sunchao commented on code in PR #6599: URL: https://github.com/apache/datafusion-comet/pull/6599#discussion_r4185787351
########## native/spark-expr/src/array_funcs/array_contains.rs: ########## @@ -0,0 +1,446 @@ +// 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. + +use std::sync::Arc; + +use super::set_bits_before; +use crate::float_semantics::{compare_floats, spark_equality}; +use arrow::array::{Array, ArrayRef, AsArray, BooleanArray, BooleanBufferBuilder, ListArray}; +use arrow::buffer::{BooleanBuffer, NullBuffer}; +use arrow::datatypes::{ArrowPrimitiveType, DataType, Float32Type, Float64Type}; +use datafusion::common::{exec_err, utils::take_function_args, Result, ScalarValue}; +use datafusion::logical_expr::{ + ColumnarValue, ScalarFunctionArgs, ScalarUDFImpl, Signature, Volatility, +}; +use num::Float; + +/// Spark's `array_contains` for arrays whose elements hold floats at any depth. +/// +/// Spark tests each element against the value with `genEqual`, in which `-0.0` equals `0.0` and +/// all NaNs are equal, at any depth of an array or struct element. The result is true when an +/// element matches, null when none does and the array holds a null element, and false otherwise; +/// a null array or value gives null. datafusion-spark's `array_contains`, which Comet uses for +/// other element types, compares the bits instead, so `0.0` misses a `-0.0` and a NaN misses a +/// NaN whose bits differ. +#[derive(Debug, Hash, Eq, PartialEq)] +pub struct SparkFloatArrayContains { + signature: Signature, +} + +impl Default for SparkFloatArrayContains { + fn default() -> Self { + Self { + signature: Signature::any(2, Volatility::Immutable), + } + } +} + +impl ScalarUDFImpl for SparkFloatArrayContains { + fn name(&self) -> &str { + "spark_array_contains" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, _arg_types: &[DataType]) -> Result<DataType> { + Ok(DataType::Boolean) + } + + fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> { + let [array, value] = take_function_args(self.name(), &args.args)?; + let all_scalars = matches!( + (array, value), + (ColumnarValue::Scalar(_), ColumnarValue::Scalar(_)) + ); + let rows = if all_scalars { 1 } else { args.number_rows }; + let array = array.to_array(rows)?; + // Spark arrays use Arrow's 32-bit List layout. + let Some(list) = array.as_list_opt::<i32>() else { + return exec_err!( + "spark_array_contains takes a list, got {}", + array.data_type() + ); + }; + let result = match value { + ColumnarValue::Scalar(needle) => contains_constant(list, needle)?, + ColumnarValue::Array(value) => array_contains(list, value)?, + }; + if all_scalars { + Ok(ColumnarValue::Scalar(ScalarValue::Boolean( + result.is_valid(0).then(|| result.value(0)), + ))) + } else { + Ok(ColumnarValue::Array(Arc::new(result))) + } + } +} + +fn array_contains(array: &ListArray, value: &ArrayRef) -> Result<BooleanArray> { + // Spark casts the value to the element type. Anything else fails in `spark_equality`. + match (array.value_type(), value.data_type()) { + (DataType::Float32, DataType::Float32) => Ok(contains_floats::<Float32Type>(array, value)), + (DataType::Float64, DataType::Float64) => Ok(contains_floats::<Float64Type>(array, value)), + _ => { + let equal = spark_equality(array.values().as_ref(), value.as_ref())?; + contains_where(array, value, equal) + } + } +} + +/// [`array_contains`] with the same value for every row, which a float array answers in one pass +/// over all the values. +fn contains_constant(array: &ListArray, needle: &ScalarValue) -> Result<BooleanArray> { + match (array.value_type(), needle) { + // A null value gives null for every row. + (_, needle) if needle.is_null() => Ok(BooleanArray::new_null(array.len())), + (DataType::Float32, ScalarValue::Float32(Some(needle))) => { + Ok(contains_float_constant::<Float32Type>(array, *needle)) + } + (DataType::Float64, ScalarValue::Float64(Some(needle))) => { + Ok(contains_float_constant::<Float64Type>(array, *needle)) + } + _ => array_contains(array, &needle.to_array_of_size(array.len())?), + } +} + +fn contains_float_constant<T>(array: &ListArray, needle: T::Native) -> BooleanArray +where + T: ArrowPrimitiveType, + T::Native: Float, +{ + let offsets = array.offsets(); + let first = offsets[0] as usize; + let values = array + .values() + .slice(first, offsets[array.len()] as usize - first); + let floats = values.as_primitive::<T>(); + let buffer = floats.values(); + // For a fixed value, Spark's genEqual is one test per element: IEEE `==`, under which the two + // zeros are equal and a NaN differs from