open local file '/arrow/cpp/submodules/parquet-testing/data/column_chunk_key_value_metadata.parquet'. Detail: [errno 44] No such file or directory

pyarrow/error.pxi:92: FileNotFoundError

[jorisvandenbossche]

cc @joemarshall

[jorisvandenbossche]

For setting the env variable, could we modify os.environ directly in python code? (eg in _load_pyarrow_in_runner)

[raulcd]

@github-actions crossbow submit test-conda-python-emscripten

[raulcd]

Pushing a test skip that works locally but happy if @joemarshall or someone else comes with a solution that does not require us to skip the test.

[github-actions]

Revision: 6a5c5143315fe3ef2d1f57ee6de7b70d3307d95c

Submitted crossbow builds: ursacomputing/crossbow @ actions-e9434c5a2d

Task	Status
test-conda-python-emscripten

[jorisvandenbossche]

Skipping the test is certainly fine with me

[pitrou]

This seems relevant:
https://pyodide.org/en/stable/usage/accessing-files.html

[raulcd]

I've been able to have the test passing locally for the NodeDriver with this (and some other minor changes):

diff --git a/python/scripts/run_emscripten_tests.py b/python/scripts/run_emscripten_tests.py
index 53d3dd5..550e1a0 100644
--- a/python/scripts/run_emscripten_tests.py
+++ b/python/scripts/run_emscripten_tests.py
@@ -19,6 +19,7 @@
 import argparse
 import contextlib
 import http.server
+import json
 import os
 import queue
 import shutil
@@ -153,7 +154,8 @@ class NodeDriver:
         self.execute_js(
             f"""
         const {{ loadPyodide }} = require('{dist_dir}/pyodide.js');
-        let pyodide = await loadPyodide();
+        let pyodide = await loadPyodide({json.dumps({"env": {"PARQUET_TEST_DATA": "/test-data"}})});
+        pyodide.mountNodeFS("/test-data", "{os.environ.get("PARQUET_TEST_DATA")}");
         """
         )

but the test still fails for the ChromeDriver, I tried a couple of things, but they just crash, so I am not sure how to do the mount for the ChromeDriver.

[raulcd]

@github-actions crossbow submit test-conda-python-emscripten

[github-actions]

Revision: e918ffa

Submitted crossbow builds: ursacomputing/crossbow @ actions-295c35fb98

Task	Status
test-conda-python-emscripten

[raulcd]

I am marking this as ready for review just skipping the test, I can create an issue as a follow up in order to mount PARQUET_TEST_DATA if someone wants to take that over. @pitrou @jorisvandenbossche are you ok skipping the test to fix CI?

[pitrou]

@raulcd Can you fix this PR's title and description?

[raulcd]

@raulcd Can you fix this PR's title and description?

Sorry about that, I forgot to update the PR title + desc after several iterations. Thanks for double checking @pitrou

[pitrou]

LGTM

[kou]

+1

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit fe39c8f.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 3 possible false positives for unstable benchmarks that are known to sometimes produce them.