Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

feat: Additional uint datatype support for the SQL interface #15993

Merged
merged 1 commit into from
May 2, 2024
Merged

feat: Additional uint datatype support for the SQL interface #15993

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
137 changes: 94 additions & 43 deletions crates/polars-sql/src/sql_expr.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,8 +13,8 @@ use sqlparser::ast::ExactNumberInfo;
use sqlparser::ast::{
ArrayAgg, ArrayElemTypeDef, BinaryOperator as SQLBinaryOperator, BinaryOperator, CastFormat,
DataType as SQLDataType, DateTimeField, Expr as SQLExpr, Function as SQLFunction, Ident,
JoinConstraint, OrderByExpr, Query as Subquery, SelectItem, TimezoneInfo, TrimWhereField,
UnaryOperator, Value as SQLValue,
JoinConstraint, ObjectName, OrderByExpr, Query as Subquery, SelectItem, TimezoneInfo,
TrimWhereField, UnaryOperator, Value as SQLValue,
};
use sqlparser::dialect::GenericDialect;
use sqlparser::parser::{Parser, ParserOptions};
Expand All @@ -24,41 +24,53 @@ use crate::SQLContext;

pub(crate) fn map_sql_polars_datatype(data_type: &SQLDataType) -> PolarsResult<DataType> {
Ok(match data_type {
// ---------------------------------
// array/list
// ---------------------------------
SQLDataType::Array(ArrayElemTypeDef::AngleBracket(inner_type))
| SQLDataType::Array(ArrayElemTypeDef::SquareBracket(inner_type)) => {
DataType::List(Box::new(map_sql_polars_datatype(inner_type)?))
},
#[cfg(feature = "dtype-decimal")]
SQLDataType::Dec(info) | SQLDataType::Decimal(info) | SQLDataType::Numeric(info) => {
match *info {
ExactNumberInfo::PrecisionAndScale(p, s) => {
DataType::Decimal(Some(p as usize), Some(s as usize))
},
ExactNumberInfo::Precision(p) => DataType::Decimal(Some(p as usize), Some(0)),
ExactNumberInfo::None => DataType::Decimal(Some(38), Some(9)),
}
},
SQLDataType::BigInt(_) => DataType::Int64,
SQLDataType::Boolean => DataType::Boolean,

// ---------------------------------
// binary
// ---------------------------------
SQLDataType::Bytea
| SQLDataType::Bytes(_)
| SQLDataType::Binary(_)
| SQLDataType::Blob(_)
| SQLDataType::Varbinary(_) => DataType::Binary,
SQLDataType::Char(_)
| SQLDataType::CharVarying(_)
| SQLDataType::Character(_)
| SQLDataType::CharacterVarying(_)
| SQLDataType::Clob(_)
| SQLDataType::String(_)
| SQLDataType::Text
| SQLDataType::Uuid
| SQLDataType::Varchar(_) => DataType::String,
SQLDataType::Date => DataType::Date,
SQLDataType::Double
| SQLDataType::DoublePrecision
| SQLDataType::Float8
| SQLDataType::Float64 => DataType::Float64,

// ---------------------------------
// boolean
// ---------------------------------
SQLDataType::Boolean | SQLDataType::Bool => DataType::Boolean,

// ---------------------------------
// signed integer
// ---------------------------------
SQLDataType::Int(_) | SQLDataType::Integer(_) => DataType::Int32,
SQLDataType::Int2(_) | SQLDataType::SmallInt(_) => DataType::Int16,
SQLDataType::Int4(_) | SQLDataType::MediumInt(_) => DataType::Int32,
SQLDataType::Int8(_) | SQLDataType::BigInt(_) => DataType::Int64,
SQLDataType::TinyInt(_) => DataType::Int8,

