Add plugable handler for CREATE FUNCTION (#9333)

* Add plugable function factory

* cover `DROP FUNCTION` as well ...

... partially, as `SessionState` does not expose
unregister_udf at the moment.

* update documentation

* fix doc test

* Address PR comments (code organization)

* Address PR comments (factory interface)

* fix test after rebase

* `remove`'s gone from the trait ...

... `DROP FUNCTION` will look for function name
in all available registries (udf, udaf, udwf).

`remove` may be necessary if UDaF and UDwF do not
get `simplify` method from #9304.

* Rename FunctionDefinition and export it ...

FunctionDefinition already exists, DefinitionStatement makes more sense.

* Update datafusion/expr/src/logical_plan/ddl.rs

Co-authored-by: Andrew Lamb <[email protected]>

* Update datafusion/core/src/execution/context/mod.rs

Co-authored-by: Andrew Lamb <[email protected]>

* Update datafusion/core/tests/user_defined/user_defined_scalar_functions.rs

Co-authored-by: Andrew Lamb <[email protected]>

* Update datafusion/expr/src/logical_plan/ddl.rs

Co-authored-by: Andrew Lamb <[email protected]>

* resolve part of follow up comments

* Qualified functions are not supported anymore

* update docs and todos

* fix clippy

* address additional comments

* Add sqllogicteset for CREATE/DROP function

* Add coverage for DROP FUNCTION IF EXISTS

* fix multiline error

* revert dialect back to generic in test ...

... as `create function` gets support in latest
sqlparser.

* fmt

---------

Co-authored-by: Andrew Lamb <[email protected]>

Author: milenkovicm

datafusion/core/src/execution/context/mod.rs

@@ -73,9 +73,10 @@ use crate::datasource::{
 };
 use crate::error::{DataFusionError, Result};
 use crate::logical_expr::{
-    CreateCatalog, CreateCatalogSchema, CreateExternalTable, CreateMemoryTable,
-    CreateView, DropCatalogSchema, DropTable, DropView, Explain, LogicalPlan,
-    LogicalPlanBuilder, SetVariable, TableSource, TableType, UNNAMED_TABLE,
+    CreateCatalog, CreateCatalogSchema, CreateExternalTable, CreateFunction,
+    CreateMemoryTable, CreateView, DropCatalogSchema, DropFunction, DropTable, DropView,
+    Explain, LogicalPlan, LogicalPlanBuilder, SetVariable, TableSource, TableType,
+    UNNAMED_TABLE,
 };
 use crate::optimizer::OptimizerRule;
 use datafusion_sql::{
@@ -489,6 +490,8 @@ impl SessionContext {
                 DdlStatement::DropTable(cmd) => self.drop_table(cmd).await,
                 DdlStatement::DropView(cmd) => self.drop_view(cmd).await,
                 DdlStatement::DropCatalogSchema(cmd) => self.drop_schema(cmd).await,
+                DdlStatement::CreateFunction(cmd) => self.create_function(cmd).await,
+                DdlStatement::DropFunction(cmd) => self.drop_function(cmd).await,
             },
             // TODO what about the other statements (like TransactionStart and TransactionEnd)
             LogicalPlan::Statement(Statement::SetVariable(stmt)) => {
@@ -794,6 +797,55 @@ impl SessionContext {
         Ok(false)
     }
 
+    async fn create_function(&self, stmt: CreateFunction) -> Result<DataFrame> {
+        let function = {
+            let state = self.state.read().clone();
+            let function_factory = &state.function_factory;
+
+            match function_factory {
+                Some(f) => f.create(state.config(), stmt).await?,
+                _ => Err(DataFusionError::Configuration(
+                    "Function factory has not been configured".into(),
+                ))?,
+            }
+        };
+
+        match function {
+            RegisterFunction::Scalar(f) => {
+                self.state.write().register_udf(f)?;
+            }
+            RegisterFunction::Aggregate(f) => {
+                self.state.write().register_udaf(f)?;
+            }
+            RegisterFunction::Window(f) => {
+                self.state.write().register_udwf(f)?;
+            }
+            RegisterFunction::Table(name, f) => self.register_udtf(&name, f),
+        };
+
+        self.return_empty_dataframe()
+    }
+
+    async fn drop_function(&self, stmt: DropFunction) -> Result<DataFrame> {
+        // we don't know function type at this point
+        // decision has been made to drop all functions
+        let mut dropped = false;
+        dropped |= self.state.write().deregister_udf(&stmt.name)?.is_some();
+        dropped |= self.state.write().deregister_udaf(&stmt.name)?.is_some();
+        dropped |= self.state.write().deregister_udwf(&stmt.name)?.is_some();
+
+        // DROP FUNCTION IF EXISTS drops the specified function only if that
+        // function exists and in this way, it avoids error. While the DROP FUNCTION
+        // statement also performs the same function, it throws an
+        // error if the function does not exist.
+
+        if !stmt.if_exists && !dropped {
+            exec_err!("Function does not exist")
+        } else {
+            self.return_empty_dataframe()
+        }
+    }
+
     /// Registers a variable provider within this context.
     pub fn register_variable(
         &self,
@@ -1261,7 +1313,30 @@ impl QueryPlanner for DefaultQueryPlanner {
             .await
     }
 }
+/// A pluggable interface to handle `CREATE FUNCTION` statements
+/// and interact with [SessionState] to registers new udf, udaf or udwf.
+
+#[async_trait]
+pub trait FunctionFactory: Sync + Send {
+    /// Handles creation of user defined function specified in [CreateFunction] statement
+    async fn create(
+        &self,
+        state: &SessionConfig,
+        statement: CreateFunction,
+    ) -> Result<RegisterFunction>;
+}
 
