robin-sparkless 4.0.0

PySpark-like DataFrame API in Rust on Polars; no JVM.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37

from typing import Any, cast


def test_sql_basic_select_schema_matches_dataframe_select(spark, table_prefix) -> None:
    """BUG-021 regression: basic SQL SELECT should project the correct schema.

    This test ensures that:
    - An explicit column list in SQL (SELECT id, name, age ...) returns exactly
      those columns, in the same order.
    - SELECT * returns all columns from the underlying table.
    """
    # SparkSession not needed - using spark fixture

    try:
        # Create a DataFrame with an extra column that should not appear in the
        # projected SELECT list.
        rows = [{"id": 1, "name": "Alice", "age": 25}]
        df = spark.createDataFrame(rows)

        df = cast(Any, df.withColumn("salary", df.id * 1000))

        # Save as table and query via SQL (unique name for shared session)
        tbl = f"{table_prefix}_test_table"
        df.write.mode("overwrite").saveAsTable(tbl)

        # Explicit projection should only include the requested columns
        result = spark.sql(f"SELECT id, name, age FROM {tbl}")
        assert result.columns == ["id", "name", "age"]
        assert len(result.schema.fields) == 3

        # SELECT * should include all columns from the table
        result_star = spark.sql(f"SELECT * FROM {tbl}")
        assert set(result_star.columns) == {"id", "name", "age", "salary"}
        assert len(result_star.schema.fields) == 4
    except Exception:
        # Cleanup handled by fixture
        pass