// ---------------------------------
// unsigned integer: the following do not map to PostgreSQL types/syntax, but
// are enabled for wider compatibility (eg: "CAST(col AS BIGINT UNSIGNED)").
// ---------------------------------
SQLDataType::UnsignedInt(_) | SQLDataType::UnsignedInteger(_) => DataType::UInt32,
SQLDataType::UnsignedInt2(_) | SQLDataType::UnsignedSmallInt(_) => DataType::UInt16,
SQLDataType::UnsignedInt4(_) | SQLDataType::UnsignedMediumInt(_) => DataType::UInt32,
SQLDataType::UnsignedInt8(_) | SQLDataType::UnsignedBigInt(_) => DataType::UInt64,
SQLDataType::UnsignedTinyInt(_) => DataType::UInt8, // see also: "custom" types below

// ---------------------------------
// float
// ---------------------------------
SQLDataType::Double | SQLDataType::DoublePrecision | SQLDataType::Float8 => {
DataType::Float64
},
SQLDataType::Float(n_bytes) => match n_bytes {
Some(n) if (1u64..=24u64).contains(n) => DataType::Float32,
Some(n) if (25u64..=53u64).contains(n) => DataType::Float64,
Expand All @@ -68,12 +80,26 @@ pub(crate) fn map_sql_polars_datatype(data_type: &SQLDataType) -> PolarsResult<D
None => DataType::Float64,
},
SQLDataType::Float4 | SQLDataType::Real => DataType::Float32,
SQLDataType::Int(_) | SQLDataType::Integer(_) => DataType::Int32,
SQLDataType::Int2(_) => DataType::Int16,
SQLDataType::Int4(_) => DataType::Int32,
SQLDataType::Int8(_) => DataType::Int64,

// ---------------------------------
// decimal
// ---------------------------------
#[cfg(feature = "dtype-decimal")]
SQLDataType::Dec(info) | SQLDataType::Decimal(info) | SQLDataType::Numeric(info) => {
match *info {
ExactNumberInfo::PrecisionAndScale(p, s) => {
DataType::Decimal(Some(p as usize), Some(s as usize))
},
ExactNumberInfo::Precision(p) => DataType::Decimal(Some(p as usize), Some(0)),
ExactNumberInfo::None => DataType::Decimal(Some(38), Some(9)),
}
},

// ---------------------------------
// temporal
// ---------------------------------
SQLDataType::Date => DataType::Date,
SQLDataType::Interval => DataType::Duration(TimeUnit::Microseconds),
SQLDataType::SmallInt(_) => DataType::Int16,
SQLDataType::Time(_, tz) => match tz {
TimezoneInfo::None => DataType::Time,
_ => {
Expand All @@ -97,16 +123,41 @@ pub(crate) fn map_sql_polars_datatype(data_type: &SQLDataType) -> PolarsResult<D
},
}
},
SQLDataType::TinyInt(_) => DataType::Int8,
SQLDataType::UnsignedBigInt(_) => DataType::UInt64,
SQLDataType::UnsignedInt(_) | SQLDataType::UnsignedInteger(_) => DataType::UInt32,
SQLDataType::UnsignedInt2(_) => DataType::UInt16,
SQLDataType::UnsignedInt4(_) => DataType::UInt32,
SQLDataType::UnsignedInt8(_) => DataType::UInt64,
SQLDataType::UnsignedSmallInt(_) => DataType::UInt16,
SQLDataType::UnsignedTinyInt(_) => DataType::UInt8,

_ => polars_bail!(ComputeError: "SQL datatype {:?} is not yet supported", data_type),
// ---------------------------------
// string
// ---------------------------------
SQLDataType::Char(_)
| SQLDataType::CharVarying(_)
| SQLDataType::Character(_)
| SQLDataType::CharacterVarying(_)
| SQLDataType::Clob(_)
| SQLDataType::String(_)
| SQLDataType::Text
| SQLDataType::Uuid
| SQLDataType::Varchar(_) => DataType::String,