+/// Type of function to create
+pub enum RegisterFunction {
+    /// Scalar user defined function
+    Scalar(Arc<ScalarUDF>),
+    /// Aggregate user defined function
+    Aggregate(Arc<AggregateUDF>),
+    /// Window user defined function
+    Window(Arc<WindowUDF>),
+    /// Table user defined function
+    Table(String, Arc<dyn TableFunctionImpl>),
+}
 /// Execution context for registering data sources and executing queries.
 /// See [`SessionContext`] for a higher level API.
 ///
@@ -1306,6 +1381,12 @@ pub struct SessionState {
     table_factories: HashMap<String, Arc<dyn TableProviderFactory>>,
     /// Runtime environment
     runtime_env: Arc<RuntimeEnv>,
+
+    /// [FunctionFactory] to support pluggable user defined function handler.
+    ///
+    /// It will be invoked on `CREATE FUNCTION` statements.
+    /// thus, changing dialect o PostgreSql is required
+    function_factory: Option<Arc<dyn FunctionFactory>>,
 }
 
 impl Debug for SessionState {
@@ -1392,6 +1473,7 @@ impl SessionState {
             execution_props: ExecutionProps::new(),
             runtime_env: runtime,
             table_factories,
+            function_factory: None,
         };
 
         // register built in functions
@@ -1568,6 +1650,15 @@ impl SessionState {
         self
     }
 
+    /// Registers a [`FunctionFactory`] to handle `CREATE FUNCTION` statements
+    pub fn with_function_factory(
+        mut self,
+        function_factory: Arc<dyn FunctionFactory>,
+    ) -> Self {
+        self.function_factory = Some(function_factory);
+        self
+    }
+
     /// Replace the extension [`SerializerRegistry`]
     pub fn with_serializer_registry(
         mut self,

datafusion/core/tests/user_defined/user_defined_scalar_functions.rs

@@ -17,10 +17,12 @@
 
 use arrow::compute::kernels::numeric::add;
 use arrow_array::{
-    Array, ArrayRef, Float32Array, Float64Array, Int32Array, RecordBatch, UInt8Array,
+    Array, ArrayRef, ArrowNativeTypeOp, Float32Array, Float64Array, Int32Array,
+    RecordBatch, UInt8Array,
 };
 use arrow_schema::DataType::Float64;
 use arrow_schema::{DataType, Field, Schema};
+use datafusion::execution::context::{FunctionFactory, RegisterFunction, SessionState};
 use datafusion::prelude::*;
 use datafusion::{execution::registry::FunctionRegistry, test_util};
 use datafusion_common::cast::as_float64_array;
@@ -31,10 +33,12 @@ use datafusion_common::{
 use datafusion_expr::simplify::ExprSimplifyResult;
 use datafusion_expr::simplify::SimplifyInfo;
 use datafusion_expr::{
-    create_udaf, create_udf, Accumulator, ColumnarValue, ExprSchemable,
+    create_udaf, create_udf, Accumulator, ColumnarValue, CreateFunction, ExprSchemable,
     LogicalPlanBuilder, ScalarUDF, ScalarUDFImpl, Signature, Volatility,
 };
+use parking_lot::Mutex;
 
+use datafusion_execution::runtime_env::{RuntimeConfig, RuntimeEnv};
 use rand::{thread_rng, Rng};
 use std::any::Any;
 use std::iter;
@@ -735,6 +739,128 @@ async fn verify_udf_return_type() -> Result<()> {
     Ok(())
 }
 
