python/tests/test_sql.py

# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License.  You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied.  See the License for the
# specific language governing permissions and limitations
# under the License.
import gzip
from pathlib import Path

import numpy as np
import pyarrow as pa
import pyarrow.dataset as ds
import pytest
from datafusion import SessionContext, col, udf
from datafusion.object_store import Http
from pyarrow.csv import write_csv

from . import generic as helpers


def test_no_table(ctx):
    with pytest.raises(
        ValueError,
        match=r"^Error during planning: table 'datafusion.public.b' not found$",
    ):
        ctx.sql("SELECT a FROM b").collect()


def test_register_csv(ctx, tmp_path):
    path = tmp_path / "test.csv"
    gzip_path = tmp_path / "test.csv.gz"

    table = pa.Table.from_arrays(
        [
            [1, 2, 3, 4],
            ["a", "b", "c", "d"],
            [1.1, 2.2, 3.3, 4.4],
        ],
        names=["int", "str", "float"],
    )
    write_csv(table, path)

    with Path.open(path, "rb") as csv_file, gzip.open(gzip_path, "wb") as gzipped_file:
        gzipped_file.writelines(csv_file)

    ctx.register_csv("csv", path)
    ctx.register_csv("csv1", str(path))
    ctx.register_csv(
        "csv2",
        path,
        has_header=True,
        delimiter=",",
        schema_infer_max_records=10,
    )
    ctx.register_csv(
        "csv_gzip",
        gzip_path,
        file_extension="gz",
        file_compression_type="gzip",
    )

    alternative_schema = pa.schema(
        [
            ("some_int", pa.int16()),
            ("some_bytes", pa.string()),
            ("some_floats", pa.float32()),
        ]
    )
    ctx.register_csv("csv3", path, schema=alternative_schema)

    assert ctx.catalog().schema().names() == {
        "csv",
        "csv1",
        "csv2",
        "csv3",
        "csv_gzip",
    }

    for table in ["csv", "csv1", "csv2", "csv_gzip"]:
        result = ctx.sql(f"SELECT COUNT(int) AS cnt FROM {table}").collect()
        result = pa.Table.from_batches(result)
        assert result.to_pydict() == {"cnt": [4]}

    result = ctx.sql("SELECT * FROM csv3").collect()
    result = pa.Table.from_batches(result)
    assert result.schema == alternative_schema

    with pytest.raises(ValueError, match="delimiter must be a single character"):
        ctx.register_csv("csv4", path, delimiter="wrong")

    with pytest.raises(
        ValueError,
        match="file_compression_type must be one of: gzip, bz2, xz, zstd",
    ):
        ctx.register_csv("csv4", path, file_compression_type="rar")


def test_register_csv_list(ctx, tmp_path):
    path = tmp_path / "test.csv"
    second_path = tmp_path / "test2.csv"

    int_values = [1, 2, 3, 4]
    table = pa.Table.from_arrays(
        [
            int_values,
            ["a", "b", "c", "d"],
            [1.1, 2.2, 3.3, 4.4],
        ],
        names=["int", "str", "float"],
    )
    write_csv(table, path)
    write_csv(table, second_path)
    ctx.register_csv("csv", path)

    csv_df = ctx.table("csv")
    expected_count = csv_df.count() * 2
    ctx.register_csv(
        "double_csv",
        path=[
            path,
            second_path,
        ],
    )

    double_csv_df = ctx.table("double_csv")
    actual_count = double_csv_df.count()
    assert actual_count == expected_count

    int_sum = ctx.sql("select sum(int) from double_csv").to_pydict()[
        "sum(double_csv.int)"
    ][0]
    assert int_sum == 2 * sum(int_values)


def test_register_http_csv(ctx):
    url = "https://raw.githubusercontent.com/ibis-project/testing-data/refs/heads/master/csv/diamonds.csv"
    ctx.register_object_store("", Http(url))
    ctx.register_csv("remote", url)
    assert ctx.table_exist("remote")
    res, *_ = ctx.sql("SELECT COUNT(*) AS total FROM remote").to_pylist()
    assert res["total"] > 0


def test_register_parquet(ctx, tmp_path):
    path = helpers.write_parquet(tmp_path / "a.parquet", helpers.data())
    ctx.register_parquet("t", path)
    ctx.register_parquet("t1", str(path))
    assert ctx.catalog().schema().names() == {"t", "t1"}

    result = ctx.sql("SELECT COUNT(a) AS cnt FROM t").collect()
    result = pa.Table.from_batches(result)
    assert result.to_pydict() == {"cnt": [100]}


