Refactor to_local_time() signature away from user_defined (#18707)

## Which issue does this PR close?

<!--
We generally require a GitHub issue to be filed for all bug fixes and
enhancements and this helps us generate change logs for our releases.
You can link an issue to this PR using the GitHub syntax. For example
`Closes #123` indicates that this PR will close issue #123.
-->

Part of #12725

## Rationale for this change

<!--
Why are you proposing this change? If this is already explained clearly
in the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand
your changes and offer better suggestions for fixes.
-->

Prefer to avoid user_defined for consistency in function definitions.

## What changes are included in this PR?

<!--
There is no need to duplicate the description in the issue here but it
is sometimes worth providing a summary of the individual changes in this
PR.
-->

Refactor signature of `to_local_time` away from user_defined.

Also some other refactors for `to_local_time`.

## Are these changes tested?

<!--
We typically require tests for all PRs in order to:
1. Prevent the code from being accidentally broken by subsequent changes
2. Serve as another way to document the expected behavior of the code

If tests are not included in your PR, please explain why (for example,
are they covered by existing tests)?
-->

Existing tests.

## Are there any user-facing changes?

<!--
If there are user-facing changes then we may require documentation to be
updated before approving the PR.
-->

No.

<!--
If there are any breaking changes to public APIs, please add the `api
change` label.
-->

Author: Jefffrey

datafusion/functions/src/datetime/to_local_time.rs

@@ -20,7 +20,7 @@ use std::ops::Add;
 use std::sync::Arc;
 
 use arrow::array::timezone::Tz;
-use arrow::array::{Array, ArrayRef, PrimitiveBuilder};
+use arrow::array::{ArrayRef, PrimitiveBuilder};
 use arrow::datatypes::DataType::Timestamp;
 use arrow::datatypes::TimeUnit::{Microsecond, Millisecond, Nanosecond, Second};
 use arrow::datatypes::{
@@ -31,11 +31,12 @@ use chrono::{DateTime, MappedLocalTime, Offset, TimeDelta, TimeZone, Utc};
 
 use datafusion_common::cast::as_primitive_array;
 use datafusion_common::{
-    exec_err, internal_datafusion_err, plan_err, utils::take_function_args, Result,
+    exec_err, internal_datafusion_err, internal_err, utils::take_function_args, Result,
     ScalarValue,
 };
 use datafusion_expr::{
-    ColumnarValue, Documentation, ScalarUDFImpl, Signature, Volatility,
+    Coercion, ColumnarValue, Documentation, ScalarUDFImpl, Signature, TypeSignatureClass,
+    Volatility,
 };
 use datafusion_macros::user_doc;
 
@@ -111,133 +112,163 @@ impl Default for ToLocalTimeFunc {
 impl ToLocalTimeFunc {
     pub fn new() -> Self {
         Self {
-            signature: Signature::user_defined(Volatility::Immutable),
+            signature: Signature::coercible(
+                vec![Coercion::new_exact(TypeSignatureClass::Timestamp)],
+                Volatility::Immutable,
+            ),
         }
     }
+}
 
-    fn to_local_time(&self, args: &[ColumnarValue]) -> Result<ColumnarValue> {
-        let [time_value] = take_function_args(self.name(), args)?;
+impl ScalarUDFImpl for ToLocalTimeFunc {
+    fn as_any(&self) -> &dyn Any {
+        self
+    }
 