// ---------------------------------
// custom
// ---------------------------------
SQLDataType::Custom(ObjectName(idents), _) => match idents.as_slice() {
[Ident { value, .. }] => match value.to_lowercase().as_str() {
// these integer types are not supported by the PostgreSQL core distribution,
// but they ARE available via `pguint` (https://github.com/petere/pguint), an
// extension maintained by one of the PostgreSQL core developers.
"uint1" => DataType::UInt8,
"uint2" => DataType::UInt16,
"uint4" | "uint" => DataType::UInt32,
"uint8" => DataType::UInt64,
// `pguint` also provides a 1 byte (8bit) integer type alias
"int1" => DataType::Int8,
_ => {
polars_bail!(ComputeError: "SQL datatype {:?} is not currently supported", value)
},
},
_ => polars_bail!(ComputeError: "SQL datatype {:?} is not currently supported", idents),
},
_ => polars_bail!(ComputeError: "SQL datatype {:?} is not currently supported", data_type),
})
}

Expand Down Expand Up @@ -500,7 +551,7 @@ impl SQLExprVisitor<'_> {
return Ok(expr.str().json_decode(None, None));
}
let polars_type = map_sql_polars_datatype(data_type)?;
Ok(expr.cast(polars_type))
Ok(expr.strict_cast(polars_type))
}

/// Visit a SQL literal.
Expand Down
109 changes: 103 additions & 6 deletions py-polars/tests/unit/sql/test_cast.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,8 +1,11 @@
from __future__ import annotations

from typing import Any

import pytest

import polars as pl
import polars.selectors as cs
from polars.exceptions import ComputeError
from polars.testing import assert_frame_equal

Expand All @@ -14,6 +17,7 @@ def test_cast() -> None:
"b": [1.1, 2.2, 3.3, 4.4, 5.5],
"c": ["a", "b", "c", "d", "e"],
"d": [True, False, True, False, True],
"e": [-1, 0, None, 1, 2],
}
)
# test various dtype casts, using standard ("CAST <col> AS <dtype>")
Expand All @@ -25,41 +29,116 @@ def test_cast() -> None:
-- float
CAST(a AS DOUBLE PRECISION) AS a_f64,
a::real AS a_f32,
b::float(24) AS b_f32,
b::float(25) AS b_f64,
e::float8 AS e_f64,
e::float4 AS e_f32,

-- integer
CAST(b AS TINYINT) AS b_i8,
CAST(b AS SMALLINT) AS b_i16,
b::bigint AS b_i64,
d::tinyint AS d_i8,
a::int1 AS a_i8,
a::int2 AS a_i16,
a::int4 AS a_i32,
a::int8 AS a_i64,

-- unsigned integer
CAST(a AS TINYINT UNSIGNED) AS a_u8,
d::uint1 AS d_u8,
a::uint2 AS a_u16,
b::uint4 AS b_u32,
b::uint8 AS b_u64,
CAST(a AS BIGINT UNSIGNED) AS a_u64,

-- string/binary
CAST(a AS CHAR) AS a_char,
CAST(b AS VARCHAR) AS b_varchar,
c::blob AS c_blob,
c::bytes AS c_bytes,
c::VARBINARY AS c_varbinary,
CAST(d AS CHARACTER VARYING) AS d_charvar,