every number, or `is_nan` for a NaN value. + let matches = if needle.is_nan() { + BooleanBuffer::collect_bool(buffer.len(), |index| buffer[index].is_nan()) + } else { + BooleanBuffer::collect_bool(buffer.len(), |index| buffer[index] == needle) + }; + // A null element never matches. + let matches = match floats.nulls() { + Some(nulls) => &matches & nulls.inner(), + None => matches, + }; + // One pass over each bitmap gives every row's match count and valid count. + let matched = set_bits_before(&matches, offsets); + let valid = floats + .nulls() + .map(|nulls| set_bits_before(nulls.inner(), offsets)); + let mut values = BooleanBufferBuilder::new(array.len()); + let mut validity = BooleanBufferBuilder::new(array.len()); + for (row, bounds) in offsets.windows(2).enumerate() { + let found = matched[row + 1] > matched[row]; + let has_null = valid + .as_ref() + .is_some_and(|valid| valid[row + 1] - valid[row] < bounds[1] - bounds[0]); + values.append(found); + validity.append(array.is_valid(row) && (found || !has_null)); + } + BooleanArray::new(values.finish(), Some(NullBuffer::new(validity.finish()))) +} + +/// Tests each row's floats against the row's value in Spark's `genEqual`, straight from the +/// flattened values. A row is null when the array or the value is. +fn contains_floats<T>(array: &ListArray, value: &ArrayRef) -> BooleanArray +where + T: ArrowPrimitiveType, + T::Native: Float, +{ + let values = array.values().as_primitive::<T>(); + let needles = value.as_primitive::<T>().values(); + let (buffer, nulls) = (values.values(), values.nulls()); + let row_nulls = NullBuffer::union(array.nulls(), value.nulls()); + let mut found_rows = BooleanBufferBuilder::new(array.len()); + let mut validity = BooleanBufferBuilder::new(array.len()); + for (row, bounds) in array.offsets().windows(2).enumerate() { + let (mut found, mut has_null) = (false, false); + let row_valid = row_nulls.as_ref().is_none_or(|nulls| nulls.is_valid(row)); + if row_valid { + let needle = needles[row]; + for index in bounds[0] as usize..bounds[1] as usize { + if nulls.is_some_and(|nulls| nulls.is_null(index)) { + has_null = true; + } else if compare_floats(buffer[index], needle).is_eq() { + found = true; + break; + } + } + } + found_rows.append(found); + validity.append(row_valid && (found || !has_null)); + } + BooleanArray::new( + found_rows.finish(), + Some(NullBuffer::new(validity.finish())), + ) +} + +/// Tests each row's non-null elements with `equal(element, row)`. A row is null when the array or +/// the value is. +fn contains_where<F>(array: &ListArray, value: &ArrayRef, equal: F) -> Result<BooleanArray> +where + F: Fn(usize, usize) -> bool, +{ + let element_nulls = array.values().logical_nulls(); + let row_nulls = NullBuffer::union(array.nulls(), value.nulls()); + let mut found_rows = BooleanBufferBuilder::new(array.len()); + let mut validity = BooleanBufferBuilder::new(array.len()); + for (row, bounds) in array.offsets().windows(2).enumerate() { + let (mut found, mut has_null) = (false, false); + let row_valid = row_nulls.as_ref().is_none_or(|nulls| nulls.is_valid(row)); + if row_valid { + for element in bounds[0] as usize..bounds[1] as usize { + if element_nulls + .as_ref() + .is_some_and(|nulls| nulls.is_null(element)) + { + has_null = true; + } else if equal(element, row) { + found = true; + break; + } + } + } + found_rows.append(found); + validity.append(row_valid && (found || !has_null)); + } + Ok(BooleanArray::new( + found_rows.finish(), + Some(NullBuffer::new(validity.finish())), + )) +} + +#[cfg(test)] +mod tests { + use super::super::test_util::{list, EDGE_VALUES}; Review Comment: [P2] Import `EDGE_VALUES` from its current module. The merged base moved this constant to `crate::float_semantics`, but this new test module still imports it from `array_funcs::test_util`. Consequently, even a filtered Rust unit-test run fails during compilation with E0432, preventing the entire crate’s unit tests from running. Keep `list` imported from `test_util` and import `EDGE_VALUES` from `crate::float_semantics`, as `array_remove` and `sort_array` now do. Evidence: At the exact head, `cargo test --manifest-path native/Cargo.toml -p datafusion-comet-spark-expr --lib array_funcs::array_contains --locked --offline` exited 101: `error[E0432]: unresolved import super::super::test_util::EDGE_VALUES` at array_contains.rs:237. The base already defines `EDGE_VALUES` in native/spark-expr/src/float_semantics/mod.rs:140 and has no corresponding definition in array_funcs::test_util. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
