SQL: Extend DATE_TRUNC to also operate on intervals(elastic - #46632 ) (#47720)

The function is extended to operate on intervals according to the PostgreSQL: https://www.postgresql.org/docs/9.1/functions-datetime.html#FUNCTIONS-DATETIME-TRUNC

Closes : #46632

Author: musteaf

docs/reference/sql/functions/date-time.asciidoc

@@ -500,18 +500,19 @@ include-tagged::{sql-specs}/docs/docs.csv-spec[datePartDateTimeTzOffsetMinus]
 --------------------------------------------------
 DATE_TRUNC(
     string_exp, <1>
-    datetime_exp) <2>
+    datetime_exp/interval_exp) <2>
 --------------------------------------------------
 
 *Input*:
 
-<1> string expression denoting the unit to which the date/datetime should be truncated to
-<2> date/datetime expression
+<1> string expression denoting the unit to which the date/datetime/interval should be truncated to
+<2> date/datetime/interval expression
 
-*Output*: datetime
+*Output*: datetime/interval
 
-*Description*: Truncate the date/datetime to the specified unit by setting all fields that are less significant than the specified
+*Description*: Truncate the date/datetime/interval to the specified unit by setting all fields that are less significant than the specified
 one to zero (or one, for day, day of week and month). If any of the two arguments is `null` a `null` is returned.
+If the first argument is `week` and the second argument is of `interval` type, an error is thrown since the `interval` data type doesn't support a `week` time unit.
 
 [cols="^,^"]
 |===
@@ -563,6 +564,21 @@ include-tagged::{sql-specs}/docs/docs.csv-spec[truncateDateDecades]
 include-tagged::{sql-specs}/docs/docs.csv-spec[truncateDateQuarter]
 --------------------------------------------------
 
+[source, sql]
+--------------------------------------------------
+include-tagged::{sql-specs}/docs/docs.csv-spec[truncateIntervalCenturies]
+--------------------------------------------------
+
+[source, sql]
+--------------------------------------------------
+include-tagged::{sql-specs}/docs/docs.csv-spec[truncateIntervalHour]
+--------------------------------------------------
+
+[source, sql]
+--------------------------------------------------
+include-tagged::{sql-specs}/docs/docs.csv-spec[truncateIntervalDay]
+--------------------------------------------------
+
 [[sql-functions-datetime-day]]
 ==== `DAY_OF_MONTH/DOM/DAY`
 

x-pack/plugin/ql/src/main/java/org/elasticsearch/xpack/ql/util/DateUtils.java

@@ -34,7 +34,7 @@ public class DateUtils {
     public static final ZoneId UTC = ZoneId.of("Z");
 
     public static final String EMPTY = "";
-    
+
     public static final DateTimeFormatter ISO_DATE_WITH_MILLIS = new DateTimeFormatterBuilder()
             .parseCaseInsensitive()
             .append(ISO_LOCAL_DATE)
@@ -72,17 +72,17 @@ public class DateUtils {
             .appendOffsetId()
             .toFormatter(Locale.ROOT);
 
-    private static final int SECONDS_PER_MINUTE = 60;
-    private static final int SECONDS_PER_HOUR = SECONDS_PER_MINUTE * 60;
-    private static final int SECONDS_PER_DAY = SECONDS_PER_HOUR * 24;
+    public static final int SECONDS_PER_MINUTE = 60;
+    public static final int SECONDS_PER_HOUR = SECONDS_PER_MINUTE * 60;
+    public static final int SECONDS_PER_DAY = SECONDS_PER_HOUR * 24;
 
     private DateUtils() {}
 
     public static String toString(Object value) {
         if (value == null) {
             return "null";
         }
-        
+
         if (value instanceof ZonedDateTime) {
             return ((ZonedDateTime) value).format(ISO_DATE_WITH_MILLIS);
         }

x-pack/plugin/sql/qa/src/main/resources/datetime.csv-spec

@@ -499,6 +499,16 @@ DATE_TRUNC('week', '2019-09-04'::date) as dt_week,  DATE_TRUNC('day', '2019-09-0
 2000-01-01T00:00:00.000Z | 2000-01-01T00:00:00.000Z | 2010-01-01T00:00:00.000Z | 2019-01-01T00:00:00.000Z | 2019-07-01T00:00:00.000Z | 2019-09-01T00:00:00.000Z | 2019-09-02T00:00:00.000Z | 2019-09-04T00:00:00.000Z
 ;
 
