Skip to content

ARROW-10236: [Rust][DataFusion] Unify type casting logic in DataFusion #8400

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

Closed
alamb wants to merge 1 commit into from
Closed

ARROW-10236: [Rust][DataFusion] Unify type casting logic in DataFusion #8400

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
7 changes: 4 additions & 3 deletions rust/datafusion/src/logical_plan/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ use crate::{
use crate::{
physical_plan::{
aggregates, expressions::binary_operator_data_type, functions,
type_coercion::can_coerce_from, udf::ScalarUDF,
type_casting::can_cast_types, udf::ScalarUDF,
},
sql::parser::FileType,
};
Expand Down Expand Up @@ -323,12 +323,13 @@ impl Expr {
///
/// # Errors
///
/// This function errors when it is impossible to cast the expression to the target [arrow::datatypes::DataType].
/// This function errors when it is impossible to cast the
/// expression to the target [arrow::datatypes::DataType].
pub fn cast_to(&self, cast_to_type: &DataType, schema: &Schema) -> Result<Expr> {
let this_type = self.get_type(schema)?;
if this_type == *cast_to_type {
Ok(self.clone())
} else if can_coerce_from(cast_to_type, &this_type) {
} else if can_cast_types(&this_type, cast_to_type) {
Copy link
Contributor Author

@alamb alamb Oct 8, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

the idea is to use a common can_cast_types function to detect valid casts at plan time, and make can_cast_types consistent with the arrow cast kernel

Ok(Expr::Cast {
expr: Box::new(self.clone()),
data_type: cast_to_type.clone(),
Expand Down
29 changes: 14 additions & 15 deletions rust/datafusion/src/physical_plan/expressions.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,9 @@ use std::{cell::RefCell, convert::TryFrom};

use crate::error::{ExecutionError, Result};
use crate::logical_plan::Operator;
use crate::physical_plan::{Accumulator, AggregateExpr, PhysicalExpr};
use crate::physical_plan::{
type_casting::can_cast_types, Accumulator, AggregateExpr, PhysicalExpr,
};
use crate::scalar::ScalarValue;
use arrow::array::{
Float32Builder, Float64Builder, Int16Builder, Int32Builder, Int64Builder,
Expand Down Expand Up @@ -1532,27 +1534,23 @@ impl PhysicalExpr for CastExpr {
}
}

/// Returns a cast operation, if casting needed.
/// Returns a physical cast operation that casts `expr` to `cast_type`
/// if casting is needed.
///
/// Note that such casts may lose type information
pub fn cast(
expr: Arc<dyn PhysicalExpr>,
input_schema: &Schema,
cast_type: DataType,
) -> Result<Arc<dyn PhysicalExpr>> {
let expr_type = expr.data_type(input_schema)?;
if expr_type == cast_type {
return Ok(expr.clone());
}
if is_numeric(&expr_type) && (is_numeric(&cast_type) || cast_type == DataType::Utf8) {
Ok(Arc::new(CastExpr { expr, cast_type }))
} else if expr_type == DataType::Binary && cast_type == DataType::Utf8 {
Ok(Arc::new(CastExpr { expr, cast_type }))
} else if is_numeric(&expr_type)
&& cast_type == DataType::Timestamp(TimeUnit::Nanosecond, None)
{
Ok(expr.clone())
} else if can_cast_types(&expr_type, &cast_type) {
Ok(Arc::new(CastExpr { expr, cast_type }))
} else {
Err(ExecutionError::General(format!(
"Invalid CAST from {:?} to {:?}",
"Unsupported CAST from {:?} to {:?}",
expr_type, cast_type
)))
}
Expand Down Expand Up @@ -1992,9 +1990,10 @@ mod tests {

#[test]
fn invalid_cast() -> Result<()> {
let schema = Schema::new(vec![Field::new("a", DataType::Utf8, false)]);
let result = cast(col("a"), &schema, DataType::Int32);
result.expect_err("Invalid CAST from Utf8 to Int32");
// Ensure a useful error happens at plan time if invalid casts are used
let schema = Schema::new(vec![Field::new("a", DataType::Int32, false)]);
let result = cast(col("a"), &schema, DataType::LargeBinary);
result.expect_err("expected Invalid CAST");
Ok(())
}

Expand Down
1 change: 1 addition & 0 deletions rust/datafusion/src/physical_plan/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -207,6 +207,7 @@ pub mod planner;
pub mod projection;
pub mod sort;
pub mod string_expressions;
pub mod type_casting;
pub mod type_coercion;
pub mod udaf;
pub mod udf;
218 changes: 218 additions & 0 deletions rust/datafusion/src/physical_plan/type_casting.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,218 @@
// 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.

