feat: support date_part function and more fields for extract

e2e_test/batch/duckdb/join/iejoin/test_iejoin.test.slt.part

@@ -140,46 +140,46 @@ WITH data_table AS (
     GROUP BY bucket, low, high
     ORDER BY bucket;
 ----
-0	1577836800	1578627360	10
-1	1578627360	1579417920	9
-2	1579417920	1580208480	9
-3	1580208480	1580999040	9
-4	1580999040	1581789600	9
-5	1581789600	1582580160	9
-6	1582580160	1583370720	10
-7	1583370720	1584161280	9
-8	1584161280	1584951840	9
-9	1584951840	1585742400	9
-10	1585742400	1586532960	9
-11	1586532960	1587323520	9
-12	1587323520	1588114080	9
-13	1588114080	1588904640	10
-14	1588904640	1589695200	9
-15	1589695200	1590485760	9
-16	1590485760	1591276320	9
-17	1591276320	1592066880	9
-18	1592066880	1592857440	9
-19	1592857440	1593648000	9
-20	1593648000	1594438560	10
-21	1594438560	1595229120	9
-22	1595229120	1596019680	9
-23	1596019680	1596810240	9
-24	1596810240	1597600800	9
-25	1597600800	1598391360	9
-26	1598391360	1599181920	10
-27	1599181920	1599972480	9
-28	1599972480	1600763040	9
-29	1600763040	1601553600	9
-30	1601553600	1602344160	9
-31	1602344160	1603134720	9
-32	1603134720	1603925280	9
-33	1603925280	1604715840	10
-34	1604715840	1605506400	9
-35	1605506400	1606296960	9
-36	1606296960	1607087520	9
-37	1607087520	1607878080	9
-38	1607878080	1608668640	9
-39	1608668640	1609459200	9
+0	  1577836800.000000	1578627360.000000	10
+1	  1578627360.000000	1579417920.000000	9
+2	  1579417920.000000	1580208480.000000	9
+3	  1580208480.000000	1580999040.000000	9
+4	  1580999040.000000	1581789600.000000	9
+5	  1581789600.000000	1582580160.000000	9
+6	  1582580160.000000	1583370720.000000	10
+7	  1583370720.000000	1584161280.000000	9
+8	  1584161280.000000	1584951840.000000	9
+9	  1584951840.000000	1585742400.000000	9
+10	1585742400.000000	1586532960.000000	9
+11	1586532960.000000	1587323520.000000	9
+12	1587323520.000000	1588114080.000000	9
+13	1588114080.000000	1588904640.000000	10
+14	1588904640.000000	1589695200.000000	9
+15	1589695200.000000	1590485760.000000	9
+16	1590485760.000000	1591276320.000000	9
+17	1591276320.000000	1592066880.000000	9
+18	1592066880.000000	1592857440.000000	9
+19	1592857440.000000	1593648000.000000	9
+20	1593648000.000000	1594438560.000000	10
+21	1594438560.000000	1595229120.000000	9
+22	1595229120.000000	1596019680.000000	9
+23	1596019680.000000	1596810240.000000	9
+24	1596810240.000000	1597600800.000000	9
+25	1597600800.000000	1598391360.000000	9
+26	1598391360.000000	1599181920.000000	10
+27	1599181920.000000	1599972480.000000	9
+28	1599972480.000000	1600763040.000000	9
+29	1600763040.000000	1601553600.000000	9
+30	1601553600.000000	1602344160.000000	9
+31	1602344160.000000	1603134720.000000	9
+32	1603134720.000000	1603925280.000000	9
+33	1603925280.000000	1604715840.000000	10
+34	1604715840.000000	1605506400.000000	9
+35	1605506400.000000	1606296960.000000	9
+36	1606296960.000000	1607087520.000000	9
+37	1607087520.000000	1607878080.000000	9
+38	1607878080.000000	1608668640.000000	9
+39	1608668640.000000	1609459200.000000	9
 
 
 statement ok

proto/expr.proto

@@ -45,6 +45,7 @@ message ExprNode {
     BITWISE_SHIFT_RIGHT = 36;
     // date functions
     EXTRACT = 101;
+    DATE_PART = 102;
     TUMBLE_START = 103;
     // From f64 to timestamp.
     // e.g. `select to_timestamp(1672044740.0)`

