Remove and rewrite unnecessary tests in the Polars, Arrow, and S3FS cursor tests

Summary

Part of #1239. Files: polars/, arrow/, s3fs/ (test_cursor.py, test_async_cursor.py, test_result_set.py, test_converter.py, test_reader.py, test_util.py). At master 59169c5 with the cross-check corrections: 289 test functions: 212 KEEP, 55 REMOVE, 6 MERGE, 16 REWRITE. #1238 deletes 4 of the REMOVE rows; the tables list 51 REMOVE, 6 MERGE, 16 REWRITE. Applying them saves about 44 Athena queries per run: 17 in the S3FS per-reader tests, 14 in the Polars chunk tests, 7 in the ArrowCursor as_polars tests. The result-set, converter, and reader rows are offline.

Paths are relative to tests/pyathena/. [original] marks a test that existed at 21cbf05 (v3.15.0).

Order

After Break result sets after failed or cancelled fetches and drop the cross-thread guards #1238 (06adf23). It deletes polars/test_result_set.py::TestAthenaPolarsResultSet::{test_close_during_fetchall, test_close_during_fetchall_generator, test_close_during_first_read} and TestPolarsDataFrameIterator::test_close_during_read, modifies the KEEP TestAthenaPolarsResultSet::test_fetch_after_conversion_error, and adds TestAthenaPolarsResultSet::{test_fetch_after_failed_chunk_read, test_fetchone_interrupted_while_counting} and TestPolarsDataFrameIterator::test_read_after_failed_read. Section 3 is rebased onto it.
After Read UNLOAD timestamps at microseconds #1233 (1dc9b28). It modifies arrow/test_cursor.py::TestArrowCursor::{test_complex_as_polars, test_complex_unload_as_polars} (section 5, REMOVE still holds; only col_date moves to pl.Date), their covers test_complex_as_arrow, test_complex_unload_as_arrow, and test_temporal_types, and polars/test_result_set.py::TestAthenaPolarsResultSet::{test_storage_options_replace_defaults, test_iter_parquet_chunks_raises_when_read_fails_partway} (section 3). It changes the private _read_parquet/_iter_parquet_chunks signatures to take a pl.Schema, which the section 3 REWRITE no longer calls. It also changes polars/test_cursor.py::TestPolarsCursor::test_temporal_types and adds tests that were not audited: TestArrowCursor::test_unload_nested_timestamps, TestPolarsCursor::{test_unload_nested_timestamps, test_unload_schema_option, test_unload_timestamps_with_read_options_polars_1}.
Read CSV dates as date32 and temporal text NULL as null in ArrowCursor #1236 is merged (master 7e5d9ee). It changes col_date to pa.date32() in TestArrowCursor::test_complex_as_arrow and to pl.Date in test_complex_as_polars, removes the #1216 xfail from test_temporal_types, and adds TestArrowCursor::test_temporal_text_null_with_binary and arrow/test_converter.py::test_to_null (not audited). Re-checked at 7e5d9ee: the section 5 REMOVE of test_complex_as_polars still holds, because the Polars dtypes follow from the Arrow table that test_complex_as_arrow asserts.
s3fs/test_cursor.py::TestS3FSCursor::test_on_start_query_execution (section 9) is removed with or after the top-level sub-issue's REWRITE of test_connection.py::test_cursor_accepts_connection_callbacks, which must keep running the callback for every class in CURSOR_CLASSES.
Covers within this sub-issue that are MERGE/REWRITE targets: polars/test_converter.py::test_to_datetimes_not_strict needs the section 3 REWRITE of test_read_csv_ignore_errors and test_read_csv_truncates_timestamps (keep the non-strict read and the OperationalError); TestArrowCursor::test_many_as_polars needs the rewritten TestArrowCursor::test_as_polars; TestS3FSCursor::test_quoted_string_with_comma merges into the rewritten TestS3FSCursor::test_contain_tab_character.

