datafusion_functions_aggregate/
lib.rs

1// Licensed to the Apache Software Foundation (ASF) under one
2// or more contributor license agreements.  See the NOTICE file
3// distributed with this work for additional information
4// regarding copyright ownership.  The ASF licenses this file
5// to you under the Apache License, Version 2.0 (the
6// "License"); you may not use this file except in compliance
7// with the License.  You may obtain a copy of the License at
8//
9//   http://www.apache.org/licenses/LICENSE-2.0
10//
11// Unless required by applicable law or agreed to in writing,
12// software distributed under the License is distributed on an
13// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
14// KIND, either express or implied.  See the License for the
15// specific language governing permissions and limitations
16// under the License.
17
18#![doc(
19    html_logo_url = "https://raw.githubusercontent.com/apache/datafusion/19fe44cf2f30cbdd63d4a4f52c74055163c6cc38/docs/logos/standalone_logo/logo_original.svg",
20    html_favicon_url = "https://raw.githubusercontent.com/apache/datafusion/19fe44cf2f30cbdd63d4a4f52c74055163c6cc38/docs/logos/standalone_logo/logo_original.svg"
21)]
22#![cfg_attr(docsrs, feature(doc_auto_cfg))]
23// Make sure fast / cheap clones on Arc are explicit:
24// https://github.com/apache/datafusion/issues/11143
25#![deny(clippy::clone_on_ref_ptr)]
26
27//! Aggregate Function packages for [DataFusion].
28//!
29//! This crate contains a collection of various aggregate function packages for DataFusion,
30//! implemented using the extension API. Users may wish to control which functions
31//! are available to control the binary size of their application as well as
32//! use dialect specific implementations of functions (e.g. Spark vs Postgres)
33//!
34//! Each package is implemented as a separate
35//! module, activated by a feature flag.
36//!
37//! [DataFusion]: https://crates.io/crates/datafusion
38//!
39//! # Available Packages
40//! See the list of [modules](#modules) in this crate for available packages.
41//!
42//! # Using A Package
43//! You can register all functions in all packages using the [`register_all`] function.
44//!
45//! Each package also exports an `expr_fn` submodule to help create [`Expr`]s that invoke
46//! functions using a fluent style. For example:
47//!
48//![`Expr`]: datafusion_expr::Expr
49//!
50//! # Implementing A New Package
51//!
52//! To add a new package to this crate, you should follow the model of existing
53//! packages. The high level steps are:
54//!
55//! 1. Create a new module with the appropriate [AggregateUDF] implementations.
56//!
57//! 2. Use the macros in [`macros`] to create standard entry points.
58//!
59//! 3. Add a new feature to `Cargo.toml`, with any optional dependencies
60//!
61//! 4. Use the `make_package!` macro to expose the module when the
62//!    feature is enabled.
63
64#[macro_use]
65pub mod macros;
66
67pub mod approx_distinct;
68pub mod approx_median;
69pub mod approx_percentile_cont;
70pub mod approx_percentile_cont_with_weight;
71pub mod array_agg;
72pub mod average;
73pub mod bit_and_or_xor;
74pub mod bool_and_or;
75pub mod correlation;
76pub mod count;
77pub mod covariance;
78pub mod first_last;
79pub mod grouping;
80pub mod hyperloglog;
81pub mod median;
82pub mod min_max;
83pub mod nth_value;
84pub mod regr;
85pub mod stddev;
86pub mod string_agg;
87pub mod sum;
88pub mod variance;
89
90pub mod planner;
91
92use crate::approx_percentile_cont::approx_percentile_cont_udaf;
93use crate::approx_percentile_cont_with_weight::approx_percentile_cont_with_weight_udaf;
94use datafusion_common::Result;
95use datafusion_execution::FunctionRegistry;
96use datafusion_expr::AggregateUDF;
97use log::debug;
98use std::sync::Arc;
99
100/// Fluent-style API for creating `Expr`s
101pub mod expr_fn {
102    pub use super::approx_distinct::approx_distinct;
103    pub use super::approx_median::approx_median;
104    pub use super::approx_percentile_cont::approx_percentile_cont;
105    pub use super::approx_percentile_cont_with_weight::approx_percentile_cont_with_weight;
106    pub use super::array_agg::array_agg;
107    pub use super::average::avg;
108    pub use super::bit_and_or_xor::bit_and;
109    pub use super::bit_and_or_xor::bit_or;
110    pub use super::bit_and_or_xor::bit_xor;
111    pub use super::bool_and_or::bool_and;
112    pub use super::bool_and_or::bool_or;