@pytest.mark.filterwarnings(
    "ignore:using literals for table_partition_cols data types:DeprecationWarning"
)
@pytest.mark.parametrize(
    ("path_to_str", "legacy_data_type"), [(True, False), (False, False), (False, True)]
)
def test_register_parquet_partitioned(ctx, tmp_path, path_to_str, legacy_data_type):
    dir_root = tmp_path / "dataset_parquet_partitioned"
    dir_root.mkdir(exist_ok=False)
    (dir_root / "grp=a").mkdir(exist_ok=False)
    (dir_root / "grp=b").mkdir(exist_ok=False)

    table = pa.Table.from_arrays(
        [
            [1, 2, 3, 4],
            ["a", "b", "c", "d"],
            [1.1, 2.2, 3.3, 4.4],
        ],
        names=["int", "str", "float"],
    )
    pa.parquet.write_table(table.slice(0, 3), dir_root / "grp=a/file.parquet")
    pa.parquet.write_table(table.slice(3, 4), dir_root / "grp=b/file.parquet")

    dir_root = str(dir_root) if path_to_str else dir_root

    partition_data_type = "string" if legacy_data_type else pa.string()

    if legacy_data_type:
        with pytest.warns(DeprecationWarning):  # noqa: PT030
            ctx.register_parquet(
                "datapp",
                dir_root,
                table_partition_cols=[("grp", partition_data_type)],
                parquet_pruning=True,
                file_extension=".parquet",
            )
    else:
        ctx.register_parquet(
            "datapp",
            dir_root,
            table_partition_cols=[("grp", partition_data_type)],
            parquet_pruning=True,
            file_extension=".parquet",
        )

    assert ctx.catalog().schema().names() == {"datapp"}

    result = ctx.sql("SELECT grp, COUNT(*) AS cnt FROM datapp GROUP BY grp").collect()
    result = pa.Table.from_batches(result)

    rd = result.to_pydict()
    assert dict(zip(rd["grp"], rd["cnt"], strict=False)) == {"a": 3, "b": 1}


@pytest.mark.parametrize("path_to_str", [True, False])
def test_register_dataset(ctx, tmp_path, path_to_str):
    path = helpers.write_parquet(tmp_path / "a.parquet", helpers.data())
    path = str(path) if path_to_str else path
    dataset = ds.dataset(path, format="parquet")

    ctx.register_dataset("t", dataset)
    assert ctx.catalog().schema().names() == {"t"}

    result = ctx.sql("SELECT COUNT(a) AS cnt FROM t").collect()
    result = pa.Table.from_batches(result)
    assert result.to_pydict() == {"cnt": [100]}


def test_register_json(ctx, tmp_path):
    path = Path(__file__).parent.resolve()
    test_data_path = Path(path) / "data_test_context" / "data.json"
    gzip_path = tmp_path / "data.json.gz"

    with (
        Path.open(test_data_path, "rb") as json_file,
        gzip.open(gzip_path, "wb") as gzipped_file,
    ):
        gzipped_file.writelines(json_file)

    ctx.register_json("json", test_data_path)
    ctx.register_json("json1", str(test_data_path))
    ctx.register_json(
        "json2",
        test_data_path,
        schema_infer_max_records=10,
    )
    ctx.register_json(
        "json_gzip",
        gzip_path,
        file_extension="gz",
        file_compression_type="gzip",
    )
    ctx.register_json(
        "json_gzip1",
        str(gzip_path),
        file_extension="gz",
        file_compression_type="gzip",
    )