Findings

1. `polars/test_cursor.py`

test	change	criteria	reason	covered by
`TestPolarsCursor::test_as_arrow`	REMOVE	2, 6	`as_arrow()` is `as_polars().to_arrow()`, so this checks Polars' `to_arrow` (2 queries).	`TestPolarsCursor::test_whole_result_reused` (asserts `as_arrow()` values)
`TestPolarsCursor::test_iter_chunks`, `TestPolarsCursor::test_iter_chunks_many_rows`	REMOVE	2	Row totals and "is a DataFrame" checks of CSV chunking with another chunksize add nothing to the consistency test.	`TestPolarsCursor::test_iter_chunks_data_consistency` (with the merged size bound)
`TestPolarsCursor::test_iter_chunks_without_chunksize`	REMOVE	2	Without a chunksize, `iter_chunks()` yields one whole chunk, which is already asserted.	`TestPolarsCursor::test_whole_result_reused` (`[3]` chunk heights)
`TestPolarsCursor::test_iter_chunks_unload`	REMOVE	2	A chunked UNLOAD row total is weaker than the exact sizes, schema, and equality checks of the cover.	`TestPolarsCursor::test_unload_chunks`
`TestPolarsCursor::test_iter_chunks_chunk_sizes`	MERGE	2	Its only extra assertion is `chunk.height <= chunksize`.	into `TestPolarsCursor::test_iter_chunks_data_consistency`
`TestPolarsCursor::test_fetchone_with_chunksize`, `TestPolarsCursor::test_fetchmany_with_chunksize`, `TestPolarsCursor::test_fetchall_with_chunksize`, `TestPolarsCursor::test_iterator_with_chunksize`, `TestPolarsCursor::test_fetchone_with_chunksize_unload`, `TestPolarsCursor::test_iterator_with_chunksize_unload`	REMOVE	2	Row-count-only checks of fetching across chunks; the cover checks values and `rownumber` for CSV and UNLOAD (6 queries).	`TestPolarsCursor::test_fetch_across_chunks`; description: `TestPolarsCursor::test_description_with_chunksize_unload`; iteration: `TestPolarsCursor::test_iterator`

2. `polars/test_async_cursor.py`

test	change	criteria	reason	covered by
`TestAsyncPolarsCursor::test_as_arrow`	REMOVE	2, 6	The same Polars `to_arrow` wrapper through the async future (2 queries).	`polars/test_cursor.py::TestPolarsCursor::test_whole_result_reused`
`TestAsyncPolarsCursor::test_read_options`	REWRITE	1, 6	Keeps the option-plumbing assertions and drops `cursor._executor._max_workers == 1`, which only checks that the constructor stored its argument.	—

3. `polars/test_result_set.py`

#1238 deletes TestAthenaPolarsResultSet::test_close_during_fetchall, TestAthenaPolarsResultSet::test_close_during_fetchall_generator, TestAthenaPolarsResultSet::test_close_during_first_read, and TestPolarsDataFrameIterator::test_close_during_read; they are not listed.

test	change	criteria	reason	covered by
`TestAthenaPolarsResultSet::test_fetchmany_size_not_integer`	REMOVE	2	The size is validated by the base `AthenaResultSet.fetchmany` (`operator.index`) before the Polars `_take_rows` runs.	`test_result_set.py::TestAthenaResultSet::test_fetchmany_size_not_integer`
`TestPolarsDataFrameIterator::test_itertuples_without_column_names`	REMOVE	7	Exercises the `repeat((), length)` branch for a chunk with rows but no column names, which Athena results never have.	—
`TestAthenaPolarsResultSet::test_storage_options_replace_defaults`, `TestAthenaPolarsResultSet::test_csv_storage_options_replace_defaults`	MERGE	1, 5	They assert the kwargs of mocked `polars.scan_csv`/`read_parquet`/`scan_parquet`/`fsspec.open` and that private properties are not computed; the contract that `storage_options` replace PyAthena's is observable end to end.	into `polars/test_cursor.py::TestPolarsCursor::test_as_polars_with_read_kwargs` (parametrized over unload and chunksize)
`TestAthenaPolarsResultSet::test_iter_csv_chunks_raises_when_read_fails_partway`, `TestAthenaPolarsResultSet::test_iter_parquet_chunks_raises_when_read_fails_partway`, `TestAthenaPolarsResultSet::test_csv_read_kwargs_replace_defaults`, `TestAthenaPolarsResultSet::test_txt_new_columns_with_schema_overrides`, `TestAthenaPolarsResultSet::test_read_csv_truncates_timestamps`, `TestAthenaPolarsResultSet::test_read_csv_ignore_errors`	REWRITE	1	Build the result set through its constructor over a local output location, patching only the `_pre_fetch` metadata and `S3FileSystem` with `LocalFileSystem`, and assert through `as_polars()`/`iter_chunks()`: a failed read raises instead of truncating, read kwargs replace the defaults, `schema_overrides` apply to `new_columns`, timestamps truncate (`OperationalError` on a strict failure), and `ignore_errors` reads text; drop `scan_parquet.called is bool(read_options)` and the `__cause__` AttributeError check.	—
`TestPolarsDataFrameIterator::test_itertuples_nested_value_python_does_not_hold`	REWRITE	8	Asserts that the rows before the bad nested value are returned and that the next row raises, without naming `pl.exceptions.PanicException`, Polars' internal failure type.	—