+#[derive(Debug, Default)]
+struct MockFunctionFactory {
+    pub captured_expr: Mutex<Option<Expr>>,
+}
+
+#[async_trait::async_trait]
+impl FunctionFactory for MockFunctionFactory {
+    #[doc = r" Crates and registers a function from [CreateFunction] statement"]
+    #[must_use]
+    #[allow(clippy::type_complexity, clippy::type_repetition_in_bounds)]
+    async fn create(
+        &self,
+        _config: &SessionConfig,
+        statement: CreateFunction,
+    ) -> datafusion::error::Result<RegisterFunction> {
+        // In this example, we always create a function that adds its arguments
+        // with the name specified in `CREATE FUNCTION`. In a real implementation
+        // the body of the created UDF would also likely be a function of the contents
+        // of the `CreateFunction`
+        let mock_add = Arc::new(|args: &[datafusion_expr::ColumnarValue]| {
+            let args = datafusion_expr::ColumnarValue::values_to_arrays(args)?;
+            let base =
+                datafusion_common::cast::as_float64_array(&args[0]).expect("cast failed");
+            let exponent =
+                datafusion_common::cast::as_float64_array(&args[1]).expect("cast failed");
+
+            let array = base
+                .iter()
+                .zip(exponent.iter())
+                .map(|(base, exponent)| match (base, exponent) {
+                    (Some(base), Some(exponent)) => Some(base.add_wrapping(exponent)),
+                    _ => None,
+                })
+                .collect::<arrow_array::Float64Array>();
+            Ok(datafusion_expr::ColumnarValue::from(
+                Arc::new(array) as arrow_array::ArrayRef
+            ))
+        });
+
+        let args = statement.args.unwrap();
+        let mock_udf = create_udf(
+            &statement.name,
+            vec![args[0].data_type.clone(), args[1].data_type.clone()],
+            Arc::new(statement.return_type.unwrap()),
+            datafusion_expr::Volatility::Immutable,
+            mock_add,
+        );
+
+        // capture expression so we can verify
+        // it has been parsed
+        *self.captured_expr.lock() = statement.params.return_;
+
+        Ok(RegisterFunction::Scalar(Arc::new(mock_udf)))
+    }
+}
+
+#[tokio::test]
+async fn create_scalar_function_from_sql_statement() -> Result<()> {
+    let function_factory = Arc::new(MockFunctionFactory::default());
+    let runtime_config = RuntimeConfig::new();
+    let runtime_environment = RuntimeEnv::new(runtime_config)?;
+
+    let session_config = SessionConfig::new();
+    let state =
+        SessionState::new_with_config_rt(session_config, Arc::new(runtime_environment))
+            .with_function_factory(function_factory.clone());
+
+    let ctx = SessionContext::new_with_state(state);
+    let options = SQLOptions::new().with_allow_ddl(false);
+
+    let sql = r#"
+    CREATE FUNCTION better_add(DOUBLE, DOUBLE)
+        RETURNS DOUBLE
+        RETURN $1 + $2
+    "#;
+
+    // try to `create function` when sql options have allow ddl disabled
+    assert!(ctx.sql_with_options(sql, options).await.is_err());
+
+    // Create the `better_add` function dynamically via CREATE FUNCTION statement
+    assert!(ctx.sql(sql).await.is_ok());
+    // try to `drop function` when sql options have allow ddl disabled
+    assert!(ctx
+        .sql_with_options("drop function better_add", options)
+        .await
+        .is_err());
+
+    ctx.sql("select better_add(2.0, 2.0)").await?.show().await?;
+
+    // check if we sql expr has been converted to datafusion expr
+    let captured_expression = function_factory.captured_expr.lock().clone().unwrap();
+    assert_eq!("$1 + $2", captured_expression.to_string());
+
+    // statement drops  function
+    assert!(ctx.sql("drop function better_add").await.is_ok());
+    // no function, it panics
+    assert!(ctx.sql("drop function better_add").await.is_err());
+    // no function, it dies not care
+    assert!(ctx.sql("drop function if exists better_add").await.is_ok());
+    // query should fail as there is no function
+    assert!(ctx.sql("select better_add(2.0, 2.0)").await.is_err());
+
+    // tests expression parsing
+    // if expression is not correct
+    let bad_expression_sql = r#"
+    CREATE FUNCTION bad_expression_fun(DOUBLE, DOUBLE)
+        RETURNS DOUBLE
+        RETURN $1 $3
+    "#;
+    assert!(ctx.sql(bad_expression_sql).await.is_err());
+
+    // tests bad function definition
+    let bad_definition_sql = r#"
+    CREATE FUNCTION bad_definition_fun(DOUBLE, DOUBLE)
+        RET BAD_TYPE
+        RETURN $1 + $3
+    "#;
+    assert!(ctx.sql(bad_definition_sql).await.is_err());
+
+    Ok(())
+}
+
 fn create_udf_context() -> SessionContext {
     let ctx = SessionContext::new();
     // register a custom UDF