src/common/src/array/arrow.rs

@@ -353,10 +353,10 @@ impl FromIntoArrow for Interval {
 
     fn into_arrow(self) -> Self::ArrowType {
         arrow_array::types::IntervalMonthDayNanoType::make_value(
-            self.get_months(),
-            self.get_days(),
+            self.months(),
+            self.days(),
             // TODO: this may overflow and we need `try_into`
-            self.get_usecs() * 1000,
+            self.usecs() * 1000,
         )
     }
 }

src/common/src/array/interval_array.rs

@@ -35,17 +35,17 @@ mod tests {
         }
         let ret_arr = array_builder.finish();
         for v in ret_arr.iter().flatten() {
-            assert_eq!(v.get_months(), 12);
-            assert_eq!(v.get_days(), 0);
+            assert_eq!(v.months(), 12);
+            assert_eq!(v.days(), 0);
         }
         let ret_arr = IntervalArray::from_iter([Some(Interval::from_ymd(1, 0, 0)), None]);
         let v = ret_arr.value_at(0).unwrap();
-        assert_eq!(v.get_months(), 12);
-        assert_eq!(v.get_days(), 0);
+        assert_eq!(v.months(), 12);
+        assert_eq!(v.days(), 0);
         let v = ret_arr.value_at(1);
         assert_eq!(v, None);
         let v = unsafe { ret_arr.value_at_unchecked(0).unwrap() };
-        assert_eq!(v.get_months(), 12);
-        assert_eq!(v.get_days(), 0);
+        assert_eq!(v.months(), 12);
+        assert_eq!(v.days(), 0);
     }
 }

src/common/src/types/chrono_wrapper.rs

@@ -525,13 +525,13 @@ impl CheckedAdd<Interval> for Timestamp {
 
     fn checked_add(self, rhs: Interval) -> Option<Timestamp> {
         let mut date = self.0.date();
-        if rhs.get_months() != 0 {
+        if rhs.months() != 0 {
             // NaiveDate don't support add months. We need calculate manually
             let mut day = date.day() as i32;
             let mut month = date.month() as i32;
             let mut year = date.year();
             // Calculate the number of year in this interval
-            let interval_months = rhs.get_months();
+            let interval_months = rhs.months();
             let year_diff = interval_months / 12;
             year += year_diff;
 
@@ -556,8 +556,8 @@ impl CheckedAdd<Interval> for Timestamp {
             date = NaiveDate::from_ymd_opt(year, month as u32, day as u32)?;
         }
         let mut datetime = NaiveDateTime::new(date, self.0.time());
-        datetime = datetime.checked_add_signed(Duration::days(rhs.get_days().into()))?;
-        datetime = datetime.checked_add_signed(Duration::microseconds(rhs.get_usecs()))?;
+        datetime = datetime.checked_add_signed(Duration::days(rhs.days().into()))?;
+        datetime = datetime.checked_add_signed(Duration::microseconds(rhs.usecs()))?;
 
         Some(Timestamp::new(datetime))
     }

src/common/src/types/interval.rs

@@ -57,6 +57,7 @@ impl Interval {
         usecs: i64::MIN,
     };
 
+    /// Creates a new `Interval` from the given number of months, days, and microseconds.
     pub fn from_month_day_usec(months: i32, days: i32, usecs: i64) -> Self {
         Interval {
             months,
@@ -65,22 +66,96 @@ impl Interval {
         }
     }
 
-    pub fn get_days(&self) -> i32 {
-        self.days
+    /// Returns the total number of whole months.
+    ///
+    /// Note the difference between [`months`] and [`months_field`].
+    /// ```
+    /// # use risingwave_common::types::Interval;
+    /// let interval: Interval = "5 yrs 1 month".parse().unwrap();
+    /// assert_eq!(interval.months(), 61);
+    /// assert_eq!(interval.months_field(), 1);
+    /// ```
+    pub fn months(&self) -> i32 {
+        self.months
     }
 
-    pub fn get_months(&self) -> i32 {
-        self.months
+    /// Returns the number of days.
+    pub fn days(&self) -> i32 {
+        self.days
     }
 