+selectDateTruncWithInterval
+SELECT DATE_TRUNC('hour', INTERVAL '1 12:43:21' DAY TO SECONDS) as dt_hour, DATE_TRUNC('minute', INTERVAL '1 12:43:21' DAY TO SECONDS) as dt_min,
+DATE_TRUNC('seconds', INTERVAL '1 12:43:21' DAY TO SECONDS) as dt_sec, DATE_TRUNC('ms', INTERVAL '1 12:43:21' DAY TO SECONDS)::string as dt_millis,
+DATE_TRUNC('mcs', INTERVAL '1 12:43:21' DAY TO SECONDS)::string as dt_micro, DATE_TRUNC('nanoseconds', INTERVAL '1 12:43:21' DAY TO SECONDS)::string as dt_nano;
+
+    dt_hour    |    dt_min     |    dt_sec     |   dt_millis   |   dt_micro    |    dt_nano
+---------------+---------------+---------------+---------------+---------------+---------------
++1 12:00:00    |+1 12:43:00    |+1 12:43:21    |+1 12:43:21    |+1 12:43:21    |+1 12:43:21
+;
+
 selectDateTruncWithField
 schema::emp_no:i|birth_date:ts|dt_mil:ts|dt_cent:ts|dt_dec:ts|dt_year:ts|dt_quarter:ts|dt_month:ts|dt_week:ts|dt_day:ts
 SELECT emp_no, birth_date, DATE_TRUNC('millennium', birth_date) as dt_mil, DATE_TRUNC('centuries', birth_date) as dt_cent,
@@ -585,6 +595,21 @@ SELECT emp_no, hire_date, DATE_TRUNC('quarter', hire_date) as dt FROM test_emp O
 10076   | 1985-07-09 00:00:00.000Z | 1985-07-01 00:00:00.000Z
 ;
 
+dateTruncOrderByWithInterval
+schema::first_name:s|dt:ts|hire_date:ts|languages:byte
+SELECT first_name, hire_date + DATE_TRUNC('centuries', CASE WHEN languages = 5 THEN INTERVAL '18-3' YEAR TO MONTH
+WHEN languages = 4 THEN INTERVAL '108-4' YEAR TO MONTH WHEN languages = 3 THEN INTERVAL '212-3' YEAR TO MONTH
+ELSE INTERVAL '318-6' YEAR TO MONTH END) as dt, hire_date, languages FROM test_emp WHERE emp_no <= 10006 ORDER BY dt NULLS LAST LIMIT 5;
+
+   first_name |            dt            |        hire_date         | languages
+--------------+--------------------------+--------------------------+-----------
+Bezalel       | 1985-11-21 00:00:00.000Z | 1985-11-21T00:00:00.000Z | 5
+Chirstian     | 1986-12-01 00:00:00.000Z | 1986-12-01T00:00:00.000Z | 5
+Parto         | 2086-08-28 00:00:00.000Z | 1986-08-28T00:00:00.000Z | 4
+Anneke        | 2189-06-02 00:00:00.000Z | 1989-06-02T00:00:00.000Z | 3
+Georgi        | 2286-06-26 00:00:00.000Z | 1986-06-26T00:00:00.000Z | 2
+;
+
 dateTruncFilter
 schema::emp_no:i|hire_date:ts|dt:ts
 SELECT emp_no, hire_date, DATE_TRUNC('quarter', hire_date) as dt FROM test_emp WHERE DATE_TRUNC('quarter', hire_date) > '1994-07-01T00:00:00.000Z'::timestamp ORDER BY emp_no;
@@ -601,6 +626,24 @@ SELECT emp_no, hire_date, DATE_TRUNC('quarter', hire_date) as dt FROM test_emp W
 10093   | 1996-11-05 00:00:00.000Z | 1996-10-01 00:00:00.000Z
 ;
 