-- boolean
e::bool AS e_bool,
e::boolean AS e_boolean
FROM df
"""
)
assert res.schema == {
"a_f64": pl.Float64,
"a_f32": pl.Float32,
"b_f32": pl.Float32,
"b_f64": pl.Float64,
"e_f64": pl.Float64,
"e_f32": pl.Float32,
"b_i8": pl.Int8,
"b_i16": pl.Int16,
"b_i64": pl.Int64,
"d_i8": pl.Int8,
"a_i8": pl.Int8,
"a_i16": pl.Int16,
"a_i32": pl.Int32,
"a_i64": pl.Int64,
"a_u8": pl.UInt8,
"d_u8": pl.UInt8,
"a_u16": pl.UInt16,
"b_u32": pl.UInt32,
"b_u64": pl.UInt64,
"a_u64": pl.UInt64,
"a_char": pl.String,
"b_varchar": pl.String,
"c_blob": pl.Binary,
"c_bytes": pl.Binary,
"c_varbinary": pl.Binary,
"d_charvar": pl.String,
"e_bool": pl.Boolean,
"e_boolean": pl.Boolean,
}
assert res.rows() == [
(1.0, 1.0, 1, 1, 1, 1, "1", "1.1", b"a", b"a", b"a", "true"),
(2.0, 2.0, 2, 2, 2, 0, "2", "2.2", b"b", b"b", b"b", "false"),
(3.0, 3.0, 3, 3, 3, 1, "3", "3.3", b"c", b"c", b"c", "true"),
(4.0, 4.0, 4, 4, 4, 0, "4", "4.4", b"d", b"d", b"d", "false"),
(5.0, 5.0, 5, 5, 5, 1, "5", "5.5", b"e", b"e", b"e", "true"),
assert res.select(cs.by_dtype(pl.Float32)).rows() == pytest.approx(
[
(1.0, 1.100000023841858, -1.0),
(2.0, 2.200000047683716, 0.0),
(3.0, 3.299999952316284, None),
(4.0, 4.400000095367432, 1.0),
(5.0, 5.5, 2.0),
]
)
assert res.select(cs.by_dtype(pl.Float64)).rows() == [
(1.0, 1.1, -1.0),
(2.0, 2.2, 0.0),
(3.0, 3.3, None),
(4.0, 4.4, 1.0),
(5.0, 5.5, 2.0),
]
assert res.select(cs.integer()).rows() == [
(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1),
(2, 2, 2, 0, 2, 2, 2, 2, 2, 0, 2, 2, 2, 2),
(3, 3, 3, 1, 3, 3, 3, 3, 3, 1, 3, 3, 3, 3),
(4, 4, 4, 0, 4, 4, 4, 4, 4, 0, 4, 4, 4, 4),
(5, 5, 5, 1, 5, 5, 5, 5, 5, 1, 5, 5, 5, 5),
]
assert res.select(cs.string()).rows() == [
("1", "1.1", "true"),
("2", "2.2", "false"),
("3", "3.3", "true"),
("4", "4.4", "false"),
("5", "5.5", "true"),
]
assert res.select(cs.binary()).rows() == [
(b"a", b"a", b"a"),
(b"b", b"b", b"b"),
(b"c", b"c", b"c"),
(b"d", b"d", b"d"),
(b"e", b"e", b"e"),
]
assert res.select(cs.boolean()).rows() == [
(True, True),
(False, False),
(None, None),
(True, True),
(True, True),
]

with pytest.raises(ComputeError, match="unsupported use of FORMAT in CAST"):
Expand All @@ -68,6 +147,24 @@ def test_cast() -> None:
)


@pytest.mark.parametrize(
("values", "cast_op", "error"),
[
([1.0, -1.0], "values::uint8", "conversion from `f64` to `u64` failed"),
([10, 0, -1], "values::uint4", "conversion from `i64` to `u32` failed"),
([int(1e8)], "values::int1", "conversion from `i64` to `i8` failed"),
(["a", "b"], "values::date", "conversion from `str` to `date` failed"),
(["a", "b"], "values::time", "conversion from `str` to `time` failed"),
(["a", "b"], "values::int4", "conversion from `str` to `i32` failed"),
],
)
def test_cast_errors(values: Any, cast_op: str, error: str) -> None:
df = pl.DataFrame({"values": values})

with pytest.raises(ComputeError, match=error):
df.sql(f"SELECT {cast_op} FROM df")


def test_cast_json() -> None:
df = pl.DataFrame({"txt": ['{"a":[1,2,3],"b":["x","y","z"],"c":5.0}']})

Expand Down
Loading