Simplify type signatures using TypeSignatureClass for mixed type function signature

datafusion/common/src/types/native.rs

@@ -245,6 +245,8 @@ impl LogicalType for NativeType {
             (Self::FixedSizeBinary(size), _) => FixedSizeBinary(*size),
             (Self::String, LargeBinary) => LargeUtf8,
             (Self::String, BinaryView) => Utf8View,
+            // We don't cast to another kind of string type if the origin one is already a string type
+            (Self::String, Utf8 | LargeUtf8 | Utf8View) => origin.to_owned(),
             (Self::String, data_type) if can_cast_types(data_type, &Utf8View) => Utf8View,
             (Self::String, data_type) if can_cast_types(data_type, &LargeUtf8) => {
                 LargeUtf8
@@ -433,4 +435,29 @@ impl NativeType {
             UInt8 | UInt16 | UInt32 | UInt64 | Int8 | Int16 | Int32 | Int64
         )
     }
+
+    #[inline]
+    pub fn is_timestamp(&self) -> bool {
+        matches!(self, NativeType::Timestamp(_, _))
+    }
+
+    #[inline]
+    pub fn is_date(&self) -> bool {
+        matches!(self, NativeType::Date)
+    }
+
+    #[inline]
+    pub fn is_time(&self) -> bool {
+        matches!(self, NativeType::Time(_))
+    }
+
+    #[inline]
+    pub fn is_interval(&self) -> bool {
+        matches!(self, NativeType::Interval(_))
+    }
+
+    #[inline]
+    pub fn is_duration(&self) -> bool {
+        matches!(self, NativeType::Duration(_))
+    }
 }

datafusion/expr-common/src/signature.rs

@@ -18,8 +18,10 @@
 //! Signature module contains foundational types that are used to represent signatures, types,
 //! and return types of functions in DataFusion.
 
+use std::fmt::Display;
+
 use crate::type_coercion::aggregates::NUMERICS;
-use arrow::datatypes::DataType;
+use arrow::datatypes::{DataType, IntervalUnit, TimeUnit};
 use datafusion_common::types::{LogicalTypeRef, NativeType};
 use itertools::Itertools;
 
@@ -112,7 +114,7 @@ pub enum TypeSignature {
     /// For example, `Coercible(vec![logical_float64()])` accepts
     /// arguments like `vec![DataType::Int32]` or `vec![DataType::Float32]`
     /// since i32 and f32 can be casted to f64
-    Coercible(Vec<LogicalTypeRef>),
+    Coercible(Vec<TypeSignatureClass>),
     /// The arguments will be coerced to a single type based on the comparison rules.
     /// For example, i32 and i64 has coerced type Int64.
     ///
@@ -154,6 +156,33 @@ impl TypeSignature {
     }
 }
 
+/// Represents the class of types that can be used in a function signature.
+///
+/// This is used to specify what types are valid for function arguments in a more flexible way than
+/// just listing specific DataTypes. For example, TypeSignatureClass::Timestamp matches any timestamp
+/// type regardless of timezone or precision.
+///
+/// Used primarily with TypeSignature::Coercible to define function signatures that can accept
+/// arguments that can be coerced to a particular class of types.
+#[derive(Debug, Clone, Eq, PartialEq, PartialOrd, Hash)]
+pub enum TypeSignatureClass {
+    Timestamp,
+    Date,
+    Time,
+    Interval,
+    Duration,
+    Native(LogicalTypeRef),
+    // TODO:
+    // Numeric
+    // Integer
+}
+
+impl Display for TypeSignatureClass {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "TypeSignatureClass::{self:?}")
+    }
+}
+
 #[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Hash)]
 pub enum ArrayFunctionSignature {
     /// Specialized Signature for ArrayAppend and similar functions
@@ -180,7 +209,7 @@ pub enum ArrayFunctionSignature {
     MapArray,
 }
 
-impl std::fmt::Display for ArrayFunctionSignature {
+impl Display for ArrayFunctionSignature {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
         match self {
             ArrayFunctionSignature::ArrayAndElement => {
@@ -255,7 +284,7 @@ impl TypeSignature {
     }
 
     /// Helper function to join types with specified delimiter.
-    pub fn join_types<T: std::fmt::Display>(types: &[T], delimiter: &str) -> String {
+    pub fn join_types<T: Display>(types: &[T], delimiter: &str) -> String {
         types
             .iter()
             .map(|t| t.to_string())
@@ -290,7 +319,30 @@ impl TypeSignature {
                 .collect(),
             TypeSignature::Coercible(types) => types
                 .iter()
-                .map(|logical_type| get_data_types(logical_type.native()))
+                .map(|logical_type| match logical_type {
+                    TypeSignatureClass::Native(l) => get_data_types(l.native()),
+                    TypeSignatureClass::Timestamp => {
+                        vec![
+                            DataType::Timestamp(TimeUnit::Nanosecond, None),
+                            DataType::Timestamp(
+                                TimeUnit::Nanosecond,
+                                Some(TIMEZONE_WILDCARD.into()),
+                            ),
+                        ]
+                    }
+                    TypeSignatureClass::Date => {
+                        vec![DataType::Date64]
+                    }
+                    TypeSignatureClass::Time => {
+                        vec![DataType::Time64(TimeUnit::Nanosecond)]
+                    }
+                    TypeSignatureClass::Interval => {
+                        vec![DataType::Interval(IntervalUnit::DayTime)]
+                    }
+                    TypeSignatureClass::Duration => {
+                        vec![DataType::Duration(TimeUnit::Nanosecond)]
+                    }
+                })
                 .multi_cartesian_product()
                 .collect(),
             TypeSignature::Variadic(types) => types
@@ -424,7 +476,10 @@ impl Signature {
         }
     }
     /// Target coerce types in order
-    pub fn coercible(target_types: Vec<LogicalTypeRef>, volatility: Volatility) -> Self {
+    pub fn coercible(
+        target_types: Vec<TypeSignatureClass>,
+        volatility: Volatility,
+    ) -> Self {
         Self {
             type_signature: TypeSignature::Coercible(target_types),
             volatility,
@@ -618,8 +673,10 @@ mod tests {
             ]
         );
 
