Merge #699

699: state_agg API tweaks r=Smittyvb a=Smittyvb - `interpolated_duration_in`/`duration_in`/`interpolated_state_periods`/`state_periods` have the first two arguments swapped: now the aggregate is first and the state is second - `into_values`/`into_int_values` now returns a table with intervals instead of microseconds Co-authored-by: Smitty <smitty@timescale.com>
timescale · Feb 9, 2023 · f80b630 · f80b630
2 parents cf27d8d + c6fab48
commit f80b630
Show file tree

Hide file tree

Showing 4 changed files with 210 additions and 211 deletions.
diff --git a/Changelog.md b/Changelog.md
@@ -17,6 +17,8 @@ This changelog should be updated as part of a PR if the work is worth noting (mo
 - [#692](https://github.com/timescale/timescaledb-toolkit/pull/692): Support specifying a range to `duration_in` to specify a time range to get states in for state aggregates
 - [#692](https://github.com/timescale/timescaledb-toolkit/pull/692): Removed `next` parameter from interpolated state aggregate functions
 - [#692](https://github.com/timescale/timescaledb-toolkit/pull/692): Renamed `state_agg` to `compact_state_agg` and `timeline_agg` to `state_agg`
+- [#699](https://github.com/timescale/timescaledb-toolkit/pull/699): `interpolated_duration_in`/`duration_in`/`interpolated_state_periods`/`state_periods` have the first two arguments swapped: now the aggregate is first and the state is second
+- [#699](https://github.com/timescale/timescaledb-toolkit/pull/699): `into_values`/`into_int_values` now returns a table with intervals instead of microseconds
 
 #### Shout-outs
 

diff --git a/docs/state_agg.md b/docs/state_agg.md
@@ -46,15 +46,15 @@ INSERT INTO states_test_5 VALUES
 Compute the amount of time spent in a state as INTERVAL.
 
 ```SQL
-SELECT toolkit_experimental.duration_in('ERROR', toolkit_experimental.compact_state_agg(ts, state)) FROM states_test;
+SELECT toolkit_experimental.duration_in(toolkit_experimental.compact_state_agg(ts, state), 'ERROR') FROM states_test;
 ```
 ```output
  interval
 ----------
  00:00:03
 ```
 ```SQL
-SELECT toolkit_experimental.duration_in(2, toolkit_experimental.compact_state_agg(ts, state)) FROM states_test_4;
+SELECT toolkit_experimental.duration_in(toolkit_experimental.compact_state_agg(ts, state), 2) FROM states_test_4;
 ```
 ```output
  interval
@@ -67,7 +67,7 @@ Extract as number of seconds:
 ```SQL
 SELECT
   EXTRACT(epoch FROM
-    toolkit_experimental.duration_in('ERROR', toolkit_experimental.compact_state_agg(ts, state))
+    toolkit_experimental.duration_in(toolkit_experimental.compact_state_agg(ts, state), 'ERROR')
   )::INTEGER
 FROM states_test;
 ```
@@ -79,39 +79,39 @@ FROM states_test;
 
 #### duration_in for a range
 ```SQL
-SELECT toolkit_experimental.duration_in('OK', toolkit_experimental.state_agg(ts, state), '2020-01-01 00:01:00+00', '2 days') FROM states_test;
+SELECT toolkit_experimental.duration_in(toolkit_experimental.state_agg(ts, state), 'OK', '2020-01-01 00:01:00+00', '2 days') FROM states_test;
 ```
 ```output
  duration_in
 -------------
  00:00:57
 ```
 ```SQL
-SELECT toolkit_experimental.duration_in('OK', toolkit_experimental.state_agg(ts, state), '2020-01-01 00:01:00+00', NULL) FROM states_test;
+SELECT toolkit_experimental.duration_in(toolkit_experimental.state_agg(ts, state), 'OK', '2020-01-01 00:01:00+00', NULL) FROM states_test;
 ```
 ```output
  duration_in
 -------------
  00:00:57
 ```
 ```SQL
-SELECT toolkit_experimental.duration_in('OK', toolkit_experimental.state_agg(ts, state), '2020-01-01 00:01:00+00') FROM states_test;
+SELECT toolkit_experimental.duration_in(toolkit_experimental.state_agg(ts, state), 'OK', '2020-01-01 00:01:00+00') FROM states_test;
 ```
 ```output
  duration_in
 -------------
  00:00:57
 ```
 ```SQL
-SELECT toolkit_experimental.duration_in(51351, toolkit_experimental.state_agg(ts, state), '2020-01-01 00:01:00+00', '2 days') FROM states_test_4;
+SELECT toolkit_experimental.duration_in(toolkit_experimental.state_agg(ts, state), 51351, '2020-01-01 00:01:00+00', '2 days') FROM states_test_4;
 ```
 ```output
  duration_in
 -------------
  00:00:57
 ```
 ```SQL
-SELECT toolkit_experimental.duration_in(51351, toolkit_experimental.state_agg(ts, state), '2020-01-01 00:01:00+00', NULL) FROM states_test_4;
+SELECT toolkit_experimental.duration_in(toolkit_experimental.state_agg(ts, state), 51351, '2020-01-01 00:01:00+00', NULL) FROM states_test_4;
 ```
 ```output
  duration_in
@@ -120,7 +120,7 @@ SELECT toolkit_experimental.duration_in(51351, toolkit_experimental.state_agg(ts
 ```
 