//! This module provides a way of checking what type casts are
//! supported at planning time for DataFusion. Since DataFusion uses
//! the Arrow `cast` compute kernel, the supported casts are the same
//! as the Arrow casts.
//!
//! The rules in this module are designed to be redundant with the
//! rules in the Arrow `cast` kernel. The redundancy is needed so that
//! DataFusion can generate an error at plan time rather than during
//! execution (which could happen many hours after execution starts,
//! when the query finally reaches that point)
//!

use arrow::datatypes::*;

/// Return true if a value of type `from_type` can be cast into a
/// value of `to_type`. Note that such as cast may be lossy. For
/// lossless type conversions, see the `type_coercion` module
///
/// See the module level documentation for more detail on casting
pub fn can_cast_types(from_type: &DataType, to_type: &DataType) -> bool {
use self::DataType::*;
if from_type == to_type {
return true;
}

// Note this is meant to mirror the structure in arrow/src/compute/kernels/cast.rs
match (from_type, to_type) {
Copy link
Member

@jorgecarleitao jorgecarleitao Oct 8, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is something that I have been thinking about: in DataFusion, because we use dynamically typed arrays, we often have to annotate which types are supported by each arrow kernel / operation.

Thus, we need to duplicate these matches to enumerate the valid types accepted by the kernels, so that we error during planning.

I wonder if these functions shouldn't be closer to the implementation (i.e. in the arrow::compute::kernels), or, alternatively, if we could find an idiom that would allow us to write these matches one time (again, in compute::kernel).

It seems to me that the pattern is:

for compute

match array.data_type() {
   ... => compute(array)
}

for datatypes:

match datatype {
    ... => return Some(return_datatype)
}

one idea would be to use

match datatype {
    ... => return Some((closure_that_computes, return_datatype))
}

so that both use-cases could be written in a single match (and reduce the risk of mis-typing / change in one place without a change in another place).

This comment is not specific to this PR, which I need to go through: I was just curious about your thoughts on this, since you have been touching in a couple of these recently.

Copy link
Contributor Author

@alamb alamb Oct 8, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@jorgecarleitao -- I think the idea of using a single match in cast.rs is a good one. I'll try and whip up a proposed PR with that pattern for discussion (though probably not until tomorrow morning US Eastern time, or the weekend)

Copy link
Contributor

@nevi-me nevi-me Oct 8, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I was also going to suggest if we move can_cast_types to Arrow

Copy link
Contributor Author

@alamb alamb Oct 13, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@jorgecarleitao and @nevi-me -- here is the the pattern I came up with:

pub fn cast(array: &ArrayRef, to_type: &DataType) -> Result<ArrayRef> {
    let from_type = array.data_type();
    let func = get_cast_function(from_type, to_type)?;
    func(array)
}

/// Returns true if the cast from `array.array` type is possible, false otherwise
pub fn can_cast(array: &ArrayRef, to_type: &DataType, ) -> bool {
    let from_type = array.data_type();
    get_cast_function(from_type, to_type).is_ok()
}


type CastFunction = Fn(&ArrayRef) -> Result<ArrayRef>;

/// Returns a function that can cast `array` to `to_type`.
///
/// All type checking for supported types should be done in
/// get_cast_func itself. Thus, if Ok(func) is returned, func(array)
/// should be able to succeed for arrays. In other words, Err(_)
/// should be returned by `get_cast_function`, NOT `func(array)` if
/// the types are incompatible.
///
/// can_cast relies on this functionality
fn get_cast_function(from_type: &DataType, to_type: &DataType) -> Result<CastFunction> {
    use DataType::*;

