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.

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

Specialize single column primitive group values #7043

Merged
merged 8 commits into from
Jul 21, 2023
Merged
Show file tree
Hide file tree
Changes from 5 commits
Commits
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
1 change: 1 addition & 0 deletions datafusion/core/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -73,6 +73,7 @@ flate2 = { version = "1.0.24", optional = true }
futures = "0.3"
glob = "0.3.0"
hashbrown = { version = "0.14", features = ["raw"] }
half = { version = "2.1", default-features = false }
indexmap = "2.0.0"
itertools = "0.11"
lazy_static = { version = "^1.4.0" }
Expand Down
65 changes: 65 additions & 0 deletions datafusion/core/src/physical_plan/aggregates/group_values/mod.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,65 @@
// 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 arrow_array::{downcast_primitive, ArrayRef};
use arrow_schema::SchemaRef;
use datafusion_common::Result;
use datafusion_physical_expr::EmitTo;

mod primitive;
use primitive::GroupValuesPrimitive;

mod row;
use row::GroupValuesRows;

/// An interning store for group keys
pub trait GroupValues: Send {
/// Calculates the `groups` for each input row of `cols`
fn intern(&mut self, cols: &[ArrayRef], groups: &mut Vec<usize>) -> Result<()>;

/// Returns the number of bytes used by this [`GroupValues`]
fn size(&self) -> usize;

/// Returns true if this [`GroupValues`] is empty
fn is_empty(&self) -> bool;

/// The number of values stored in this [`GroupValues`]
fn len(&self) -> usize;

/// Emits the group values
fn emit(&mut self, emit_to: EmitTo) -> Result<Vec<ArrayRef>>;
}

pub fn new_group_values(schema: SchemaRef) -> Result<Box<dyn GroupValues>> {
if schema.fields.len() == 1 {
let d = schema.fields[0].data_type();

macro_rules! downcast_helper {
($t:ty, $d:ident) => {
return Ok(Box::new(GroupValuesPrimitive::<$t>::new($d.clone())))
};
}

// TODO: More primitives
Copy link
Contributor

Choose a reason for hiding this comment

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

Is this still relevant?

downcast_primitive! {
d => (downcast_helper, d),
_ => {}
}
}

Ok(Box::new(GroupValuesRows::try_new(schema)?))
}
191 changes: 191 additions & 0 deletions datafusion/core/src/physical_plan/aggregates/group_values/primitive.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,191 @@
// 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 crate::physical_plan::aggregates::group_values::GroupValues;
use ahash::RandomState;
use arrow::array::BooleanBufferBuilder;
use arrow::buffer::NullBuffer;
use arrow::datatypes::i256;
use arrow_array::cast::AsArray;
use arrow_array::{ArrayRef, ArrowNativeTypeOp, ArrowPrimitiveType, PrimitiveArray};
use arrow_schema::DataType;
use datafusion_common::Result;
use datafusion_execution::memory_pool::proxy::VecAllocExt;
use datafusion_physical_expr::EmitTo;
use half::f16;
use hashbrown::raw::RawTable;
use std::sync::Arc;

/// A trait to allow hashing of floating point numbers
Copy link
Contributor

Choose a reason for hiding this comment

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

Can you explain why this doesn't use create_hashes? And perhaps add comments in the code about the rationale?

If it is important not to use create_hashes I recommend

  1. Move this code to hash_utils.rs so it is easier to find
  2. Implement the force_hash_collisions version (that always hashes things to 0) to ensure appropriate coverage

Here is an example of force_hash_collisions
https://github.com/apache/arrow-datafusion/blob/368f6e606a3cfca8e04638b8d5ff0ff116a20b57/datafusion/physical-expr/src/hash_utils.rs#L214-L224

Copy link
Contributor Author

Choose a reason for hiding this comment

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

We can't use create_hashes as we are generating the hashes from the native values, not an array

trait HashValue {
fn hash(self, state: &RandomState) -> u64;
}

macro_rules! hash_integer {
($($t:ty),+) => {
$(impl HashValue for $t {
fn hash(self, state: &RandomState) -> u64 {
state.hash_one(self)
}
})+
};
}
hash_integer!(i8, i16, i32, i64, i128, i256);
hash_integer!(u8, u16, u32, u64);

macro_rules! hash_float {
($($t:ty),+) => {
$(impl HashValue for $t {
fn hash(self, state: &RandomState) -> u64 {
state.hash_one(self.to_bits())
}
})+
};
}

hash_float!(f16, f32, f64);