-        let arg_type = time_value.data_type();
-        match arg_type {
-            Timestamp(_, None) => {
-                // if no timezone specified, just return the input
-                Ok(time_value.clone())
-            }
-            // If has timezone, adjust the underlying time value. The current time value
-            // is stored as i64 in UTC, even though the timezone may not be in UTC. Therefore,
-            // we need to adjust the time value to the local time. See [`adjust_to_local_time`]
-            // for more details.
-            //
-            // Then remove the timezone in return type, i.e. return None
-            Timestamp(_, Some(timezone)) => {
-                let tz: Tz = timezone.parse()?;
-
-                match time_value {
-                    ColumnarValue::Scalar(ScalarValue::TimestampNanosecond(
-                        Some(ts),
-                        Some(_),
-                    )) => {
-                        let adjusted_ts =
-                            adjust_to_local_time::<TimestampNanosecondType>(*ts, tz)?;
-                        Ok(ColumnarValue::Scalar(ScalarValue::TimestampNanosecond(
-                            Some(adjusted_ts),
-                            None,
-                        )))
-                    }
-                    ColumnarValue::Scalar(ScalarValue::TimestampMicrosecond(
-                        Some(ts),
-                        Some(_),
-                    )) => {
-                        let adjusted_ts =
-                            adjust_to_local_time::<TimestampMicrosecondType>(*ts, tz)?;
-                        Ok(ColumnarValue::Scalar(ScalarValue::TimestampMicrosecond(
-                            Some(adjusted_ts),
-                            None,
-                        )))
-                    }
-                    ColumnarValue::Scalar(ScalarValue::TimestampMillisecond(
-                        Some(ts),
-                        Some(_),
-                    )) => {
-                        let adjusted_ts =
-                            adjust_to_local_time::<TimestampMillisecondType>(*ts, tz)?;
-                        Ok(ColumnarValue::Scalar(ScalarValue::TimestampMillisecond(
-                            Some(adjusted_ts),
-                            None,
-                        )))
-                    }
-                    ColumnarValue::Scalar(ScalarValue::TimestampSecond(
-                        Some(ts),
-                        Some(_),
-                    )) => {
-                        let adjusted_ts =
-                            adjust_to_local_time::<TimestampSecondType>(*ts, tz)?;
-                        Ok(ColumnarValue::Scalar(ScalarValue::TimestampSecond(
-                            Some(adjusted_ts),
-                            None,
-                        )))
-                    }
-                    ColumnarValue::Array(array) => {
-                        fn transform_array<T: ArrowTimestampType>(
-                            array: &ArrayRef,
-                            tz: Tz,
-                        ) -> Result<ColumnarValue> {
-                            let mut builder = PrimitiveBuilder::<T>::new();
-
-                            let primitive_array = as_primitive_array::<T>(array)?;
-                            for ts_opt in primitive_array.iter() {
-                                match ts_opt {
-                                    None => builder.append_null(),
-                                    Some(ts) => {
-                                        let adjusted_ts: i64 =
-                                            adjust_to_local_time::<T>(ts, tz)?;
-                                        builder.append_value(adjusted_ts)
-                                    }
-                                }
-                            }
-
-                            Ok(ColumnarValue::Array(Arc::new(builder.finish())))
-                        }
-
-                        match array.data_type() {
-                            Timestamp(_, None) => {
-                                // if no timezone specified, just return the input
-                                Ok(time_value.clone())
-                            }
-                            Timestamp(Nanosecond, Some(_)) => {
-                                transform_array::<TimestampNanosecondType>(array, tz)
-                            }
-                            Timestamp(Microsecond, Some(_)) => {
-                                transform_array::<TimestampMicrosecondType>(array, tz)
-                            }
-                            Timestamp(Millisecond, Some(_)) => {
-                                transform_array::<TimestampMillisecondType>(array, tz)
-                            }
-                            Timestamp(Second, Some(_)) => {
-                                transform_array::<TimestampSecondType>(array, tz)
-                            }
-                            _ => {
-                                exec_err!("to_local_time function requires timestamp argument in array, got {:?}", array.data_type())
-                            }
-                        }
-                    }
-                    _ => {
-                        exec_err!(
-                        "to_local_time function requires timestamp argument, got {:?}",
-                        time_value.data_type()
-                    )
-                    }
-                }
-            }
-            _ => {
-                exec_err!(
-                    "to_local_time function requires timestamp argument, got {:?}",
-                    arg_type
