fix: push down date range filters wrapped by redundant date conversions (both operands, v2 + Calcite)

integ-test/src/test/java/org/opensearch/sql/ppl/CastFunctionIT.java

@@ -5,6 +5,7 @@
 
 package org.opensearch.sql.ppl;
 
+import static org.junit.Assert.assertEquals;
 import static org.opensearch.sql.legacy.TestsConstants.TEST_INDEX_DATATYPE_NONNUMERIC;
 import static org.opensearch.sql.legacy.TestsConstants.TEST_INDEX_DATATYPE_NUMERIC;
 import static org.opensearch.sql.legacy.TestsConstants.TEST_INDEX_DATE_FORMATS;
@@ -19,6 +20,7 @@
 import static org.opensearch.sql.util.MatcherUtils.verifySchema;
 
 import java.io.IOException;
+import java.util.List;
 import java.util.Locale;
 import org.json.JSONObject;
 import org.junit.Test;
@@ -471,4 +473,34 @@ public void testCastDoubleAsString() throws IOException {
     verifySchema(actual, schema("s", "string"));
     verifyDataRows(actual, rows("0.0"));
   }
+
+  /**
+   * A redundant date/time cast on the filtered field changes only how the predicate is executed
+   * (native range query instead of a per-document script), never which rows match. Compare the
+   * wrapped forms against the bare-field form to keep that guarantee enforced.
+   */
+  @Test
+  public void testRedundantDateCastOnFilteredFieldDoesNotChangeRows() throws IOException {
+    String template =
+        "source=%s | where %s | sort strict_date_optional_time | fields strict_date_optional_time";
+    JSONObject bare =
+        executeQuery(
+            String.format(
+                Locale.ROOT,
+                template,
+                TEST_INDEX_DATE_FORMATS,
+                "strict_date_optional_time >= '1984-04-12 09:07:42'"));
+
+    for (String wrapped :
+        List.of(
+            "timestamp(strict_date_optional_time) >= timestamp('1984-04-12 09:07:42')",
+            "cast(strict_date_optional_time as timestamp) >= timestamp('1984-04-12 09:07:42')")) {
+      JSONObject actual =
+          executeQuery(String.format(Locale.ROOT, template, TEST_INDEX_DATE_FORMATS, wrapped));
+      assertEquals(
+          "wrapping the filtered field in a redundant date cast must not change the result",
+          bare.getJSONArray("datarows").toString(),
+          actual.getJSONArray("datarows").toString());
+    }
+  }
 }

integ-test/src/test/java/org/opensearch/sql/ppl/ExplainIT.java

@@ -76,6 +76,54 @@ public void testFilterByCompareStringTimestampPushDownExplain() throws IOExcepti
                 + "| where birthdate < '2018-11-09 00:00:00.000000000' "));
   }
 
+  /**
+   * Wrapping an already timestamp-typed field in timestamp() is a no-op, so the comparison must
+   * still push down to a native range query instead of falling back to a per-document script.
+   */
+  @Test
+  public void testFilterTimestampWrappedFieldPushDownExplain() throws IOException {
+    String expected = loadExpectedPlan("explain_filter_push_timestamp_wrapped_field.yaml");
+    assertYamlEqualsIgnoreId(
+        expected,
+        explainQueryYaml(
+            "source=opensearch-sql_test_index_bank"
+                + "| where timestamp(birthdate) > cast('2016-12-08 00:00:00' as timestamp) "
+                + "| where timestamp(birthdate) < cast('2018-11-09 00:00:00' as timestamp) "));
+  }
+
+  /**
+   * `field > timestamp('...')` is the shape clients emit for a time-range bound (for example the
+   * Grafana OpenSearch data source's PPL time filter). PPL coerces the string argument first, so
+   * the bound arrives as timestamp(cast_to_timestamp('...')); applying timestamp() to a value that
+   * is already a timestamp is a no-op, so the comparison must still push down to a native range
+   * query.
+   */
+  @Test
+  public void testFilterTimestampWrappedBoundPushDownExplain() throws IOException {
+    String expected = loadExpectedPlan("explain_filter_push_timestamp_wrapped_bound.yaml");
+    assertYamlEqualsIgnoreId(
+        expected,
+        explainQueryYaml(
+            "source=opensearch-sql_test_index_bank"
+                + "| where birthdate > timestamp('2016-12-08 00:00:00') "
+                + "| where birthdate < timestamp('2018-11-09 00:00:00') "));
+  }
+
+  /**
+   * last_day() takes a single date argument and returns a date, but it changes the value, so it is
+   * not a redundant conversion and must not be folded to the bare field -- it stays on the script
+   * path rather than becoming a range query.
+   */
+  @Test
+  public void testFilterLastDayOverDateFieldNoPushDownExplain() throws IOException {
+    String expected = loadExpectedPlan("explain_filter_last_day_no_push.yaml");
+    assertYamlEqualsIgnoreId(
+        expected,
+        explainQueryYaml(
+            "source=opensearch-sql_test_index_date_formats | fields yyyy-MM-dd"
+                + "| where last_day(yyyy-MM-dd) = date('2018-11-30') "));
+  }
+
   @Test
   public void testFilterByCompareStringDatePushDownExplain() throws IOException {
     String expected = loadExpectedPlan("explain_filter_push_compare_date_string.yaml");

