feat(functions): refactor from_parts registration and add comprehensive DST tests

Author: tharunn0

src/query/expression/src/utils/date_helper.rs

@@ -2069,8 +2069,8 @@ pub fn pg_format_to_strftime(pg_format_string: &str) -> String {
 }
 
 /// Normalize month/day values that may be outside normal ranges.
-/// Snowflake allows e.g. month=0 (meaning Dec of previous year),
-/// month=13 (meaning Jan of next year), day=100 (100th day from month start), etc.
+/// Supports values like month=0 (meaning Dec of previous year),
+/// month=13 (meaning Jan of next year), and day=100 (100th day from month start).
 pub fn normalize_date_parts(year: i64, month: i64, day: i64) -> std::result::Result<Date, String> {
     // Normalize month: month is 1-based, so month=0 means Dec of previous year,
     // month=-1 means Nov of previous year, month=13 means Jan of next year.

src/query/functions/src/scalars/timestamp/src/datetime.rs

@@ -2786,8 +2786,16 @@ fn register_timestamp_from_parts(registry: &mut FunctionRegistry) {
             Vec<DataType>,
             fn(&[Value<AnyType>], &mut EvalContext) -> Value<AnyType>,
         ) = match args_type.len() {
-            6 => (vec![int64_type; 6], timestamp_from_parts_6 as _),
-            7 => (vec![int64_type; 7], timestamp_from_parts_7 as _),
+            // 6 args: year, month, day, hour, minute, second
+            6 => (vec![int64_type; 6], |args, ctx| {
+                timestamp_from_parts_fn(args, ctx, false)
+            }),
+
+            // 7 args: includes nanosecond
+            7 => (vec![int64_type; 7], |args, ctx| {
+                timestamp_from_parts_fn(args, ctx, true)
+            }),
+
             _ => return None,
         };
 
@@ -2809,14 +2817,6 @@ fn register_timestamp_from_parts(registry: &mut FunctionRegistry) {
     registry.register_function_factory("timestamp_from_parts", factory);
 }
 
-fn timestamp_from_parts_6(args: &[Value<AnyType>], ctx: &mut EvalContext) -> Value<AnyType> {
-    timestamp_from_parts_fn(args, ctx, false)
-}
-
-fn timestamp_from_parts_7(args: &[Value<AnyType>], ctx: &mut EvalContext) -> Value<AnyType> {
-    timestamp_from_parts_fn(args, ctx, true)
-}
-
 fn register_timestamp_tz_from_parts(registry: &mut FunctionRegistry) {
     registry.register_aliases("timestamp_tz_from_parts", &["timestamptzfromparts"]);
 
@@ -2828,20 +2828,36 @@ fn register_timestamp_tz_from_parts(registry: &mut FunctionRegistry) {
             Vec<DataType>,
             fn(&[Value<AnyType>], &mut EvalContext) -> Value<AnyType>,
         ) = match args_type.len() {
-            6 => (vec![int64_type; 6], timestamp_tz_from_parts_6 as _),
+            // 6 args: no nanoseconds, no timezone
+            6 => (vec![int64_type; 6], |args, ctx| {
+                timestamp_tz_from_parts_fn(args, ctx, false, false)
+            }),
 
+            // 7 args: timezone provided
             7 if args_type[6].remove_nullable() == DataType::String => {
                 let mut v = vec![int64_type; 6];
                 v.push(DataType::String);
-                (v, timestamp_tz_from_parts_7_tz as _)
+
+                // year, month, day, hour, minute, second, timezone
+                (v, |args, ctx| {
+                    timestamp_tz_from_parts_fn(args, ctx, false, true)
+                })
             }
 
-            7 => (vec![int64_type; 7], timestamp_tz_from_parts_7_nano as _),
+            // 7 args: nanoseconds
+            7 => (vec![int64_type; 7], |args, ctx| {
+                timestamp_tz_from_parts_fn(args, ctx, true, false)
+            }),
 
+            // 8 args: nanoseconds + timezone
             8 => {
                 let mut v = vec![int64_type; 7];
                 v.push(DataType::String);
-                (v, timestamp_tz_from_parts_8 as _)
+
+                // year, month, day, hour, minute, second, nanosecond, timezone
+                (v, |args, ctx| {
+                    timestamp_tz_from_parts_fn(args, ctx, true, true)
+                })
             }
 