-                )
+    fn name(&self) -> &str {
+        "to_local_time"
+    }
+
+    fn signature(&self) -> &Signature {
+        &self.signature
+    }
+
+    fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> {
+        match &arg_types[0] {
+            DataType::Null => Ok(Timestamp(Nanosecond, None)),
+            Timestamp(timeunit, _) => Ok(Timestamp(*timeunit, None)),
+            dt => internal_err!(
+                "The to_local_time function can only accept timestamp as the arg, got {dt}"
+            ),
+        }
+    }
+
+    fn invoke_with_args(
+        &self,
+        args: datafusion_expr::ScalarFunctionArgs,
+    ) -> Result<ColumnarValue> {
+        let [time_value] = take_function_args(self.name(), &args.args)?;
+        to_local_time(time_value)
+    }
+
+    fn documentation(&self) -> Option<&Documentation> {
+        self.doc()
+    }
+}
+
+fn transform_array<T: ArrowTimestampType>(
+    array: &ArrayRef,
+    tz: Tz,
+) -> Result<ColumnarValue> {
+    let primitive_array = as_primitive_array::<T>(array)?;
+    let mut builder = PrimitiveBuilder::<T>::with_capacity(primitive_array.len());
+    for ts_opt in primitive_array.iter() {
+        match ts_opt {
+            None => builder.append_null(),
+            Some(ts) => {
+                let adjusted_ts: i64 = adjust_to_local_time::<T>(ts, tz)?;
+                builder.append_value(adjusted_ts)
             }
         }
     }
+
+    Ok(ColumnarValue::Array(Arc::new(builder.finish())))
+}
+
+fn to_local_time(time_value: &ColumnarValue) -> Result<ColumnarValue> {
+    let arg_type = time_value.data_type();
+
+    let tz: Tz = match &arg_type {
+        Timestamp(_, Some(timezone)) => timezone.parse()?,
+        Timestamp(_, None) => {
+            // if no timezone specified, just return the input
+            return Ok(time_value.clone());
+        }
+        DataType::Null => {
+            return Ok(ColumnarValue::Scalar(ScalarValue::TimestampNanosecond(
+                None, None,
+            )));
+        }
+        dt => {
+            return internal_err!(
+                "to_local_time function requires timestamp argument, got {dt}"
+            )
+        }
+    };
+
+    // If has timezone, adjust the underlying time value. The current time value
+    // is stored as i64 in UTC, even though the timezone may not be in UTC. Therefore,
+    // we need to adjust the time value to the local time. See [`adjust_to_local_time`]
+    // for more details.
+    //
+    // Then remove the timezone in return type, i.e. return None
+    match time_value {
+        ColumnarValue::Scalar(ScalarValue::TimestampSecond(None, Some(_))) => Ok(
+            ColumnarValue::Scalar(ScalarValue::TimestampSecond(None, None)),
+        ),
+        ColumnarValue::Scalar(ScalarValue::TimestampMillisecond(None, Some(_))) => Ok(
+            ColumnarValue::Scalar(ScalarValue::TimestampMillisecond(None, None)),
+        ),
+        ColumnarValue::Scalar(ScalarValue::TimestampMicrosecond(None, Some(_))) => Ok(
+            ColumnarValue::Scalar(ScalarValue::TimestampMicrosecond(None, None)),
+        ),
+        ColumnarValue::Scalar(ScalarValue::TimestampNanosecond(None, Some(_))) => Ok(
+            ColumnarValue::Scalar(ScalarValue::TimestampNanosecond(None, None)),
+        ),
+        ColumnarValue::Scalar(ScalarValue::TimestampNanosecond(Some(ts), Some(_))) => {
+            let adjusted_ts = adjust_to_local_time::<TimestampNanosecondType>(*ts, tz)?;
+            Ok(ColumnarValue::Scalar(ScalarValue::TimestampNanosecond(
+                Some(adjusted_ts),
+                None,
+            )))
+        }
+        ColumnarValue::Scalar(ScalarValue::TimestampMicrosecond(Some(ts), Some(_))) => {
+            let adjusted_ts = adjust_to_local_time::<TimestampMicrosecondType>(*ts, tz)?;
+            Ok(ColumnarValue::Scalar(ScalarValue::TimestampMicrosecond(
+                Some(adjusted_ts),
+                None,
+            )))
+        }
+        ColumnarValue::Scalar(ScalarValue::TimestampMillisecond(Some(ts), Some(_))) => {
+            let adjusted_ts = adjust_to_local_time::<TimestampMillisecondType>(*ts, tz)?;
+            Ok(ColumnarValue::Scalar(ScalarValue::TimestampMillisecond(
+                Some(adjusted_ts),
+                None,
+            )))
+        }
+        ColumnarValue::Scalar(ScalarValue::TimestampSecond(Some(ts), Some(_))) => {
+            let adjusted_ts = adjust_to_local_time::<TimestampSecondType>(*ts, tz)?;
+            Ok(ColumnarValue::Scalar(ScalarValue::TimestampSecond(
+                Some(adjusted_ts),
+                None,
+            )))
+        }
+        ColumnarValue::Array(array)
+            if matches!(array.data_type(), Timestamp(Nanosecond, Some(_))) =>
+        {
+            transform_array::<TimestampNanosecondType>(array, tz)
+        }
+        ColumnarValue::Array(array)
+            if matches!(array.data_type(), Timestamp(Microsecond, Some(_))) =>
+        {
+            transform_array::<TimestampMicrosecondType>(array, tz)
+        }
+        ColumnarValue::Array(array)
+            if matches!(array.data_type(), Timestamp(Millisecond, Some(_))) =>
+        {
+            transform_array::<TimestampMillisecondType>(array, tz)
+        }
+        ColumnarValue::Array(array)
+            if matches!(array.data_type(), Timestamp(Second, Some(_))) =>
+        {
+            transform_array::<TimestampSecondType>(array, tz)
+        }
+        _ => {
+            internal_err!(
+                "to_local_time function requires timestamp argument, got {arg_type}"
+            )
+        }
+    }
 }
 