integ-test/src/test/resources/expectedOutput/calcite/explain_filter_last_day_no_push.yaml

@@ -0,0 +1,8 @@
+calcite:
+  logical: |
+    LogicalSystemLimit(fetch=[10000], type=[QUERY_SIZE_LIMIT])
+      LogicalFilter(condition=[=(LAST_DAY($0), DATE('2018-11-30':VARCHAR))])
+        LogicalProject(yyyy-MM-dd=[$83])
+          CalciteLogicalIndexScan(table=[[OpenSearch, opensearch-sql_test_index_date_formats]])
+  physical: |
+    CalciteEnumerableIndexScan(table=[[OpenSearch, opensearch-sql_test_index_date_formats]], PushDownContext=[[PROJECT->[yyyy-MM-dd], SCRIPT->=(LAST_DAY($0), '2018-11-30'), LIMIT->10000], OpenSearchRequestBuilder(sourceBuilder={"from":0,"size":10000,"timeout":"1m","query":{"script":{"script":{"source":"{\"langType\":\"calcite\",\"script\":\"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\"}","lang":"opensearch_compounded_script","params":{"utcTimestamp": 0,"SOURCES":[0,2],"DIGESTS":["yyyy-MM-dd","2018-11-30"]}},"boost":1.0}},"_source":{"includes":["yyyy-MM-dd"]}}, requestedTotalSize=10000, pageSize=null, startFrom=0)])

integ-test/src/test/resources/expectedOutput/calcite/explain_filter_push_timestamp_wrapped_bound.yaml

@@ -0,0 +1,9 @@
+calcite:
+  logical: |
+    LogicalSystemLimit(fetch=[10000], type=[QUERY_SIZE_LIMIT])
+      LogicalProject(account_number=[$0], firstname=[$1], address=[$2], birthdate=[$3], gender=[$4], city=[$5], lastname=[$6], balance=[$7], employer=[$8], state=[$9], age=[$10], email=[$11], male=[$12])
+        LogicalFilter(condition=[<($3, TIMESTAMP('2018-11-09 00:00:00':VARCHAR))])
+          LogicalFilter(condition=[>($3, TIMESTAMP('2016-12-08 00:00:00':VARCHAR))])
+            CalciteLogicalIndexScan(table=[[OpenSearch, opensearch-sql_test_index_bank]])
+  physical: |
+    CalciteEnumerableIndexScan(table=[[OpenSearch, opensearch-sql_test_index_bank]], PushDownContext=[[PROJECT->[account_number, firstname, address, birthdate, gender, city, lastname, balance, employer, state, age, email, male], FILTER->SEARCH($3, Sarg[('2016-12-08 00:00:00':VARCHAR..'2018-11-09 00:00:00':VARCHAR)]:VARCHAR), LIMIT->10000], OpenSearchRequestBuilder(sourceBuilder={"from":0,"size":10000,"timeout":"1m","query":{"range":{"birthdate":{"from":"2016-12-08T00:00:00.000Z","to":"2018-11-09T00:00:00.000Z","include_lower":false,"include_upper":false,"format":"date_time","boost":1.0}}},"_source":{"includes":["account_number","firstname","address","birthdate","gender","city","lastname","balance","employer","state","age","email","male"]}}, requestedTotalSize=10000, pageSize=null, startFrom=0)])

integ-test/src/test/resources/expectedOutput/calcite/explain_filter_push_timestamp_wrapped_field.yaml