-        let type_signature =
-            TypeSignature::Coercible(vec![logical_string(), logical_int64()]);
+        let type_signature = TypeSignature::Coercible(vec![
+            TypeSignatureClass::Native(logical_string()),
+            TypeSignatureClass::Native(logical_int64()),
+        ]);
         let possible_types = type_signature.get_possible_types();
         assert_eq!(
             possible_types,

datafusion/expr/src/type_coercion/functions.rs

@@ -22,14 +22,18 @@ use arrow::{
     datatypes::{DataType, TimeUnit},
 };
 use datafusion_common::{
-    exec_err, internal_datafusion_err, internal_err, plan_err,
+    exec_err, internal_datafusion_err, internal_err, not_impl_err, plan_err,
     types::{LogicalType, NativeType},
     utils::{coerced_fixed_size_list_to_list, list_ndims},
     Result,
 };
 use datafusion_expr_common::{
-    signature::{ArrayFunctionSignature, FIXED_SIZE_LIST_WILDCARD, TIMEZONE_WILDCARD},
-    type_coercion::binary::{comparison_coercion_numeric, string_coercion},
+    signature::{
+        ArrayFunctionSignature, TypeSignatureClass, FIXED_SIZE_LIST_WILDCARD,
+        TIMEZONE_WILDCARD,
+    },
+    type_coercion::binary::comparison_coercion_numeric,
+    type_coercion::binary::string_coercion,
 };
 use std::sync::Arc;
 
@@ -568,35 +572,65 @@ fn get_valid_types(
             // Make sure the corresponding test is covered
             // If this function becomes COMPLEX, create another new signature!
             fn can_coerce_to(
-                logical_type: &NativeType,
-                target_type: &NativeType,
-            ) -> bool {
-                if logical_type == target_type {
-                    return true;
-                }
+                current_type: &DataType,
+                target_type_class: &TypeSignatureClass,
+            ) -> Result<DataType> {
+                let logical_type: NativeType = current_type.into();
 
-                if logical_type == &NativeType::Null {
-                    return true;
-                }
+                match target_type_class {
+                    TypeSignatureClass::Native(native_type) => {
+                        let target_type = native_type.native();
+                        if &logical_type == target_type {
+                            return target_type.default_cast_for(current_type);
+                        }
 
-                if target_type.is_integer() && logical_type.is_integer() {
-                    return true;
-                }
+                        if logical_type == NativeType::Null {
+                            return target_type.default_cast_for(current_type);
+                        }
+
+                        if target_type.is_integer() && logical_type.is_integer() {
+                            return target_type.default_cast_for(current_type);
+                        }
 
-                false
+                        internal_err!(
+                            "Expect {} but received {}",
+                            target_type_class,
+                            current_type
+                        )
+                    }
+                    // Not consistent with Postgres and DuckDB but to avoid regression we implicit cast string to timestamp
+                    TypeSignatureClass::Timestamp
+                        if logical_type == NativeType::String =>
+                    {
+                        Ok(DataType::Timestamp(TimeUnit::Nanosecond, None))
+                    }
+                    TypeSignatureClass::Timestamp if logical_type.is_timestamp() => {
+                        Ok(current_type.to_owned())
+                    }
+                    TypeSignatureClass::Date if logical_type.is_date() => {
+                        Ok(current_type.to_owned())
+                    }
+                    TypeSignatureClass::Time if logical_type.is_time() => {
+                        Ok(current_type.to_owned())
+                    }
+                    TypeSignatureClass::Interval if logical_type.is_interval() => {
+                        Ok(current_type.to_owned())
+                    }
+                    TypeSignatureClass::Duration if logical_type.is_duration() => {
+                        Ok(current_type.to_owned())
+                    }
+                    _ => {
+                        not_impl_err!("Got logical_type: {logical_type} with target_type_class: {target_type_class}")
+                    }
+                }
             }
 
             let mut new_types = Vec::with_capacity(current_types.len());
-            for (current_type, target_type) in
+            for (current_type, target_type_class) in
                 current_types.iter().zip(target_types.iter())
             {
-                let logical_type: NativeType = current_type.into();
-                let target_logical_type = target_type.native();
-                if can_coerce_to(&logical_type, target_logical_type) {
-                    let target_type =
-                        target_logical_type.default_cast_for(current_type)?;
-                    new_types.push(target_type);
-                }
+                let target_type = can_coerce_to(current_type, target_type_class)?;
+                new_types.push(target_type);
             }
 
             vec![new_types]

datafusion/functions/Cargo.toml

@@ -75,6 +75,7 @@ datafusion-common = { workspace = true }
 datafusion-doc = { workspace = true }
 datafusion-execution = { workspace = true }
 datafusion-expr = { workspace = true }
+datafusion-expr-common = { workspace = true }
 datafusion-macros = { workspace = true }
 hashbrown = { workspace = true, optional = true }
 hex = { version = "0.4", optional = true }