113    pub use super::correlation::corr;
114    pub use super::count::count;
115    pub use super::count::count_distinct;
116    pub use super::covariance::covar_pop;
117    pub use super::covariance::covar_samp;
118    pub use super::first_last::first_value;
119    pub use super::first_last::last_value;
120    pub use super::grouping::grouping;
121    pub use super::median::median;
122    pub use super::min_max::max;
123    pub use super::min_max::min;
124    pub use super::nth_value::nth_value;
125    pub use super::regr::regr_avgx;
126    pub use super::regr::regr_avgy;
127    pub use super::regr::regr_count;
128    pub use super::regr::regr_intercept;
129    pub use super::regr::regr_r2;
130    pub use super::regr::regr_slope;
131    pub use super::regr::regr_sxx;
132    pub use super::regr::regr_sxy;
133    pub use super::regr::regr_syy;
134    pub use super::stddev::stddev;
135    pub use super::stddev::stddev_pop;
136    pub use super::sum::sum;
137    pub use super::variance::var_pop;
138    pub use super::variance::var_sample;
139}
140
141/// Returns all default aggregate functions
142pub fn all_default_aggregate_functions() -> Vec<Arc<AggregateUDF>> {
143    vec![
144        array_agg::array_agg_udaf(),
145        first_last::first_value_udaf(),
146        first_last::last_value_udaf(),
147        covariance::covar_samp_udaf(),
148        covariance::covar_pop_udaf(),
149        correlation::corr_udaf(),
150        sum::sum_udaf(),
151        min_max::max_udaf(),
152        min_max::min_udaf(),
153        median::median_udaf(),
154        count::count_udaf(),
155        regr::regr_slope_udaf(),
156        regr::regr_intercept_udaf(),
157        regr::regr_count_udaf(),
158        regr::regr_r2_udaf(),
159        regr::regr_avgx_udaf(),
160        regr::regr_avgy_udaf(),
161        regr::regr_sxx_udaf(),
162        regr::regr_syy_udaf(),
163        regr::regr_sxy_udaf(),
164        variance::var_samp_udaf(),
165        variance::var_pop_udaf(),
166        stddev::stddev_udaf(),
167        stddev::stddev_pop_udaf(),
168        approx_median::approx_median_udaf(),
169        approx_distinct::approx_distinct_udaf(),
170        approx_percentile_cont_udaf(),
171        approx_percentile_cont_with_weight_udaf(),
172        string_agg::string_agg_udaf(),
173        bit_and_or_xor::bit_and_udaf(),
174        bit_and_or_xor::bit_or_udaf(),
175        bit_and_or_xor::bit_xor_udaf(),
176        bool_and_or::bool_and_udaf(),
177        bool_and_or::bool_or_udaf(),
178        average::avg_udaf(),
179        grouping::grouping_udaf(),
180        nth_value::nth_value_udaf(),
181    ]
182}
183
184/// Registers all enabled packages with a [`FunctionRegistry`]
185pub fn register_all(registry: &mut dyn FunctionRegistry) -> Result<()> {
186    let functions: Vec<Arc<AggregateUDF>> = all_default_aggregate_functions();
187
188    functions.into_iter().try_for_each(|udf| {
189        let existing_udaf = registry.register_udaf(udf)?;
190        if let Some(existing_udaf) = existing_udaf {
191            debug!("Overwrite existing UDAF: {}", existing_udaf.name());
192        }
193        Ok(()) as Result<()>
194    })?;
195
196    Ok(())
197}
198
199#[cfg(test)]
200mod tests {
201    use crate::all_default_aggregate_functions;
202    use datafusion_common::Result;
203    use std::collections::HashSet;
204
205    #[test]
206    fn test_no_duplicate_name() -> Result<()> {
207        let mut names = HashSet::new();
208        let migrated_functions = ["array_agg", "count", "max", "min"];
209        for func in all_default_aggregate_functions() {
210            // TODO: remove this
211            // These functions are in intermediate migration state, skip them
212            if migrated_functions.contains(&func.name().to_lowercase().as_str()) {
213                continue;
214            }
215            assert!(
216                names.insert(func.name().to_string().to_lowercase()),
217                "duplicate function name: {}",
218                func.name()
219            );
220            for alias in func.aliases() {
221                assert!(
222                    names.insert(alias.to_string().to_lowercase()),
223                    "duplicate function name: {}",
224                    alias
225                );
226            }
227        }
228        Ok(())
229    }
230}
datafusion_functions_aggregate/lib.rs

datafusion_functions_aggregate/
lib.rs
