Add an ARRAY<string> column to one_row_complex through its definition only

[laughingman7743]

WHAT

This PR adds col_array_string ARRAY<string>, with the value ["a", "b"], to ONE_ROW_COMPLEX in tests/pyathena/tables.py. The diff is two lines, both in the table definition: the column and its value. No test or rule changes.

WHY

Part of #848, for #834. Stacked on #876.

#848's validation plan asks to "show that adding a column to a shared table is a one-place change". Before #874, #875, and #876, adding this column meant editing four places:

1. the Hive text file one_row_complex.gz, which uses \002 separators;
2. the DDL template;
3. about 20 hand-written expected rows, schemas, and dtype lists;
4. the SQLAlchemy column count.

After this change, the tests below cover the new column in every representation.

Test	What it now also asserts for col_array_string
TestCursor.test_complex, S3FS ×2, pandas.util.as_pandas	["a", "b"]: the cursors parse Athena's rendering [a, b] into strings
TestSQLAlchemyAthena.test_reflect_select	reflected AthenaArray with String items; value ["a", "b"]
PandasCursor CSV	"[a, b]" as a pandas string column
PandasCursor UNLOAD	["a", "b"], with dtype object
ArrowCursor CSV (rows, as_arrow, as_polars)	"[a, b]", pa.string() / pl.String
ArrowCursor UNLOAD	["a", "b"], with pa.list_(pa.field("array_element", pa.string())) / pl.List(pl.String)
PolarsCursor	nothing: arrays are outside the columns these tests select, as before

TEST

Tested commit: bcbf874, then rebased onto the simplified #875/#876 as cbd12b1 (tested again), then onto later #875/#876 heads; current head c3b08bb. The diff is unchanged.

• just lint passed.
• Against AWS, from this worktree, pytest -n 4 on tests/pyathena/test_cursor.py, both s3fs cursor test files, pandas/test_util.py, sqlalchemy/test_base.py, and the pandas, arrow, and polars test_cursor.py files, with -k "complex or as_pandas or reflect or table_names or view_names or like or escape or char_length or cast_as_varbinary or filter": 104 passed on bcbf874 and again on cbd12b1. This covers every test that derives one_row_complex expectations, plus the other tests that read one_row_complex or list the schema's tables.
• Against AWS, pytest -n 4 tests/pyathena/test_glue.py tests/pyathena/test_cursor.py tests/pyathena/aio/test_cursor.py tests/pyathena/sqlalchemy/test_base.py tests/pyathena/aio/sqlalchemy/test_base.py -k "glue or throttl or list_table_metadata or get_table_metadata": 43 passed. These tests compare the schema-wide Glue and Athena metadata, which now include the new column.
• Not run locally: the full suites. When the PR is Ready, AWS CI runs the PyAthena suite. It skips the SQLAlchemy compliance suites, because no SQLAlchemy path changes here.

🤖 Generated with Claude Code

[laughingman7743]

Self-review round one (implementation behavior): CLEAN

Base 3202ab5 (#876 head, the stacked base), head bcbf874. The full diff is 2 lines.

• Generation: Table.data_file writes ["a", "b"] as list<string> and the new round-trip check accepts it. The DDL gains col_array_string ARRAY<string>.
• Derived tests: all 19 whole-row tests pick the column up through Selection. The PolarsCursor tests leave it out through COMPLEX_FAMILIES.
• Other readers of the table:
  • The tests that select explicit one_row_complex columns are unaffected.
  • meta.reflect() over the schema (test_get_table_names/test_get_view_names) reflects the new array type.
  • The Glue-vs-Athena and throttled-metadata comparisons list the table with the new column on both sides.
• AWS results: 104 passed for the derived-expectation, reflection, and one_row_complex column tests, and 43 passed for the Glue, throttling, and table-metadata tests.

[laughingman7743]

Self-review round two (claims, callers, AWS operations): CLEAN

Base 3202ab5, head bcbf874.

Claims checked against the AWS run on this head

• Each row of the description's table is a derived expectation that passed on AWS:
  • ["a", "b"] for the Python-object cursors and SQLAlchemy
  • "[a, b]" for the CSV paths
  • pa.list_(pa.field("array_element", pa.string())) / pl.List(pl.String) for Arrow UNLOAD
  • object dtype for pandas UNLOAD
• "Two lines, both in the table definition": the diff touches only tests/pyathena/tables.py.
• "Four places before": the gzipped Hive text row, the Jinja DDL, the hand-written expectations in about 20 tests, and len(one_row_complex.c) == 16. All four existed before Define the shared test tables in Python and generate their data #874.

Operational effect: none beyond one more column in one Parquet file. The number of setup statements and uploads is unchanged.

CI scope: no SQLAlchemy or Spark path changes, so on Ready the PyAthena suite runs without Spark.

[laughingman7743]

Independent review (relayed Codex result): CLEAN

Reviewer: Codex CLI 0.157.1 (codex exec -s read-only), model gpt-6-astra, session 01a0e7b6-1295-7331-87c3-e21bcaa43fd8.

Scope: base 3202ab5, head bcbf874. The review ran on a detached snapshot without .env, and the prompt left out the PR framing. The snapshot was still clean at the head afterwards. This was a static review only.

Covered, as reported:

• DDL, Parquet generation, and the Arrow round-trip guard.
• All consumers under tests/, including metadata, Glue, reflection, and schema listings, plus assertions that depend on column count or position.
• The derived rows, descriptions, pandas dtypes, Arrow schemas, Polars dtypes, and SQLAlchemy array element types.
• The default, S3FS, pandas, Arrow, and Polars converter and result-set paths, including UNLOAD.

Result: "No broken or newly vacuous assertions found. The new value's expectations match the applicable cursor paths, and ["a", "b"] is losslessly representable by pa.list_(pa.string())."

[laughingman7743]

Closing without merging, by the maintainer's decision. Deriving the expected results from the table definitions required tests/pyathena/expected.py to re-model how each cursor returns every type: Athena's text rendering, JSON parsing, and the pandas, Arrow, and Polars types. The module grew large enough to need tests of its own, which is more complexity than the one-place column addition is worth. The hand-written expectations stay. See #848 for the summary. The branch is kept for reference.