 ```SQL
-SELECT toolkit_experimental.duration_in('OK', toolkit_experimental.state_agg(ts, state), '2020-01-01 00:00:15+00', '30 seconds') FROM states_test;
+SELECT toolkit_experimental.duration_in(toolkit_experimental.state_agg(ts, state), 'OK', '2020-01-01 00:00:15+00', '30 seconds') FROM states_test;
 ```
 ```output
  duration_in
@@ -129,7 +129,7 @@ SELECT toolkit_experimental.duration_in('OK', toolkit_experimental.state_agg(ts,
 ```
 
 ```SQL
-SELECT toolkit_experimental.duration_in('OK', toolkit_experimental.state_agg(ts, state), '2020-01-01 00:00:15+00', '1 minute 1 second') FROM states_test;
+SELECT toolkit_experimental.duration_in(toolkit_experimental.state_agg(ts, state), 'OK', '2020-01-01 00:00:15+00', '1 minute 1 second') FROM states_test;
 ```
 ```output
  duration_in
@@ -147,23 +147,23 @@ SELECT state, duration FROM toolkit_experimental.into_values(
 ```output
  state | duration
 -------+-----------
- ERROR |   3000000
- OK    | 106000000
- START |  11000000
- STOP  |         0
+ ERROR |  00:00:03
+ OK    |  00:01:46
+ START |  00:00:11
+ STOP  |  00:00:00
 ```
 ```SQL
 SELECT state, duration FROM toolkit_experimental.into_int_values(
-    (SELECT toolkit_experimental.compact_state_agg(ts, state) FROM states_test_4))
+    (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test_4))
     ORDER BY state, duration;
 ```
 ```output
  state | duration
 -------+-----------
-   -9 |         0
-    2 |   3000000
-    4 |  11000000
-51351 | 106000000
+   -9 |  00:00:00
+    2 |  00:00:03
+    4 |  00:00:11
+51351 |  00:01:46
 ```
 