4. `polars/test_converter.py`

test	change	criteria	reason	covered by
`test_to_datetimes_not_strict`	REMOVE	2, 8	The non-strict path is covered through the reader, and the strict half asserts Polars' `InvalidOperationError`.	`polars/test_result_set.py::TestAthenaPolarsResultSet::test_read_csv_ignore_errors`; strict failure: `polars/test_result_set.py::TestAthenaPolarsResultSet::test_read_csv_truncates_timestamps` (both rewritten in section 3)

5. `arrow/test_cursor.py`

test	change	criteria	reason	covered by
`TestArrowCursor::test_as_polars`	REWRITE	4, 6	`ArrowCursor.as_polars()` is `pl.from_arrow(self._table)`; keep one CSV query that asserts the Polars values and drop the unload parameter (2 → 1 query).	—
`TestArrowCursor::test_many_as_polars`	REMOVE	2, 4, 6	The 10,000-row values are asserted on the Arrow table, and the Polars side is `pl.from_arrow` (2 queries).	`TestArrowCursor::test_many_as_arrow` [original], `TestArrowCursor::test_as_polars` (rewritten)
`TestArrowCursor::test_complex_as_polars`, `TestArrowCursor::test_complex_unload_as_polars`	REMOVE	6, 8, 2	They assert the Polars dtypes and values that `pl.from_arrow` makes from the Arrow table, which tests Polars; `test_complex_unload_as_polars` failed once on a Polars-version `Map` dtype (#1239).	`TestArrowCursor::test_complex_as_arrow` [original], `TestArrowCursor::test_complex_unload_as_arrow` [original]
`TestArrowCursor::test_timeout_parameters`, `TestArrowCursor::test_timeout_parameters_float`	REMOVE	1, 6, 4	They assert the private `_connect_timeout`/`_request_timeout` of the cursor and result set; the float variant tests Python (2 queries).	`TestArrowCursor::test_read_options` (offline, same plumbing with the `execute()` override)
`TestArrowCursor::test_fetch_converts_each_value_once`	REWRITE	1	Regression for `d1dee80` (values converted twice): assert the rows of a non-idempotent converter, e.g. `lambda v: v + "!"` gives `[("a!",), ("b!",)]`, instead of counting converter calls.	—

6. `arrow/test_async_cursor.py`

test	change	criteria	reason	covered by
`TestAsyncArrowCursor::test_binary_null_vs_empty`	REMOVE	2, 4	Same `AthenaArrowResultSet` and query as the sync test; the async cursor adds no conversion code.	`arrow/test_cursor.py::TestArrowCursor::test_binary_null_vs_empty`

7. `arrow/test_result_set.py`

test	change	criteria	reason	covered by
`TestAthenaArrowResultSet::test_fetch_after_close`	REMOVE	5, 2	The query execution is FAILED, so the result set has no rows and the test passes even if `close()` does nothing.	`test_result_set.py::TestAthenaResultSet::test_fetch_after_close`
`TestAthenaArrowResultSet::test_fetch_across_batches`	MERGE	2, 1	Its unique part is mixing methods (`fetchmany(2)` then `fetchone()`) and the calls after exhaustion (`fetchmany() == []`, `fetchall() == []`, `fetchone() is None`), which depend on Arrow's own `_fetch`.	into `TestAthenaArrowResultSet::test_fetch_across_empty_batches` (add the mixed and after-exhaustion asserts)
`TestAthenaArrowResultSet::test_open_csv_closes_file_when_stream_fails`	REMOVE	1, 7	Calls the private `_open_csv` on an `__new__` instance and asserts `file.close.assert_called_once_with()` on a mock.	—
`TestAthenaArrowResultSet::test_read_csv_timestamps_with_the_same_name`	MERGE	1	Positional typing of same-named TIMESTAMP columns, tested by bypassing `__init__` and patching `_fetch_all_rows_as_csv`; add a duplicated TIMESTAMP column to the integration query, whose `managed` parameter already reads through GetQueryResults.	into `arrow/test_cursor.py::TestArrowCursor::test_duplicate_column_names`

