Consider timezones with UTC and +00:00 to be the same (apache#10960)

* feat: add temporal_coercion check * fix: add return stmt * chore: add slts * fix: remove println * Update datafusion/expr/src/type_coercion/binary.rs --------- Co-authored-by: Andrew Lamb <[email protected]>
influxdata · Jun 24, 2024 · 21c0f6e · 21c0f6e
1 parent b26f680
commit 21c0f6e
Show file tree

Hide file tree

Showing 2 changed files with 32 additions and 5 deletions.
diff --git a/datafusion/expr/src/type_coercion/binary.rs b/datafusion/expr/src/type_coercion/binary.rs
@@ -1050,12 +1050,16 @@ fn temporal_coercion(lhs_type: &DataType, rhs_type: &DataType) -> Option<DataTyp
         }
         (Timestamp(lhs_unit, lhs_tz), Timestamp(rhs_unit, rhs_tz)) => {
             let tz = match (lhs_tz, rhs_tz) {
-                // can't cast across timezones
                 (Some(lhs_tz), Some(rhs_tz)) => {
-                    if lhs_tz != rhs_tz {
-                        return None;
-                    } else {
-                        Some(lhs_tz.clone())
+                    match (lhs_tz.as_ref(), rhs_tz.as_ref()) {
+                        // UTC and "+00:00" are the same by definition. Most other timezones
+                        // do not have a 1-1 mapping between timezone and an offset from UTC
+                        ("UTC", "+00:00") | ("+00:00", "UTC") => Some(lhs_tz.clone()),
+                        (lhs, rhs) if lhs == rhs => Some(lhs_tz.clone()),
+                        // can't cast across timezones
+                        _ => {
+                            return None;
+                        }
                     }
                 }
                 (Some(lhs_tz), None) => Some(lhs_tz.clone()),

diff --git a/datafusion/sqllogictest/test_files/timestamps.slt b/datafusion/sqllogictest/test_files/timestamps.slt
@@ -2801,3 +2801,26 @@ query B
 select current_time = current_time;
 ----
 true
+
+# Test temporal coercion for UTC
+query ?
+select arrow_cast('2024-06-17T11:00:00', 'Timestamp(Nanosecond, Some("UTC"))') - arrow_cast('2024-06-17T12:00:00', 'Timestamp(Microsecond, Some("UTC"))');
+----
+0 days -1 hours 0 mins 0.000000 secs
+
+query ?
+select arrow_cast('2024-06-17T13:00:00', 'Timestamp(Nanosecond, Some("+00:00"))') - arrow_cast('2024-06-17T12:00:00', 'Timestamp(Microsecond, Some("UTC"))');
+----
+0 days 1 hours 0 mins 0.000000 secs
+
+query ?
+select arrow_cast('2024-06-17T13:00:00', 'Timestamp(Nanosecond, Some("UTC"))') - arrow_cast('2024-06-17T12:00:00', 'Timestamp(Microsecond, Some("+00:00"))');
+----
+0 days 1 hours 0 mins 0.000000 secs
+
+# not supported: coercion across timezones
+query error
+select arrow_cast('2024-06-17T13:00:00', 'Timestamp(Nanosecond, Some("UTC"))') - arrow_cast('2024-06-17T12:00:00', 'Timestamp(Microsecond, Some("+01:00"))');
+
+query error
+select arrow_cast('2024-06-17T13:00:00', 'Timestamp(Nanosecond, Some("+00:00"))') - arrow_cast('2024-06-17T12:00:00', 'Timestamp(Microsecond, Some("+01:00"))');