-    pub fn get_usecs(&self) -> i64 {
+    /// Returns the total number of microseconds in a day.
+    pub fn usecs(&self) -> i64 {
         self.usecs
     }
 
-    pub fn get_usecs_of_day(&self) -> u64 {
+    /// Calculates the remaining number of microseconds in a day.
+    ///
+    /// Note the difference between [`usecs`] and [`usecs_of_day`].
+    /// ```
+    /// # use risingwave_common::types::Interval;
+    /// let interval: Interval = "-1:00:00".parse().unwrap();
+    /// assert_eq!(interval.usecs(), -1 * 60 * 60 * 1_000_000);
+    /// assert_eq!(interval.usecs_of_day(), 23 * 60 * 60 * 1_000_000);
+    /// ```
+    pub fn usecs_of_day(&self) -> u64 {
         self.usecs.rem_euclid(USECS_PER_DAY) as u64
     }
 
+    /// Returns the years field.
+    pub fn years_field(&self) -> i32 {
+        self.months / 12
+    }
+
+    /// Returns the months field. range: 0-11
+    pub fn months_field(&self) -> i32 {
+        self.months % 12
+    }
+
+    /// Returns the days field.
+    pub fn days_field(&self) -> i32 {
+        self.days
+    }
+
+    /// Returns the hours field. range: -23..=23
+    pub fn hours_field(&self) -> i32 {
+        (self.usecs / USECS_PER_SEC / 3600 % 24) as i32
+    }
+
+    /// Returns the minutes field. range: -59..=-59
+    pub fn minutes_field(&self) -> i32 {
+        (self.usecs / USECS_PER_SEC / 60 % 60) as i32
+    }
+
+    /// Returns the seconds field, including fractional parts, in microseconds.
+    /// range: -59,999,999..=59,999,999
+    pub fn seconds_in_micros(&self) -> i32 {
+        (self.usecs % (USECS_PER_SEC * 60)) as i32
+    }
+
+    /// Returns the microseconds since 1970-01-01 00:00:00+00.
+    pub fn epoch_in_micros(&self) -> i128 {
+        // https://github.com/postgres/postgres/blob/REL_15_2/src/backend/utils/adt/timestamp.c#L5304
+
+        const DAYS_PER_YEAR_X4: i32 = 365 * 4 + 1;
+        const DAYS_PER_MONTH: i32 = 30;
+        const SECS_PER_DAY: i32 = 86400;
+        const MONTHS_PER_YEAR: i32 = 12;
+
+        // To do this calculation in integer arithmetic even though
+        // DAYS_PER_YEAR is fractional, multiply everything by 4 and then
+        // divide by 4 again at the end.  This relies on DAYS_PER_YEAR
+        // being a multiple of 0.25 and on SECS_PER_DAY being a multiple
+        // of 4.
+        let secs_from_day_month = ((DAYS_PER_YEAR_X4 as i64)
+            * (self.months / MONTHS_PER_YEAR) as i64
+            + (4 * DAYS_PER_MONTH as i64) * (self.months % MONTHS_PER_YEAR) as i64
+            + 4 * self.days as i64)
+            * (SECS_PER_DAY / 4) as i64;
+
+        secs_from_day_month as i128 * USECS_PER_SEC as i128 + self.usecs as i128
+    }
+
     pub fn to_protobuf<T: Write>(self, output: &mut T) -> ArrayResult<usize> {
         output.write_i32::<BigEndian>(self.months)?;
         output.write_i32::<BigEndian>(self.days)?;
@@ -1483,9 +1558,9 @@ mod tests {
                 }
                 Some((rhs_months, rhs_days, rhs_usecs, rhs_str)) => {
                     // We should test individual fields rather than using custom `Eq`
-                    assert_eq!(actual_deserialize.unwrap().get_months(), rhs_months);
-                    assert_eq!(actual_deserialize.unwrap().get_days(), rhs_days);
-                    assert_eq!(actual_deserialize.unwrap().get_usecs(), rhs_usecs);
+                    assert_eq!(actual_deserialize.unwrap().months(), rhs_months);
+                    assert_eq!(actual_deserialize.unwrap().days(), rhs_days);
+                    assert_eq!(actual_deserialize.unwrap().usecs(), rhs_usecs);
                     assert_eq!(actual_deserialize.unwrap().to_string(), rhs_str);
                 }
             }
@@ -1495,9 +1570,9 @@ mod tests {
         let input = Interval::from_month_day_usec(i32::MIN, -30, 1);
         let actual_deserialize = IntervalCmpValue::from(input).as_justified();
         // It has a justified interval within range, and can be obtained by our deserialization.
-        assert_eq!(actual_deserialize.unwrap().get_months(), i32::MIN);
-        assert_eq!(actual_deserialize.unwrap().get_days(), -29);
-        assert_eq!(actual_deserialize.unwrap().get_usecs(), -USECS_PER_DAY + 1);
+        assert_eq!(actual_deserialize.unwrap().months(), i32::MIN);
+        assert_eq!(actual_deserialize.unwrap().days(), -29);
+        assert_eq!(actual_deserialize.unwrap().usecs(), -USECS_PER_DAY + 1);
     }
 