8. `arrow/test_reader.py`

test	change	criteria	reason	covered by
`TestParquetChunkReader::test_releases_files_opened_ahead`	REWRITE	1	Asserts that more files were opened than received and that all are closed, instead of the exact count `4 if corrupt else 2`, which pins a lookahead of one; the `close` parameter folds into the cover.	close parameter: `TestParquetChunkReader::test_close`
`TestParquetChunkReader::test_open_failure_opens_no_further_file`	REMOVE	1, 7	Patches the private `_open` and asserts which paths were attempted after a failure.	—
`TestParquetChunkReader::test_prefetched_file`	REWRITE	1	Keeps the one-GET request contract for a whole-read file (`fetched == [(0, size)]`) and drops `fs.opened == [(None, "none")]`, the internal `_open` arguments.	—
`TestParquetChunkReader::test_streamed_file_row_groups`	REWRITE	1, 8	Asserts laziness instead of ranges that depend on PyArrow's pre-buffer coalescing: after the first chunk fewer bytes than the file size are fetched, and the total is 300 rows.	—

9. `s3fs/test_cursor.py`

test	change	criteria	reason	covered by
`TestS3FSCursor::test_query_id`	REMOVE	2, 6	`query_id` is None before `execute()` and set after, in base cursor code.	`TestS3FSCursor::test_on_start_query_execution_execute` (`callback_query_id == cursor.query_id`)
`TestS3FSCursor::test_description_without_execute`	REMOVE	2, 6	`description` is inherited unchanged from the base cursor.	`test_cursor.py::TestCursor::test_description_initial` [original]
`TestS3FSCursor::test_description_with_select`	REMOVE	2	A one-column description is a subset of the 19-column description already asserted.	`TestS3FSCursor::test_complex`, `TestS3FSCursor::test_show_columns`
`TestS3FSCursor::test_executemany`	REWRITE	—	Permanently `@pytest.mark.skip` because it needs a nonexistent `insert_test` table; use the `empty_table` fixture and assert the inserted rows as `arrow/test_cursor.py::TestArrowCursor::test_executemany` does, or delete it.	—
`TestS3FSCursor::test_on_start_query_execution`	REMOVE	2, 4	The cursor-level callback is stored and called by shared `common.py` code, not S3FS code.	`test_connection.py::test_cursor_accepts_connection_callbacks` (rewritten by the top-level sub-issue), `test_cursor.py::TestCursor::test_connection_level_callback`
`TestS3FSCursor::test_contain_tab_character`	REWRITE	4	One SELECT with the tab, newline, and tab+newline columns, asserting each value for both readers (4 → 1 query per reader); the docstring's tab-delimiter claim is wrong, since CSV results use commas.	—
`TestS3FSCursor::test_quoted_string_with_comma`	MERGE	2, 4	`'a,b,c'` becomes another column of the rewritten special-character query.	into `TestS3FSCursor::test_contain_tab_character` (rewritten)
`TestS3FSCursor::test_basic_query_with_reader`, `TestS3FSCursor::test_multiple_columns_with_reader`	REMOVE	2, 4	Both readers returning a basic row is already proven on a real query for both readers (2 queries each).	`TestS3FSCursor::test_null_vs_empty_string`; default reader: `TestS3FSCursor::test_complex`
`TestS3FSCursor::test_null_with_empty_string_as_null_reader`, `TestS3FSCursor::test_null_with_athena_reader`, `TestS3FSCursor::test_empty_string_with_empty_string_as_null_reader`, `TestS3FSCursor::test_empty_string_with_athena_reader`, `TestS3FSCursor::test_mixed_values_with_athena_reader`	REMOVE	2, 4	Each checks one cell of the NULL vs `''` matrix that one parametrized query checks in full (5 queries).	`TestS3FSCursor::test_null_vs_empty_string`

10. `s3fs/test_async_cursor.py`

test	change	criteria	reason	covered by
`TestAsyncS3FSCursor::test_null_vs_empty_string`	REMOVE	2, 4	Same `AthenaS3FSResultSet` and readers as the sync test; the `csv_reader` plumbing to the async result set is covered offline (2 queries).	`s3fs/test_cursor.py::TestS3FSCursor::test_null_vs_empty_string`, `TestAsyncS3FSCursor::test_read_options`
`TestAsyncS3FSCursor::test_complex`, `TestAsyncS3FSCursor::test_show_columns`	REMOVE	2, 4	Identical queries and assertions to the sync tests on the same result-set class.	`s3fs/test_cursor.py::TestS3FSCursor::test_complex`, `s3fs/test_cursor.py::TestS3FSCursor::test_show_columns`