    alternative_schema = pa.schema(
        [
            ("some_int", pa.int16()),
            ("some_bytes", pa.string()),
            ("some_floats", pa.float32()),
        ]
    )
    ctx.register_json("json3", path, schema=alternative_schema)

    assert ctx.catalog().schema().names() == {
        "json",
        "json1",
        "json2",
        "json3",
        "json_gzip",
        "json_gzip1",
    }

    for table in ["json", "json1", "json2", "json_gzip"]:
        result = ctx.sql(f'SELECT COUNT("B") AS cnt FROM {table}').collect()
        result = pa.Table.from_batches(result)
        assert result.to_pydict() == {"cnt": [3]}

    result = ctx.sql("SELECT * FROM json3").collect()
    result = pa.Table.from_batches(result, alternative_schema)
    assert result.schema == alternative_schema

    with pytest.raises(
        ValueError,
        match="file_compression_type must be one of: gzip, bz2, xz, zstd",
    ):
        ctx.register_json("json4", gzip_path, file_compression_type="rar")


def test_register_avro(ctx):
    path = "testing/data/avro/alltypes_plain.avro"
    ctx.register_avro("alltypes_plain", path)
    result = ctx.sql(
        "SELECT SUM(tinyint_col) as tinyint_sum FROM alltypes_plain"
    ).collect()
    result = pa.Table.from_batches(result).to_pydict()
    assert result["tinyint_sum"][0] > 0

    alternative_schema = pa.schema(
        [
            pa.field("id", pa.int64()),
        ]
    )

    ctx.register_avro(
        "alltypes_plain_schema",
        path,
        schema=alternative_schema,
    )
    result = ctx.sql("SELECT * FROM alltypes_plain_schema").collect()
    result = pa.Table.from_batches(result)
    assert result.schema == alternative_schema


def test_execute(ctx, tmp_path):
    data = [1, 1, 2, 2, 3, 11, 12]

    # single column, "a"
    path = helpers.write_parquet(tmp_path / "a.parquet", pa.array(data))
    ctx.register_parquet("t", path)

    assert ctx.catalog().schema().names() == {"t"}

    # count
    result = ctx.sql("SELECT COUNT(a) AS cnt FROM t WHERE a IS NOT NULL").collect()
    ctx.sql("SELECT COUNT(a) AS cnt FROM t WHERE a IS NOT NULL").show()

    expected_schema = pa.schema([("cnt", pa.int64(), False)])
    expected_values = pa.array([7], type=pa.int64())
    expected = [pa.RecordBatch.from_arrays([expected_values], schema=expected_schema)]

    assert result == expected

    # where
    expected_values = pa.array([2], type=pa.int64())
    expected = [pa.RecordBatch.from_arrays([expected_values], schema=expected_schema)]
    result = ctx.sql("SELECT COUNT(a) AS cnt FROM t WHERE a > 10").collect()
    assert result == expected

    # group by
    results = ctx.sql(
        "SELECT CAST(a as int) AS a, COUNT(a) AS cnt FROM t GROUP BY a"
    ).collect()

    # group by returns batches
    result_keys = []
    result_values = []
    for result in results:
        pydict = result.to_pydict()
        result_keys.extend(pydict["a"])
        result_values.extend(pydict["cnt"])

    result_keys, result_values = (
        list(t)
        for t in zip(
            *sorted(zip(result_keys, result_values, strict=False)), strict=False
        )
    )

    assert result_keys == [1, 2, 3, 11, 12]
    assert result_values == [2, 2, 1, 1, 1]

    # order by
    result = ctx.sql(
        "SELECT a, CAST(a AS int) AS a_int FROM t ORDER BY a DESC LIMIT 2"
    ).collect()
    expected_a = pa.array([50.0219, 50.0152], pa.float64())
    expected_cast = pa.array([50, 50], pa.int32())
    expected = [pa.RecordBatch.from_arrays([expected_a, expected_cast], ["a", "a_int"])]
    np.testing.assert_equal(expected[0].column(1), expected[0].column(1))


def test_cast(ctx, tmp_path):
    """Verify that we can cast"""
    path = helpers.write_parquet(tmp_path / "a.parquet", helpers.data())
    ctx.register_parquet("t", path)

    valid_types = [
        "smallint",
        "int",
        "bigint",
        "float(32)",
        "float(64)",
        "float",
    ]

    select = ", ".join([f"CAST(9 AS {t}) AS A{i}" for i, t in enumerate(valid_types)])

