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

Add example for FunctionFactory #9482

Merged
merged 7 commits into from
Mar 10, 2024
Merged
Show file tree
Hide file tree
Changes from all 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-examples/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,7 @@ cargo run --example csv_sql
- [`deserialize_to_struct.rs`](examples/deserialize_to_struct.rs): Convert query results into rust structs using serde
- [`expr_api.rs`](examples/expr_api.rs): Create, execute, simplify and analyze `Expr`s
- [`flight_sql_server.rs`](examples/flight/flight_sql_server.rs): Run DataFusion as a standalone process and execute SQL queries from JDBC clients
- [`function_factory.rs`](examples/function_factory.rs): Register `CREATE FUNCTION` handler to implement SQL macros
Copy link
Contributor

Choose a reason for hiding this comment

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

I added the example

- [`make_date.rs`](examples/make_date.rs): Examples of using the make_date function
- [`memtable.rs`](examples/memtable.rs): Create an query data in memory using SQL and `RecordBatch`es
- [`pruning.rs`](examples/parquet_sql.rs): Use pruning to rule out files based on statistics
Expand Down
232 changes: 232 additions & 0 deletions datafusion-examples/examples/function_factory.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,232 @@
// 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 datafusion::error::Result;
use datafusion::execution::config::SessionConfig;
use datafusion::execution::context::{FunctionFactory, RegisterFunction, SessionContext};
use datafusion_common::tree_node::{Transformed, TreeNode};
use datafusion_common::{exec_err, internal_err, DataFusionError};
use datafusion_expr::simplify::ExprSimplifyResult;
use datafusion_expr::simplify::SimplifyInfo;
use datafusion_expr::{CreateFunction, Expr, ScalarUDF, ScalarUDFImpl, Signature};
use std::result::Result as RResult;
use std::sync::Arc;

/// This example shows how to utilize [FunctionFactory] to implement simple
/// SQL-macro like functions using a `CREATE FUNCTION` statement. The same
/// functionality can support functions defined in any language or library.
///
/// Apart from [FunctionFactory], this example covers
/// [ScalarUDFImpl::simplify()] which is often used at the same time, to replace
/// a function call with another expression at rutime.
///
/// This example is rather simple and does not cover all cases required for a
/// real implementation.
#[tokio::main]
async fn main() -> Result<()> {
// First we must configure the SessionContext with our function factory
let ctx = SessionContext::new()
// register custom function factory
.with_function_factory(Arc::new(CustomFunctionFactory::default()));

// With the function factory, we can now call `CREATE FUNCTION` SQL functions

// Let us register a function called f which takes a single argument and
// returns that value plus one
let sql = r#"
CREATE FUNCTION f1(BIGINT)
RETURNS BIGINT
RETURN $1 + 1
"#;

ctx.sql(sql).await?.show().await?;

// Now, let us register a function called f2 which takes two arguments and
// returns the first argument added to the result of calling f1 on that
// argument
let sql = r#"
CREATE FUNCTION f2(BIGINT, BIGINT)
RETURNS BIGINT
RETURN $1 + f1($2)
"#;

ctx.sql(sql).await?.show().await?;

// Invoke f2, and we expect to see 1 + (1 + 2) = 4
Copy link
Contributor

Choose a reason for hiding this comment

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

I removed some of the other things that didn't really help teach people how to use this feature any more (they were more like ensuring all features got covered) to keep it shorter

// Note this function works on columns as well as constants.
let sql = r#"
SELECT f2(1, 2)
"#;
ctx.sql(sql).await?.show().await?;

// Now we clean up the session by dropping the functions
ctx.sql("DROP FUNCTION f1").await?.show().await?;
ctx.sql("DROP FUNCTION f2").await?.show().await?;

Ok(())
}

/// This is our FunctionFactory that is responsible for converting `CREATE
/// FUNCTION` statements into function instances
#[derive(Debug, Default)]
struct CustomFunctionFactory {}

#[async_trait::async_trait]
impl FunctionFactory for CustomFunctionFactory {
/// This function takes the parsed `CREATE FUNCTION` statement and returns
/// the function instance.
async fn create(
&self,
_state: &SessionConfig,
statement: CreateFunction,
) -> Result<RegisterFunction> {
let f: ScalarFunctionWrapper = statement.try_into()?;