11. `s3fs/test_result_set.py`

test	change	criteria	reason	covered by
`TestAthenaS3FSResultSet::test_fetch_across_batches`	REMOVE	2	`fetchmany`/`fetchone`/`fetchall`/`rownumber` over a mocked CSV are asserted by the cover across 8 parameters.	`TestAthenaS3FSResultSet::test_reader_null_conversion`
`TestAthenaS3FSResultSet::test_get_csv_open_options`	REWRITE	1, 6	Asserts only the user-visible precedence (prefetch on for PyAthena's filesystems, off with `block_size` or a user subclass) instead of calling the private `_get_csv_open_options` and comparing it with the `_PREFETCH_OPTIONS` constant.	—

12. `s3fs/test_reader.py`

test	change	criteria	reason	covered by
`TestEmptyStringAsNullCSVReader::test_basic_parsing`, `TestEmptyStringAsNullCSVReader::test_tab_delimiter`, `TestEmptyStringAsNullCSVReader::test_quoted_field_with_comma`, `TestEmptyStringAsNullCSVReader::test_quoted_field_with_newline`, `TestEmptyStringAsNullCSVReader::test_escaped_quote`, `TestEmptyStringAsNullCSVReader::test_empty_file`	REMOVE	6	`EmptyStringAsNullCSVReader` wraps Python's `csv.reader`, so these test the standard library.	basic and close: `TestEmptyStringAsNullCSVReader::test_context_manager`; tab: `s3fs/test_result_set.py::TestAthenaS3FSResultSet::test_reader_null_conversion[.txt-EmptyStringAsNullCSVReader]`
`TestEmptyStringAsNullCSVReader::test_implements_iterator_protocol`, `TestAthenaCSVReader::test_implements_iterator_protocol`	REMOVE	6	`isinstance(reader, Iterator)` asserts the class declaration.	—
`TestEmptyStringAsNullCSVReader::test_close`, `TestAthenaCSVReader::test_close`	REMOVE	2	`__exit__` calls `close()`, and the cover asserts that iteration stops afterwards.	`TestEmptyStringAsNullCSVReader::test_context_manager`, `TestAthenaCSVReader::test_context_manager`
`TestAthenaCSVReader::test_no_trailing_null`	REMOVE	2	Identical input and output (`"a,b,c\n"`) to a row of the basic test.	`TestAthenaCSVReader::test_basic_parsing`
`TestAthenaCSVReader::test_quoted_field_with_newline`	REMOVE	2	One embedded newline is subsumed by the multi-newline and multi-row cases.	`TestAthenaCSVReader::test_quoted_field_with_multiple_newlines`, `TestAthenaCSVReader::test_multiple_rows_with_multiline_field`