             _ => return None,
@@ -2865,25 +2881,6 @@ fn register_timestamp_tz_from_parts(registry: &mut FunctionRegistry) {
     registry.register_function_factory("timestamp_tz_from_parts", factory);
 }
 
-fn timestamp_tz_from_parts_6(args: &[Value<AnyType>], ctx: &mut EvalContext) -> Value<AnyType> {
-    timestamp_tz_from_parts_fn(args, ctx, false, false)
-}
-
-fn timestamp_tz_from_parts_7_nano(
-    args: &[Value<AnyType>],
-    ctx: &mut EvalContext,
-) -> Value<AnyType> {
-    timestamp_tz_from_parts_fn(args, ctx, true, false)
-}
-
-fn timestamp_tz_from_parts_7_tz(args: &[Value<AnyType>], ctx: &mut EvalContext) -> Value<AnyType> {
-    timestamp_tz_from_parts_fn(args, ctx, false, true)
-}
-
-fn timestamp_tz_from_parts_8(args: &[Value<AnyType>], ctx: &mut EvalContext) -> Value<AnyType> {
-    timestamp_tz_from_parts_fn(args, ctx, true, true)
-}
-
 fn build_timestamp_parts(
     args: &[Value<AnyType>],
     ctx: &mut EvalContext,

tests/sqllogictests/suites/query/functions/02_0012_function_datetimes.test

@@ -1842,7 +1842,6 @@ select timestamp_from_parts(2010, 13, 1, 0, 0, 0);
 ----
 2011-01-01 00:00:00.000000
 
-
 ## Additional TIMESTAMP_TZ_FROM_PARTS tests
 
 statement ok
@@ -1893,3 +1892,63 @@ select timestamp_tz_from_parts(2013, 13, 1, 0, 0, 0, 'UTC');
 # Invalid timezone should produce error
 query error
 select timestamp_tz_from_parts(2013, 4, 5, 12, 0, 0, 'INVALID_TZ');
+
+# America/Los_Angeles winter (PST)
+query ?
+select timestamp_tz_from_parts(2023, 1, 1, 12, 0, 0, 'America/Los_Angeles');
+----
+2023-01-01 12:00:00.000000 -0800
+
+# America/Los_Angeles summer (PDT)
+query ?
+select timestamp_tz_from_parts(2023, 7, 1, 12, 0, 0, 'America/Los_Angeles');
+----
+2023-07-01 12:00:00.000000 -0700
+
+# DST start boundary (spring forward)
+query ?
+select timestamp_tz_from_parts(2023, 3, 12, 3, 0, 0, 'America/Los_Angeles');
+----
+2023-03-12 03:00:00.000000 -0700
+
+# Before DST transition
+query ?
+select timestamp_tz_from_parts(2023, 3, 12, 1, 30, 0, 'America/Los_Angeles');
+----
+2023-03-12 01:30:00.000000 -0800
+
+# After DST transition
+query ?
+select timestamp_tz_from_parts(2023, 3, 12, 4, 0, 0, 'America/Los_Angeles');
+----
+2023-03-12 04:00:00.000000 -0700
+
+# DST fall back (overlap hour)
+query ?
+select timestamp_tz_from_parts(2023, 11, 5, 1, 30, 0, 'America/Los_Angeles');
+----
+2023-11-05 01:30:00.000000 -0700
+
+# After DST fallback
+query ?
+select timestamp_tz_from_parts(2023, 11, 5, 2, 30, 0, 'America/Los_Angeles');
+----
+2023-11-05 02:30:00.000000 -0800
+
+# Asia/Shanghai (no DST behavior)
+query ?
+select timestamp_tz_from_parts(2023, 6, 1, 12, 0, 0, 'Asia/Shanghai');
+----
+2023-06-01 12:00:00.000000 +0800
+
+# Nanosecond overflow with timezone
+query ?
+select timestamp_tz_from_parts(2023, 1, 1, 12, 0, 0, 2000000000, 'UTC');
+----
+2023-01-01 12:00:02.000000 +0000
+
+# Negative nanoseconds
+query ?
+select timestamp_tz_from_parts(2023, 1, 1, 12, 0, 0, -1000000000, 'UTC');
+----
+2023-01-01 11:59:59.000000 +0000