     #[test]

src/common/src/util/value_encoding/mod.rs

@@ -293,9 +293,9 @@ fn estimate_serialize_str_size(bytes: &[u8]) -> usize {
 }
 
 fn serialize_interval(interval: &Interval, buf: &mut impl BufMut) {
-    buf.put_i32_le(interval.get_months());
-    buf.put_i32_le(interval.get_days());
-    buf.put_i64_le(interval.get_usecs());
+    buf.put_i32_le(interval.months());
+    buf.put_i32_le(interval.days());
+    buf.put_i64_le(interval.usecs());
 }
 
 fn estimate_serialize_interval_size() -> usize {

src/expr/src/vector_op/arithmetic_op.rs

@@ -234,14 +234,14 @@ pub fn interval_timestamptz_add(l: Interval, r: i64) -> Result<i64> {
 #[inline(always)]
 fn timestamptz_interval_inner(l: i64, r: Interval, f: fn(i64, i64) -> Option<i64>) -> Result<i64> {
     // Without session TimeZone, we cannot add month/day in local time. See #5826.
-    if r.get_months() != 0 || r.get_days() != 0 {
+    if r.months() != 0 || r.days() != 0 {
         return Err(ExprError::UnsupportedFunction(
             "timestamp with time zone +/- interval of days".into(),
         ));
     }
 
     let result: Option<i64> = try {
-        let delta_usecs = r.get_usecs();
+        let delta_usecs = r.usecs();
         f(l, delta_usecs)?
     };
 
@@ -280,7 +280,7 @@ pub fn time_time_sub(l: Time, r: Time) -> Result<Interval> {
 #[function("subtract(time, interval) -> time")]
 pub fn time_interval_sub(l: Time, r: Interval) -> Result<Time> {
     let time = l.0;
-    let (new_time, ignored) = time.overflowing_sub_signed(Duration::microseconds(r.get_usecs()));
+    let (new_time, ignored) = time.overflowing_sub_signed(Duration::microseconds(r.usecs()));
     if ignored == 0 {
         Ok(Time::new(new_time))
     } else {
@@ -291,7 +291,7 @@ pub fn time_interval_sub(l: Time, r: Interval) -> Result<Time> {
 #[function("add(time, interval) -> time")]
 pub fn time_interval_add(l: Time, r: Interval) -> Result<Time> {
     let time = l.0;
-    let (new_time, ignored) = time.overflowing_add_signed(Duration::microseconds(r.get_usecs()));
+    let (new_time, ignored) = time.overflowing_add_signed(Duration::microseconds(r.usecs()));
     if ignored == 0 {
         Ok(Time::new(new_time))
     } else {

src/expr/src/vector_op/cast.rs

@@ -384,7 +384,7 @@ pub fn timestamp_to_time(elem: Timestamp) -> Time {
 /// In `PostgreSQL`, casting from interval to time discards the days part.
 #[function("cast(interval) -> time")]
 pub fn interval_to_time(elem: Interval) -> Time {
-    let usecs = elem.get_usecs_of_day();
+    let usecs = elem.usecs_of_day();
     let secs = (usecs / 1_000_000) as u32;
     let nano = (usecs % 1_000_000 * 1000) as u32;
     Time::from_num_seconds_from_midnight_uncheck(secs, nano)