@@ -0,0 +1,9 @@
+calcite:
+  logical: |
+    LogicalSystemLimit(fetch=[10000], type=[QUERY_SIZE_LIMIT])
+      LogicalProject(account_number=[$0], firstname=[$1], address=[$2], birthdate=[$3], gender=[$4], city=[$5], lastname=[$6], balance=[$7], employer=[$8], state=[$9], age=[$10], email=[$11], male=[$12])
+        LogicalFilter(condition=[<(TIMESTAMP($3), TIMESTAMP('2018-11-09 00:00:00':VARCHAR))])
+          LogicalFilter(condition=[>(TIMESTAMP($3), TIMESTAMP('2016-12-08 00:00:00':VARCHAR))])
+            CalciteLogicalIndexScan(table=[[OpenSearch, opensearch-sql_test_index_bank]])
+  physical: |
+    CalciteEnumerableIndexScan(table=[[OpenSearch, opensearch-sql_test_index_bank]], PushDownContext=[[PROJECT->[account_number, firstname, address, birthdate, gender, city, lastname, balance, employer, state, age, email, male], FILTER->SEARCH(TIMESTAMP($3), Sarg[('2016-12-08 00:00:00':VARCHAR..'2018-11-09 00:00:00':VARCHAR)]:VARCHAR), LIMIT->10000], OpenSearchRequestBuilder(sourceBuilder={"from":0,"size":10000,"timeout":"1m","query":{"range":{"birthdate":{"from":"2016-12-08T00:00:00.000Z","to":"2018-11-09T00:00:00.000Z","include_lower":false,"include_upper":false,"format":"date_time","boost":1.0}}},"_source":{"includes":["account_number","firstname","address","birthdate","gender","city","lastname","balance","employer","state","age","email","male"]}}, requestedTotalSize=10000, pageSize=null, startFrom=0)])

integ-test/src/test/resources/expectedOutput/calcite_no_pushdown/explain_filter_last_day_no_push.yaml

@@ -0,0 +1,10 @@
+calcite:
+  logical: |
+    LogicalSystemLimit(fetch=[10000], type=[QUERY_SIZE_LIMIT])
+      LogicalFilter(condition=[=(LAST_DAY($0), DATE('2018-11-30':VARCHAR))])
+        LogicalProject(yyyy-MM-dd=[$83])
+          CalciteLogicalIndexScan(table=[[OpenSearch, opensearch-sql_test_index_date_formats]])
+  physical: |
+    EnumerableLimit(fetch=[10000])
+      EnumerableCalc(expr#0..94=[{inputs}], expr#95=[LAST_DAY($t83)], expr#96=['2018-11-30':EXPR_DATE VARCHAR], expr#97=[=($t95, $t96)], yyyy-MM-dd=[$t83], $condition=[$t97])
+        CalciteEnumerableIndexScan(table=[[OpenSearch, opensearch-sql_test_index_date_formats]])

integ-test/src/test/resources/expectedOutput/calcite_no_pushdown/explain_filter_push_timestamp_wrapped_bound.yaml