Notes

Flaky tests named in Audit the test suite for unnecessary and flaky tests #1239: arrow/test_cursor.py::TestArrowCursor::test_complex_unload_as_polars is REMOVE (section 5). polars/test_result_set.py::TestPolarsDataFrameIterator::test_itertuples_value_python_does_not_hold is KEEP; since 90c94d9 it matches only "year", which both Python messages contain. polars/test_cursor.py::TestPolarsCursor::test_null_vs_empty_string is KEEP; its one failure (run 38560114057) belongs to the flakiness classification.
The per-cursor copies of the original Arrow suite in the Polars and S3FS cursors (sync and async: fetch methods and iterator over CSV and UNLOAD, arraysize, description, query_execution, poll, bad_query, cancel, open_close, no_ops, empty_result, executemany, show_columns) are KEEP because they follow the original per-cursor shape. They are the largest remaining pool of queries, about 40 for Polars alone.
test_cancel in all 6 cursors uses a random 5-10 s sleep and a cross join (criteria 3 and 4). It is KEEP because cross-thread cancel is a supported contract.
test_read_options in the Polars, Arrow, and S3FS cursors (sync and async) patches the private _execute/_poll and asserts the result-set constructor kwargs. It is KEEP as the only check that block_size, cache_type, s3_max_workers, chunksize, the timeouts, and csv_reader reach the reader, with execute() overriding the cursor.
arrow/test_cursor.py::TestArrowCursor::test_iceberg_table [original] runs about 12 queries (DDL, DML, MERGE, VACUUM) through the base cursor path.
arrow/test_result_set.py::TestAthenaArrowResultSet::test_create_s3_file_system_uses_session_credentials bypasses __init__ and reads PyArrow's __reduce__ internals (criteria 1 and 8). It is KEEP as the only check for the ArrowCursor's native S3 filesystem ignores the connection's profile and role credentials #1115 credentials bug.
polars/test_result_set.py::TestPolarsDataFrameIterator::test_itertuples_large_chunk uses the private _ROW_BLOCK_SIZE; integration results never cross a 10,000-row block.
The match= text in arrow/test_reader.py::TestParquetChunkReader::{test_columns_not_top_level, test_invalid_chunksize}, the Polars and S3FS test_ambiguous_time_zone_value, and s3fs/test_result_set.py::TestAthenaS3FSResultSet::test_reader_failure_closes_file is not a documented contract; the exception types stay asserted and the match= arguments can be dropped.
Unused after the change: in polars/test_result_set.py, the patch targets pyathena.polars.result_set.fsspec.open and to_column_info after the storage-options MERGE (Break result sets after failed or cancelled fetches and drop the cross-thread guards #1238 already removes the threading, ThreadPoolExecutor, and BytesIO imports, and its test_read_after_failed_read keeps inspect); datetime in arrow/test_result_set.py; datetime and Decimal in s3fs/test_async_cursor.py; collections.abc.Iterator in s3fs/test_reader.py. ImmediateExecutor, _RecordingFile, and _RecordingFileSystem in arrow/test_reader.py stay in use. No fixture or tests/pyathena/util.py helper becomes unused.
Each row's cover is re-checked before the row is applied. A row whose cover does not hold becomes MERGE or KEEP.

Check that storage options reach chunked CSV reads and that fetches stop

Read the local Polars results through GetQueryResults and as_polars()

Check the read options and conversions that the removed tests protected

Test the cleanup and routing that the follow-up review found untested

Read the S3FS test metadata through GetQueryResults

The row of section 9 that #1252 kept until #1243, s3fs/test_cursor.py::TestS3FSCursor::test_on_start_query_execution, is removed by #1261: since #1253, test_connection.py::test_cursor_accepts_connection_callbacks runs the connection's callbacks through S3FSCursor.

Check the callback that the connection's cursor_kwargs give a cursor

Run the callbacks given through cursor_kwargs for every cursor class

Check that cursor_kwargs callbacks win over the connection's

Keep a case of callbacks given only through cursor_kwargs

Remove the S3FS cursor-level callback test

Check the callback that the connection's cursor_kwargs give a cursor

Run the callbacks given through cursor_kwargs for every cursor class

Check that cursor_kwargs callbacks win over the connection's

Keep a case of callbacks given only through cursor_kwargs

Navigation Menu

Repository navigation

Remove and rewrite unnecessary tests in the Polars, Arrow, and S3FS cursor tests #1246

Summary

Order

Findings

1. `polars/test_cursor.py`

2. `polars/test_async_cursor.py`

3. `polars/test_result_set.py`

4. `polars/test_converter.py`

5. `arrow/test_cursor.py`

6. `arrow/test_async_cursor.py`

7. `arrow/test_result_set.py`

8. `arrow/test_reader.py`

9. `s3fs/test_cursor.py`

10. `s3fs/test_async_cursor.py`

11. `s3fs/test_result_set.py`

12. `s3fs/test_reader.py`

Notes

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Remove and rewrite unnecessary tests in the Polars, Arrow, and S3FS cursor tests #1246

Description

Summary

Order

Findings

1. polars/test_cursor.py

2. polars/test_async_cursor.py

3. polars/test_result_set.py

4. polars/test_converter.py

5. arrow/test_cursor.py

6. arrow/test_async_cursor.py

7. arrow/test_result_set.py

8. arrow/test_reader.py

9. s3fs/test_cursor.py

10. s3fs/test_async_cursor.py

11. s3fs/test_result_set.py

12. s3fs/test_reader.py

Notes

Activity

laughingman7743 commented on Oct 18, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

1. `polars/test_cursor.py`

2. `polars/test_async_cursor.py`

3. `polars/test_result_set.py`

4. `polars/test_converter.py`

5. `arrow/test_cursor.py`

6. `arrow/test_async_cursor.py`

7. `arrow/test_result_set.py`

8. `arrow/test_reader.py`

9. `s3fs/test_cursor.py`

10. `s3fs/test_async_cursor.py`

11. `s3fs/test_result_set.py`

12. `s3fs/test_reader.py`