/// A [`GroupValues`] storing raw primitive values
Copy link
Contributor

Choose a reason for hiding this comment

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

Suggested change
/// A [`GroupValues`] storing raw primitive values
/// A [`GroupValues`] storing a single column of raw primitive values
///
/// This specialization is significantly faster than using the more general
/// purpose `Row`s format

pub struct GroupValuesPrimitive<T: ArrowPrimitiveType> {
/// The data type of the output array
data_type: DataType,
/// Stores the group index based on the hash of its value
map: RawTable<usize>,
alamb marked this conversation as resolved.
Show resolved Hide resolved
/// The group index of the null value if any
null_group: Option<usize>,
Copy link
Contributor

Choose a reason for hiding this comment

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

👍

/// The values for each group index
values: Vec<T::Native>,
/// The random state used to generate hashes
random_state: RandomState,
}

impl<T: ArrowPrimitiveType> GroupValuesPrimitive<T> {
pub fn new(data_type: DataType) -> Self {
assert!(PrimitiveArray::<T>::is_compatible(&data_type));
Self {
data_type,
map: RawTable::with_capacity(128),
values: Vec::with_capacity(128),
null_group: None,
random_state: Default::default(),
}
}
}

impl<T: ArrowPrimitiveType> GroupValues for GroupValuesPrimitive<T>
where
T::Native: HashValue,
{
fn intern(&mut self, cols: &[ArrayRef], groups: &mut Vec<usize>) -> Result<()> {
assert_eq!(cols.len(), 1);
groups.clear();

for v in cols[0].as_primitive::<T>() {
let group_id = match v {
None => *self.null_group.get_or_insert_with(|| {
let group_id = self.values.len();
self.values.push(Default::default());
group_id
}),
Some(key) => {
let state = &self.random_state;
let hash = key.hash(state);
let insert = self.map.find_or_find_insert_slot(
hash,
|g| unsafe { self.values.get_unchecked(*g).is_eq(key) },
Copy link
Contributor

Choose a reason for hiding this comment

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

this is awesome 🚀

|g| unsafe { self.values.get_unchecked(*g).hash(state) },
);

// SAFETY: No mutation occurred since find_or_find_insert_slot
unsafe {
match insert {
Ok(v) => *v.as_ref(),
Err(slot) => {
let g = self.values.len();
self.map.insert_in_slot(hash, slot, g);
Copy link
Contributor

Choose a reason for hiding this comment

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

I think this should still track the allocated memory (like insert_accounted did?)

Copy link
Contributor Author

Choose a reason for hiding this comment

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

It is accounted in the size method

self.values.push(key);
g
}
}
}
}
};
groups.push(group_id)
}
Ok(())
}

fn size(&self) -> usize {
self.map.capacity() * std::mem::size_of::<usize>() + self.values.allocated_size()
}

fn is_empty(&self) -> bool {
self.values.is_empty()
}

fn len(&self) -> usize {
self.values.len()
}

fn emit(&mut self, emit_to: EmitTo) -> Result<Vec<ArrayRef>> {
fn build_primitive<T: ArrowPrimitiveType>(
values: Vec<T::Native>,
null_idx: Option<usize>,
) -> PrimitiveArray<T> {
let nulls = null_idx.map(|null_idx| {
let mut buffer = BooleanBufferBuilder::new(values.len());
buffer.append_n(values.len(), true);
buffer.set_bit(null_idx, false);
unsafe { NullBuffer::new_unchecked(buffer.finish(), 1) }
});
PrimitiveArray::<T>::new(values.into(), nulls)
}

let array: PrimitiveArray<T> = match emit_to {
EmitTo::All => {
self.map.clear();
build_primitive(std::mem::take(&mut self.values), self.null_group.take())
}
EmitTo::First(n) => {
// SAFETY: self.map outlives iterator and is not modified concurrently
unsafe {
Copy link
Contributor

Choose a reason for hiding this comment

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

I think this code is largely replicated from the row version. I wonder if it could be refactored into a (templated) common function (with appropriate documentation)?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

There isn't an easy way to make this generic, as one stores tuples and one isn't... I at least can't see a way that doesn't just obfuscate the code

for bucket in self.map.iter() {
// Decrement group index by n
match bucket.as_ref().checked_sub(n) {
// Group index was >= n, shift value down
Some(sub) => *bucket.as_mut() = sub,
// Group index was < n, so remove from table
None => self.map.erase(bucket),
}
}
}
let null_group = match &mut self.null_group {
Some(v) if *v >= n => {
*v -= n;
None
}
Some(_) => self.null_group.take(),
None => None,
};
let mut split = self.values.split_off(n);
std::mem::swap(&mut self.values, &mut split);
build_primitive(split, null_group)
}
};
Ok(vec![Arc::new(array.with_data_type(self.data_type.clone()))])
}
}
Loading