Ok(RegisterFunction::Scalar(Arc::new(ScalarUDF::from(f))))
}
}

/// this function represents the newly created execution engine.
#[derive(Debug)]
struct ScalarFunctionWrapper {
/// The text of the function body, `$1 + f1($2)` in our example
name: String,
expr: Expr,
signature: Signature,
return_type: arrow_schema::DataType,
}

impl ScalarUDFImpl for ScalarFunctionWrapper {
fn as_any(&self) -> &dyn std::any::Any {
self
}

fn name(&self) -> &str {
&self.name
}

fn signature(&self) -> &datafusion_expr::Signature {
&self.signature
}

fn return_type(
&self,
_arg_types: &[arrow_schema::DataType],
) -> Result<arrow_schema::DataType> {
Ok(self.return_type.clone())
}

fn invoke(
&self,
_args: &[datafusion_expr::ColumnarValue],
) -> Result<datafusion_expr::ColumnarValue> {
// Since this function is always simplified to another expression, it
// should never actually be invoked
internal_err!("This function should not get invoked!")
}

/// The simplify function is called to simply a call such as `f2(2)`. This
Copy link
Contributor

Choose a reason for hiding this comment

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

I tried to add some comments inline that explained what the functions were doing with more words, but hopefully for someone who was seeing if this worked for their usecase

/// function parses the string and returns the resulting expression
fn simplify(
&self,
args: Vec<Expr>,
_info: &dyn SimplifyInfo,
) -> Result<ExprSimplifyResult> {
let replacement = Self::replacement(&self.expr, &args)?;

Ok(ExprSimplifyResult::Simplified(replacement))
}

fn aliases(&self) -> &[String] {
&[]
}

fn monotonicity(&self) -> Result<Option<datafusion_expr::FuncMonotonicity>> {
Ok(None)
}
}

impl ScalarFunctionWrapper {
// replaces placeholders such as $1 with actual arguments (args[0]
fn replacement(expr: &Expr, args: &[Expr]) -> Result<Expr> {
let result = expr.clone().transform(&|e| {
let r = match e {
Expr::Placeholder(placeholder) => {
let placeholder_position =
Self::parse_placeholder_identifier(&placeholder.id)?;
if placeholder_position < args.len() {
Transformed::yes(args[placeholder_position].clone())
} else {
exec_err!(
"Function argument {} not provided, argument missing!",
placeholder.id
)?
}
}
_ => Transformed::no(e),
};

Ok(r)
})?;

Ok(result.data)
}
// Finds placeholder identifier such as `$X` format where X >= 1
fn parse_placeholder_identifier(placeholder: &str) -> Result<usize> {
if let Some(value) = placeholder.strip_prefix('$') {
Ok(value.parse().map(|v: usize| v - 1).map_err(|e| {
DataFusionError::Execution(format!(
"Placeholder `{}` parsing error: {}!",
placeholder, e
))
})?)
} else {
exec_err!("Placeholder should start with `$`!")
}
}
}

/// This impl block creates a scalar function from
/// a parsed `CREATE FUNCTION` statement (`CreateFunction`)
impl TryFrom<CreateFunction> for ScalarFunctionWrapper {
type Error = DataFusionError;

fn try_from(definition: CreateFunction) -> RResult<Self, Self::Error> {
Ok(Self {
name: definition.name,
expr: definition
.params
.return_
.expect("Expression has to be defined!"),
return_type: definition
.return_type
.expect("Return type has to be defined!"),
signature: Signature::exact(
definition
.args
.unwrap_or_default()
.into_iter()
.map(|a| a.data_type)
.collect(),
definition
.params
.behavior
.unwrap_or(datafusion_expr::Volatility::Volatile),
),
})
}
}
14 changes: 14 additions & 0 deletions datafusion/core/src/execution/context/mod.rs
Original file line number Diff line number Diff line change
Expand Up @@ -349,6 +349,15 @@ impl SessionContext {
self.session_start_time
}

/// Registers a [`FunctionFactory`] to handle `CREATE FUNCTION` statements
pub fn with_function_factory(
self,
function_factory: Arc<dyn FunctionFactory>,
) -> Self {
self.state.write().set_function_factory(function_factory);
self
}

/// Registers the [`RecordBatch`] as the specified table name
pub fn register_batch(
&self,
Expand Down Expand Up @@ -1659,6 +1668,11 @@ impl SessionState {
self
}

/// Registers a [`FunctionFactory`] to handle `CREATE FUNCTION` statements
pub fn set_function_factory(&mut self, function_factory: Arc<dyn FunctionFactory>) {
self.function_factory = Some(function_factory);
}

/// Replace the extension [`SerializerRegistry`]
pub fn with_serializer_registry(
mut self,
Expand Down
Loading
Loading