    # can execute, which implies that we can cast
    ctx.sql(f"SELECT {select} FROM t").collect()


@pytest.mark.parametrize(
    ("fn", "input_types", "output_type", "input_values", "expected_values"),
    [
        (
            lambda x: x,
            [pa.float64()],
            pa.float64(),
            [-1.2, None, 1.2],
            [-1.2, None, 1.2],
        ),
        (
            lambda x: x.is_null(),
            [pa.float64()],
            pa.bool_(),
            [-1.2, None, 1.2],
            [False, True, False],
        ),
    ],
)
def test_udf(
    ctx, tmp_path, fn, input_types, output_type, input_values, expected_values
):
    # write to disk
    path = helpers.write_parquet(tmp_path / "a.parquet", pa.array(input_values))
    ctx.register_parquet("t", path)

    func = udf(fn, input_types, output_type, name="func", volatility="immutable")
    ctx.register_udf(func)

    batches = ctx.sql("SELECT func(a) AS tt FROM t").collect()
    result = batches[0].column(0)

    assert result == pa.array(expected_values)


_null_mask = np.array([False, True, False])


@pytest.mark.parametrize(
    "arr",
    [
        pa.array(["a", "b", "c"], pa.utf8(), _null_mask),
        pa.array(["a", "b", "c"], pa.large_utf8(), _null_mask),
        pa.array([b"1", b"2", b"3"], pa.binary(), _null_mask),
        pa.array([b"1111", b"2222", b"3333"], pa.large_binary(), _null_mask),
        pa.array([False, True, True], None, _null_mask),
        pa.array([0, 1, 2], None),
        helpers.data_binary_other(),
        helpers.data_date32(),
        helpers.data_with_nans(),
        pytest.param(
            pa.array([b"1111", b"2222", b"3333"], pa.binary(4), _null_mask),
            id="binary4",
        ),
        pytest.param(
            helpers.data_datetime("s"),
            id="datetime_s",
        ),
        pytest.param(
            helpers.data_datetime("ms"),
            id="datetime_ms",
        ),
        pytest.param(
            helpers.data_datetime("us"),
            id="datetime_us",
        ),
        pytest.param(
            helpers.data_datetime("ns"),
            id="datetime_ns",
        ),
        # Not writtable to parquet
        pytest.param(helpers.data_timedelta("s"), id="timedelta_s"),
        pytest.param(helpers.data_timedelta("ms"), id="timedelta_ms"),
        pytest.param(helpers.data_timedelta("us"), id="timedelta_us"),
        pytest.param(helpers.data_timedelta("ns"), id="timedelta_ns"),
    ],
)
def test_simple_select(ctx, tmp_path, arr):
    path = helpers.write_parquet(tmp_path / "a.parquet", arr)
    ctx.register_parquet("t", path)

    batches = ctx.sql("SELECT a AS tt FROM t").collect()
    result = batches[0].column(0)

    # pyarrow.parquet promotes timestamp[s] to timestamp[ms] on write
    # (https://github.com/apache/arrow/issues/41382). Compensate so the
    # comparison checks DataFusion reads what Arrow actually stored.
    if (
        isinstance(arr, pa.Array)
        and pa.types.is_timestamp(arr.type)
        and arr.type.unit == "s"
    ):
        arr = arr.cast(pa.timestamp("ms"))

    # In DF 43.0.0 we now default to having BinaryView and StringView
    # so the array that is saved to the parquet is slightly different
    # than the array read. Convert to values for comparison.
    if isinstance(result, pa.BinaryViewArray | pa.StringViewArray):
        arr = arr.tolist()
        result = result.tolist()

    np.testing.assert_equal(result, arr)