+dateTruncFilterWithInterval
+schema::first_name:s|hire_date:ts
+SELECT first_name, hire_date FROM test_emp WHERE hire_date > '2090-03-05T10:11:22.123Z'::datetime - DATE_TRUNC('centuries', INTERVAL 190 YEARS) ORDER BY first_name DESC, hire_date ASC LIMIT 10;
+
+  first_name   |          hire_date
+---------------+-------------------------
+null           | 1990-06-20 00:00:00.000Z
+null           | 1990-12-05 00:00:00.000Z
+null           | 1991-09-01 00:00:00.000Z
+null           | 1992-01-03 00:00:00.000Z
+null           | 1994-02-17 00:00:00.000Z
+Yongqiao       | 1995-03-20 00:00:00.000Z
+Yishay         | 1990-10-20 00:00:00.000Z
+Yinghua        | 1990-12-25 00:00:00.000Z
+Weiyi          | 1993-02-14 00:00:00.000Z
+Tuval          | 1995-12-15 00:00:00.000Z
+;
+
 dateTruncGroupBy
 schema::count:l|dt:ts
 SELECT count(*) as count, DATE_TRUNC('decade', hire_date) dt FROM test_emp GROUP BY dt ORDER BY 2;
@@ -611,6 +654,24 @@ SELECT count(*) as count, DATE_TRUNC('decade', hire_date) dt FROM test_emp GROUP
 41      | 1990-01-01 00:00:00.000Z
 ;
 
+dateTruncGroupByWithInterval
+schema::count:l|dt:ts
+SELECT count(*) as count, birth_date + DATE_TRUNC('hour', INTERVAL '1 12:43:21' DAY TO SECONDS) dt FROM test_emp GROUP BY dt ORDER BY 2 LIMIT 10;
+
+ count  |          dt
+--------+-------------------------
+10      | null
+1       | 1952-02-28 12:00:00.000Z
+1       | 1952-04-20 12:00:00.000Z
+1       | 1952-05-16 12:00:00.000Z
+1       | 1952-06-14 12:00:00.000Z
+1       | 1952-07-09 12:00:00.000Z
+1       | 1952-08-07 12:00:00.000Z
+1       | 1952-11-14 12:00:00.000Z
+1       | 1952-12-25 12:00:00.000Z
+1       | 1953-01-08 12:00:00.000Z
+;
+
 dateTruncHaving
 schema::gender:s|dt:ts
 SELECT gender, max(hire_date) AS dt FROM test_emp GROUP BY gender HAVING DATE_TRUNC('year', max(hire_date)) >= '1997-01-01T00:00:00.000Z'::timestamp ORDER BY 1;
@@ -621,6 +682,16 @@ null    | 1999-04-30 00:00:00.000Z
 F       | 1997-05-19 00:00:00.000Z
 ;
 
+// Awaits fix: https://github.com/elastic/elasticsearch/issues/53565
+dateTruncHavingWithInterval-Ignore
+schema::gender:s|dt:ts
+SELECT gender, max(hire_date) AS dt FROM test_emp GROUP BY gender HAVING max(hire_date) - DATE_TRUNC('hour', INTERVAL 1 YEARS) >= '1997-01-01T00:00:00.000Z'::timestamp ORDER BY 1;
+
+ gender |         dt
+--------+-------------------------
+null    | 1999-04-30 00:00:00.000Z
+;
+
 selectDatePartWithDate
 SELECT DATE_PART('year', '2019-09-04'::date) as dp_years, DATE_PART('quarter', '2019-09-04'::date) as dp_quarter, DATE_PART('month', '2019-09-04'::date) as dp_month,
 DATE_PART('dayofyear', '2019-09-04'::date) as dp_doy, DATE_PART('day', '2019-09-04'::date) as dp_day, DATE_PART('week', '2019-09-04'::date) as dp_week,

x-pack/plugin/sql/qa/src/main/resources/docs/docs.csv-spec

@@ -2675,6 +2675,36 @@ SELECT DATETRUNC('quarters', CAST('2019-09-04' AS DATE)) AS quarter;
 // end::truncateDateQuarter
 ;
 
+truncateIntervalCenturies
+// tag::truncateIntervalCenturies
+SELECT DATE_TRUNC('centuries', INTERVAL '199-5' YEAR TO MONTH) AS centuries;
+
+      centuries
+------------------
+ +100-0
+// end::truncateIntervalCenturies
+;
+
+truncateIntervalHour
+// tag::truncateIntervalHour
+SELECT DATE_TRUNC('hours', INTERVAL '17 22:13:12' DAY TO SECONDS) AS hour;
+
+      hour
+------------------
++17 22:00:00
+// end::truncateIntervalHour
+;
+
+truncateIntervalDay
+// tag::truncateIntervalDay
+SELECT DATE_TRUNC('days', INTERVAL '19 15:24:19' DAY TO SECONDS) AS day;
+
+      day
+------------------
++19 00:00:00
+// end::truncateIntervalDay
+;
+
 constantDayOfWeek
 // tag::dayOfWeek
 SELECT DAY_OF_WEEK(CAST('2018-02-19T10:23:27Z' AS TIMESTAMP)) AS day;

x-pack/plugin/sql/src/main/java/org/elasticsearch/xpack/sql/expression/SqlTypeResolutions.java

