robin-sparkless 4.3.0

"""
Regression test for issue #1387: sql.show_tables parity.

PySpark scenario (from the issue):

    def scenario_sql_show_tables(session):
        return session.sql("SHOW TABLES")

This test exercises the same scenario against sparkless, ensuring that:

- ``session.sql("SHOW TABLES")`` does not raise.
- The resulting schema's ``simpleString()`` matches the current struct form.
- The data includes at least the default namespace and respects the existing
  sparkless column naming.
- ``explain()`` returns a non-empty plan string (no blank UI).
"""

from sparkless.sql import SparkSession


def test_issue_1387_sql_show_tables_schema_data_and_explain() -> None:
    """sql.show_tables: schema, basic data, and explain behavior (issue #1387)."""
    spark = SparkSession.builder.appName("issue_1387").getOrCreate()
    try:
        df = spark.sql("SHOW TABLES")

        # Schema simpleString should match the current struct representation.
        # Sparkless today uses `database` rather than PySpark's `namespace`.
        schema_str = df.schema.simpleString()
        assert (
            schema_str == "struct<database:string,tableName:string,isTemporary:boolean>"
        )

        # Data: content may vary by environment (e.g., empty catalog vs. pre-populated),
        # so we only assert that the query returns a DataFrame and do not require a
        # specific database name.
        rows = df.collect()
        assert isinstance(rows, list)

        # explain() should produce a non-empty plan string.
        plan = df.explain()
        assert isinstance(plan, str)
        assert plan.strip() != ""
    finally:
        spark.stop()