 ### state_timeline
@@ -217,8 +217,8 @@ START | 2019-12-31 00:00:00+00 | 2019-12-31 00:00:11+00
 ```SQL
 SELECT start_time, end_time
 FROM toolkit_experimental.state_periods(
-    'OK',
-    (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test)
+    (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test),
+    'OK'
 )
 ORDER BY start_time;
 ```
@@ -232,8 +232,8 @@ start_time             | end_time
 ```SQL
 SELECT start_time, end_time
 FROM toolkit_experimental.state_periods(
-    51351,
-    (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test_4)
+    (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test_4),
+    51351
 )
 ORDER BY start_time;
 ```
@@ -247,8 +247,8 @@ start_time             | end_time
 ```SQL
 SELECT start_time, end_time
 FROM toolkit_experimental.state_periods(
-    'ANYTHING',
-    (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test)
+    (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test),
+    'ANYTHING'
 )
 ORDER BY start_time;
 ```
@@ -338,8 +338,8 @@ ERROR | 2020-01-01 00:01:00+00 | 2020-01-01 00:01:03+00
 
 ```SQL
 SELECT start_time, end_time FROM toolkit_experimental.interpolated_state_periods(
-    'OK',
     (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test),
+    'OK',
     '2019-12-31', '1 days',
     (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test_3)
 )
@@ -354,8 +354,8 @@ start_time             | end_time
 
 ```SQL
 SELECT start_time, end_time FROM toolkit_experimental.interpolated_state_periods(
-    'START',
     (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test),
+    'START',
     '2019-12-31', '5 days',
     (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test_3)
 )
@@ -369,8 +369,8 @@ start_time             | end_time
 
 ```SQL
 SELECT start_time, end_time FROM toolkit_experimental.interpolated_state_periods(
-    'STOP',
     (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test),
+    'STOP',
     '2019-12-31', '1 days',
     (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test_2)
 )
@@ -385,8 +385,8 @@ start_time             | end_time
 
 ```SQL
 SELECT start_time, end_time FROM toolkit_experimental.interpolated_state_periods(
-    'STOP',
     (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test),
+    'STOP',
     '2019-12-31', '5 days',
     (SELECT toolkit_experimental.state_agg(ts, state) FROM states_test_2)
 )
@@ -408,8 +408,8 @@ WITH buckets AS (SELECT
 FROM states_test
 GROUP BY date_trunc('minute', ts))
 SELECT toolkit_experimental.duration_in(
-    'START',
-    toolkit_experimental.rollup(buckets.sa)
+    toolkit_experimental.rollup(buckets.sa),
+    'START'
 )
 FROM buckets;
 ```
@@ -426,8 +426,8 @@ WITH buckets AS (SELECT
 FROM states_test
 GROUP BY date_trunc('minute', ts))
 SELECT toolkit_experimental.duration_in(
-    'OK',
-    toolkit_experimental.rollup(buckets.sa)
+    toolkit_experimental.rollup(buckets.sa),
+    'OK'
 )
 FROM buckets;
 ```

diff --git a/extension/src/raw.rs b/extension/src/raw.rs
@@ -128,12 +128,27 @@ impl From<i64> for Interval {
             time: interval,
             ..Default::default()
         };
-        unsafe {
+        let interval = unsafe {
             let ptr =
                 pg_sys::palloc(std::mem::size_of::<pg_sys::Interval>()) as *mut pg_sys::Interval;
             *ptr = interval;
             Interval(pg_sys::Datum::from(ptr))
-        }
+        };
+        // Now we have a valid Interval in at least one sense.  But we have the
+        // microseconds in the `time` field and `day` and `month` are both 0,
+        // which is legal.  However, directly converting one of these to TEXT
+        // comes out quite ugly if the number of microseconds is greater than 1 day:
+        //   8760:02:00
+        // Should be:
+        //   365 days 00:02:00
+        // How does postgresql do it?  It happens in src/backend/utils/adt/timestamp.c:timestamp_mi:
+        //  result->time = dt1 - dt2;
+        //  result = DatumGetIntervalP(DirectFunctionCall1(interval_justify_hours,
+        //                                                 IntervalPGetDatum(result)));
+        // So if we want the same behavior, we need to call interval_justify_hours too:
+        let function_args = vec![Some(pg_sys::Datum::from(interval))];
+        unsafe { pgx::direct_function_call(pg_sys::interval_justify_hours, function_args) }
+            .expect("interval_justify_hours does not return None")
     }
 }