    // clone array if types are the same
    if from_type == to_type {
        return Ok(|array| { Ok(array.clone()) })
    }

    match (from_type, to_type) {
        (Struct(_), _) => Err(ArrowError::ComputeError(
            "Cannot cast from struct to other types".to_string(),
        )),
        (_, Struct(_)) => Err(ArrowError::ComputeError(
            "Cannot cast to struct from other types".to_string(),
        )),
        (List(_), List(ref to)) => Ok(|array| {
            let data = array.data_ref();
            let underlying_array = make_array(data.child_data()[0].clone());
            let cast_array = cast(&underlying_array, &to)?;
            let array_data = ArrayData::new(
                *to.clone(),
                array.len(),
                Some(cast_array.null_count()),
                cast_array
                    .data()
                    .null_bitmap()
                    .clone()
                    .map(|bitmap| bitmap.bits),
                array.offset(),
                // reuse offset buffer
                data.buffers().to_vec(),
                vec![cast_array.data()],
            );
            let list = ListArray::from(Arc::new(array_data));
            Ok(Arc::new(list) as ArrayRef)
        }),
        (List(_), _) => Err(ArrowError::ComputeError(
            "Cannot cast list to non-list data types".to_string(),
        )),
        (_, List(ref to)) => {
            // cast primitive to list's primitive
            let cast_func = get_cast_function(from_type, &to)?;
            Ok(move |array| {
                let cast_array = cast_func(array, &to)?;
                // create offsets, where if array.len() = 2, we have [0,1,2]
                let offsets: Vec<i32> = (0..=array.len() as i32).collect();
                let value_offsets = Buffer::from(offsets[..].to_byte_slice());
                let list_data = ArrayData::new(
                    *to.clone(),
                    array.len(),
                    Some(cast_array.null_count()),
                    cast_array
                        .data()
                        .null_bitmap()
                        .clone()
                        .map(|bitmap| bitmap.bits),
                    0,
                    vec![value_offsets],
                    vec![cast_array.data()],
                );
                let list_array = Arc::new(ListArray::from(Arc::new(list_data))) as ArrayRef;

                Ok(list_array)
            })
        }
...

While I think it would work, I am not sure I will have enough time.

Instead, I will get a PR up that moves the can_cast_types (with duplicate match) to arrow and then try a separate refactoring PR to unify the two match arms

(Struct(_), _) => false,
(_, Struct(_)) => false,
(List(list_from), List(list_to)) => can_cast_types(list_from, list_to),
(List(_), _) => false,
(_, List(list_to)) => can_cast_types(from_type, list_to),
(Dictionary(_, from_value_type), Dictionary(_, to_value_type)) => {
can_cast_types(from_value_type, to_value_type)
}
(Dictionary(_, value_type), _) => can_cast_types(value_type, to_type),
(_, Dictionary(_, value_type)) => can_cast_types(from_type, value_type),

(_, Boolean) => is_numeric_type(from_type),
(Boolean, _) => is_numeric_type(from_type) || from_type == &Utf8,
(Utf8, _) => is_numeric_type(to_type),
(_, Utf8) => is_numeric_type(from_type) || from_type == &Binary,

// start numeric casts
(UInt8, UInt16) => true,
(UInt8, UInt32) => true,
(UInt8, UInt64) => true,
(UInt8, Int8) => true,
(UInt8, Int16) => true,
(UInt8, Int32) => true,
(UInt8, Int64) => true,
(UInt8, Float32) => true,
(UInt8, Float64) => true,

(UInt16, UInt8) => true,
(UInt16, UInt32) => true,
(UInt16, UInt64) => true,
(UInt16, Int8) => true,
(UInt16, Int16) => true,
(UInt16, Int32) => true,
(UInt16, Int64) => true,
(UInt16, Float32) => true,
(UInt16, Float64) => true,

(UInt32, UInt8) => true,
(UInt32, UInt16) => true,
(UInt32, UInt64) => true,
(UInt32, Int8) => true,
(UInt32, Int16) => true,
(UInt32, Int32) => true,
(UInt32, Int64) => true,
(UInt32, Float32) => true,
(UInt32, Float64) => true,

(UInt64, UInt8) => true,
(UInt64, UInt16) => true,
(UInt64, UInt32) => true,
(UInt64, Int8) => true,
(UInt64, Int16) => true,
(UInt64, Int32) => true,
(UInt64, Int64) => true,
(UInt64, Float32) => true,
(UInt64, Float64) => true,