@@ -16,7 +16,7 @@
 public final class SqlTypeResolutions {
 
     private SqlTypeResolutions() {}
-    
+
     public static TypeResolution isDate(Expression e, String operationName, ParamOrdinal paramOrd) {
         return isType(e, SqlDataTypes::isDateBased, operationName, paramOrd, "date", "datetime");
     }
@@ -25,6 +25,10 @@ public static TypeResolution isDateOrTime(Expression e, String operationName, Pa
         return isType(e, SqlDataTypes::isDateOrTimeBased, operationName, paramOrd, "date", "time", "datetime");
     }
 
+    public static TypeResolution isDateOrInterval(Expression e, String operationName, ParamOrdinal paramOrd) {
+        return isType(e, SqlDataTypes::isDateOrIntervalBased, operationName, paramOrd, "date", "datetime", "an interval data type");
+    }
+
     public static TypeResolution isNumericOrDate(Expression e, String operationName, ParamOrdinal paramOrd) {
         return isType(e, dt -> dt.isNumeric() || SqlDataTypes.isDateBased(dt), operationName, paramOrd,
             "date", "datetime", "numeric");

x-pack/plugin/sql/src/main/java/org/elasticsearch/xpack/sql/expression/function/scalar/datetime/BinaryDateTimeFunction.java

@@ -19,7 +19,6 @@
 import static org.elasticsearch.common.logging.LoggerMessageFormat.format;
 import static org.elasticsearch.xpack.ql.expression.TypeResolutions.isString;
 import static org.elasticsearch.xpack.ql.expression.gen.script.ParamsBuilder.paramsBuilder;
-import static org.elasticsearch.xpack.sql.expression.SqlTypeResolutions.isDate;
 
 public abstract class BinaryDateTimeFunction extends BinaryScalarFunction {
 
@@ -54,10 +53,7 @@ protected TypeResolution resolveType() {
                 }
             }
         }
-        resolution = isDate(right(), sourceText(), Expressions.ParamOrdinal.SECOND);
-        if (resolution.unresolved()) {
-            return resolution;
-        }
+
         return TypeResolution.TYPE_RESOLVED;
     }
 

x-pack/plugin/sql/src/main/java/org/elasticsearch/xpack/sql/expression/function/scalar/datetime/DatePart.java

@@ -6,6 +6,7 @@
 package org.elasticsearch.xpack.sql.expression.function.scalar.datetime;
 
 import org.elasticsearch.xpack.ql.expression.Expression;
+import org.elasticsearch.xpack.ql.expression.Expressions;
 import org.elasticsearch.xpack.ql.expression.function.scalar.BinaryScalarFunction;
 import org.elasticsearch.xpack.ql.expression.gen.pipeline.Pipe;
 import org.elasticsearch.xpack.ql.tree.NodeInfo;
@@ -23,6 +24,8 @@
 import java.util.Set;
 import java.util.function.ToIntFunction;
 
+import static org.elasticsearch.xpack.sql.expression.SqlTypeResolutions.isDate;
+
 public class DatePart extends BinaryDateTimeFunction {
 
     public enum Part implements DateTimeField {
@@ -84,6 +87,19 @@ public DataType dataType() {
         return DataTypes.INTEGER;
     }
 
+    @Override
+    protected TypeResolution resolveType() {
+        TypeResolution resolution = super.resolveType();
+        if (resolution.unresolved()) {
+            return resolution;
+        }
+        resolution = isDate(right(), sourceText(), Expressions.ParamOrdinal.SECOND);
+        if (resolution.unresolved()) {
+            return resolution;
+        }
+        return TypeResolution.TYPE_RESOLVED;
+    }
+
     @Override
     protected BinaryScalarFunction replaceChildren(Expression newDateTimePart, Expression newTimestamp) {
         return new DatePart(source(), newDateTimePart, newTimestamp, zoneId());