@pytest.mark.parametrize(
    "file_sort_order", [None, [[col("int").sort(ascending=True, nulls_first=True)]]]
)
@pytest.mark.parametrize("pass_schema", [True, False])
@pytest.mark.parametrize("path_to_str", [True, False])
def test_register_listing_table(
    ctx, tmp_path, pass_schema, file_sort_order, path_to_str
):
    dir_root = tmp_path / "dataset_parquet_partitioned"
    dir_root.mkdir(exist_ok=False)
    (dir_root / "grp=a/date=2020-10-05").mkdir(exist_ok=False, parents=True)
    (dir_root / "grp=a/date=2021-10-05").mkdir(exist_ok=False, parents=True)
    (dir_root / "grp=b/date=2020-10-05").mkdir(exist_ok=False, parents=True)

    table = pa.Table.from_arrays(
        [
            [1, 2, 3, 4, 5, 6, 7],
            ["a", "b", "c", "d", "e", "f", "g"],
            [1.1, 2.2, 3.3, 4.4, 5.5, 6.6, 7.7],
        ],
        names=["int", "str", "float"],
    )
    pa.parquet.write_table(
        table.slice(0, 3), dir_root / "grp=a/date=2020-10-05/file.parquet"
    )
    pa.parquet.write_table(
        table.slice(3, 2), dir_root / "grp=a/date=2021-10-05/file.parquet"
    )
    pa.parquet.write_table(
        table.slice(5, 10), dir_root / "grp=b/date=2020-10-05/file.parquet"
    )

    dir_root = f"file://{dir_root}/" if path_to_str else dir_root

    ctx.register_listing_table(
        "my_table",
        dir_root,
        table_partition_cols=[("grp", pa.string()), ("date", pa.date64())],
        file_extension=".parquet",
        schema=table.schema if pass_schema else None,
        file_sort_order=file_sort_order,
    )
    assert ctx.catalog().schema().names() == {"my_table"}

    result = ctx.sql(
        "SELECT grp, COUNT(*) AS count FROM my_table GROUP BY grp"
    ).collect()
    result = pa.Table.from_batches(result)

    rd = result.to_pydict()
    assert dict(zip(rd["grp"], rd["count"], strict=False)) == {"a": 5, "b": 2}

    result = ctx.sql(
        "SELECT grp, COUNT(*) AS count FROM my_table WHERE date='2020-10-05' GROUP BY grp"  # noqa: E501
    ).collect()
    result = pa.Table.from_batches(result)

    rd = result.to_pydict()
    assert dict(zip(rd["grp"], rd["count"], strict=False)) == {"a": 3, "b": 2}


def test_parameterized_named_params(ctx, tmp_path) -> None:
    path = helpers.write_parquet(tmp_path / "a.parquet", helpers.data())

    df = ctx.read_parquet(path)
    result = ctx.sql(
        "SELECT COUNT(a) AS cnt, $lit_val as lit_val FROM $replaced_df",
        lit_val=3,
        replaced_df=df,
    ).collect()
    result = pa.Table.from_batches(result)
    assert result.to_pydict() == {"cnt": [100], "lit_val": [3]}


def test_parameterized_param_values(ctx: SessionContext) -> None:
    # Test the parameters that should be handled by the parser rather
    # than our manipulation of the query string by searching for tokens
    batch = pa.RecordBatch.from_arrays(
        [pa.array([1, 2, 3, 4])],
        names=["a"],
    )

    ctx.register_record_batches("t", [[batch]])
    result = ctx.sql("SELECT a FROM t WHERE a < $val", param_values={"val": 3})
    assert result.to_pydict() == {"a": [1, 2]}


def test_parameterized_mixed_query(ctx: SessionContext) -> None:
    batch = pa.RecordBatch.from_arrays(
        [pa.array([1, 2, 3, 4])],
        names=["a"],
    )
    ctx.register_record_batches("t", [[batch]])
    registered_df = ctx.table("t")

    result = ctx.sql(
        "SELECT $col_name FROM $df WHERE a < $val",
        param_values={"val": 3},
        df=registered_df,
        col_name="a",
    )
    assert result.to_pydict() == {"a": [1, 2]}