 /// This function converts a timestamp with a timezone to a timestamp without a timezone.
@@ -343,68 +374,6 @@ fn adjust_to_local_time<T: ArrowTimestampType>(ts: i64, tz: Tz) -> Result<i64> {
     }
 }
 
-impl ScalarUDFImpl for ToLocalTimeFunc {
-    fn as_any(&self) -> &dyn Any {
-        self
-    }
-
-    fn name(&self) -> &str {
-        "to_local_time"
-    }
-
-    fn signature(&self) -> &Signature {
-        &self.signature
-    }
-
-    fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> {
-        let [time_value] = take_function_args(self.name(), arg_types)?;
-
-        match time_value {
-            Timestamp(timeunit, _) => Ok(Timestamp(*timeunit, None)),
-            _ => exec_err!(
-                "The to_local_time function can only accept timestamp as the arg, got {:?}", time_value
-            )
-        }
-    }
-
-    fn invoke_with_args(
-        &self,
-        args: datafusion_expr::ScalarFunctionArgs,
-    ) -> Result<ColumnarValue> {
-        let [time_value] = take_function_args(self.name(), args.args)?;
-
-        self.to_local_time(std::slice::from_ref(&time_value))
-    }
-
-    fn coerce_types(&self, arg_types: &[DataType]) -> Result<Vec<DataType>> {
-        if arg_types.len() != 1 {
-            return plan_err!(
-                "to_local_time function requires 1 argument, got {:?}",
-                arg_types.len()
-            );
-        }
-
-        let first_arg = arg_types[0].clone();
-        match &first_arg {
-            DataType::Null => Ok(vec![Timestamp(Nanosecond, None)]),
-            Timestamp(Nanosecond, timezone) => {
-                Ok(vec![Timestamp(Nanosecond, timezone.clone())])
-            }
-            Timestamp(Microsecond, timezone) => {
-                Ok(vec![Timestamp(Microsecond, timezone.clone())])
-            }
-            Timestamp(Millisecond, timezone) => {
-                Ok(vec![Timestamp(Millisecond, timezone.clone())])
-            }
-            Timestamp(Second, timezone) => Ok(vec![Timestamp(Second, timezone.clone())]),
-            _ => plan_err!("The to_local_time function can only accept Timestamp as the arg got {first_arg}"),
-        }
-    }
-    fn documentation(&self) -> Option<&Documentation> {
-        self.doc()
-    }
-}
-
 #[cfg(test)]
 mod tests {
     use std::sync::Arc;