(Int8, UInt8) => true,
(Int8, UInt16) => true,
(Int8, UInt32) => true,
(Int8, UInt64) => true,
(Int8, Int16) => true,
(Int8, Int32) => true,
(Int8, Int64) => true,
(Int8, Float32) => true,
(Int8, Float64) => true,

(Int16, UInt8) => true,
(Int16, UInt16) => true,
(Int16, UInt32) => true,
(Int16, UInt64) => true,
(Int16, Int8) => true,
(Int16, Int32) => true,
(Int16, Int64) => true,
(Int16, Float32) => true,
(Int16, Float64) => true,

(Int32, UInt8) => true,
(Int32, UInt16) => true,
(Int32, UInt32) => true,
(Int32, UInt64) => true,
(Int32, Int8) => true,
(Int32, Int16) => true,
(Int32, Int64) => true,
(Int32, Float32) => true,
(Int32, Float64) => true,

(Int64, UInt8) => true,
(Int64, UInt16) => true,
(Int64, UInt32) => true,
(Int64, UInt64) => true,
(Int64, Int8) => true,
(Int64, Int16) => true,
(Int64, Int32) => true,
(Int64, Float32) => true,
(Int64, Float64) => true,

(Float32, UInt8) => true,
(Float32, UInt16) => true,
(Float32, UInt32) => true,
(Float32, UInt64) => true,
(Float32, Int8) => true,
(Float32, Int16) => true,
(Float32, Int32) => true,
(Float32, Int64) => true,
(Float32, Float64) => true,

(Float64, UInt8) => true,
(Float64, UInt16) => true,
(Float64, UInt32) => true,
(Float64, UInt64) => true,
(Float64, Int8) => true,
(Float64, Int16) => true,
(Float64, Int32) => true,
(Float64, Int64) => true,
(Float64, Float32) => true,
// end numeric casts

// temporal casts
(Int32, Date32(_)) => true,
(Int32, Time32(_)) => true,
(Date32(_), Int32) => true,
(Time32(_), Int32) => true,
(Int64, Date64(_)) => true,
(Int64, Time64(_)) => true,
(Date64(_), Int64) => true,
(Time64(_), Int64) => true,
(Date32(DateUnit::Day), Date64(DateUnit::Millisecond)) => true,
(Date64(DateUnit::Millisecond), Date32(DateUnit::Day)) => true,
(Time32(TimeUnit::Second), Time32(TimeUnit::Millisecond)) => true,
(Time32(TimeUnit::Millisecond), Time32(TimeUnit::Second)) => true,
(Time32(_), Time64(_)) => true,
(Time64(TimeUnit::Microsecond), Time64(TimeUnit::Nanosecond)) => true,
(Time64(TimeUnit::Nanosecond), Time64(TimeUnit::Microsecond)) => true,
(Time64(_), Time32(to_unit)) => match to_unit {
TimeUnit::Second => true,
TimeUnit::Millisecond => true,
_ => false,
},
(Timestamp(_, _), Int64) => true,
(Int64, Timestamp(_, _)) => true,
(Timestamp(_, _), Timestamp(_, _)) => true,
(Timestamp(_, _), Date32(_)) => true,
(Timestamp(_, _), Date64(_)) => true,
// date64 to timestamp might not make sense,

// end temporal casts
(_, _) => false,
}
}

fn is_numeric_type(t: &DataType) -> bool {
Copy link
Contributor

@nevi-me nevi-me Oct 8, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is also something that can live in perhaps arrow::util or arrow::datatype (and be implemented as a trait member of DataType).
I needed to check if a datatype was listy or stringy, and the boilerplate was tedious.

use self::DataType::*;
match t {
UInt8 | UInt16 | UInt32 | UInt64 | Int8 | Int16 | Int32 | Int64 | Float32
| Float64 => true,
_ => false,
}
}

#[cfg(test)]
mod tests {
// The purpose of this test is to verify that the rules of type
// casting between Arrow and DataFusion remain in sync.

// At a high level, each test attempts to cast the input arrays
// into the target type using the cast kernel and verifies the
// compatibility between `can_cast_from` and the cast kernel

#[test]
fn test_casting() {
//let arrays = vec![];
}
}