@@ -0,0 +1,11 @@
+calcite:
+  logical: |
+    LogicalSystemLimit(fetch=[10000], type=[QUERY_SIZE_LIMIT])
+      LogicalProject(account_number=[$0], firstname=[$1], address=[$2], birthdate=[$3], gender=[$4], city=[$5], lastname=[$6], balance=[$7], employer=[$8], state=[$9], age=[$10], email=[$11], male=[$12])
+        LogicalFilter(condition=[<($3, TIMESTAMP('2018-11-09 00:00:00':VARCHAR))])
+          LogicalFilter(condition=[>($3, TIMESTAMP('2016-12-08 00:00:00':VARCHAR))])
+            CalciteLogicalIndexScan(table=[[OpenSearch, opensearch-sql_test_index_bank]])
+  physical: |
+    EnumerableLimit(fetch=[10000])
+      EnumerableCalc(expr#0..18=[{inputs}], expr#19=[Sarg[('2016-12-08 00:00:00':VARCHAR..'2018-11-09 00:00:00':VARCHAR)]:VARCHAR], expr#20=[SEARCH($t3, $t19)], proj#0..12=[{exprs}], $condition=[$t20])
+        CalciteEnumerableIndexScan(table=[[OpenSearch, opensearch-sql_test_index_bank]])

integ-test/src/test/resources/expectedOutput/calcite_no_pushdown/explain_filter_push_timestamp_wrapped_field.yaml

@@ -0,0 +1,11 @@
+calcite:
+  logical: |
+    LogicalSystemLimit(fetch=[10000], type=[QUERY_SIZE_LIMIT])
+      LogicalProject(account_number=[$0], firstname=[$1], address=[$2], birthdate=[$3], gender=[$4], city=[$5], lastname=[$6], balance=[$7], employer=[$8], state=[$9], age=[$10], email=[$11], male=[$12])
+        LogicalFilter(condition=[<(TIMESTAMP($3), TIMESTAMP('2018-11-09 00:00:00':VARCHAR))])
+          LogicalFilter(condition=[>(TIMESTAMP($3), TIMESTAMP('2016-12-08 00:00:00':VARCHAR))])
+            CalciteLogicalIndexScan(table=[[OpenSearch, opensearch-sql_test_index_bank]])
+  physical: |
+    EnumerableLimit(fetch=[10000])
+      EnumerableCalc(expr#0..18=[{inputs}], expr#19=[TIMESTAMP($t3)], expr#20=[Sarg[('2016-12-08 00:00:00':VARCHAR..'2018-11-09 00:00:00':VARCHAR)]:VARCHAR], expr#21=[SEARCH($t19, $t20)], proj#0..12=[{exprs}], $condition=[$t21])
+        CalciteEnumerableIndexScan(table=[[OpenSearch, opensearch-sql_test_index_bank]])

integ-test/src/test/resources/expectedOutput/ppl/explain_filter_last_day_no_push.yaml

@@ -0,0 +1,20 @@
+root:
+  name: ProjectOperator
+  description:
+    fields: "[yyyy-MM-dd]"
+  children:
+    - name: FilterOperator
+      description:
+        conditions: "=(last_day(yyyy-MM-dd), date(\"2018-11-30\"))"
+      children:
+        - name: ProjectOperator
+          description:
+            fields: "[yyyy-MM-dd]"
+          children:
+            - name: OpenSearchIndexScan
+              description:
+                request: "OpenSearchQueryRequest(indexName=opensearch-sql_test_index_date_formats,\
+                  \ sourceBuilder={\"from\":0,\"size\":10000,\"timeout\":\"1m\",\"\
+                  _source\":{\"includes\":[\"yyyy-MM-dd\"]}}, pitId=*,\
+                  \ cursorKeepAlive=1m, searchAfter=null, searchResponse=null)"
+              children: []

integ-test/src/test/resources/expectedOutput/ppl/explain_filter_push_timestamp_wrapped_bound.yaml

@@ -0,0 +1,19 @@
+root:
+  name: ProjectOperator
+  description:
+    fields: "[account_number, firstname, address, birthdate, gender, city, lastname,\
+      \ balance, employer, state, age, email, male]"
+  children:
+    - name: OpenSearchIndexScan
+      description:
+        request: "OpenSearchQueryRequest(indexName=opensearch-sql_test_index_bank,\
+          \ sourceBuilder={\"from\":0,\"size\":10000,\"timeout\":\"1m\",\"query\"\
+          :{\"bool\":{\"filter\":[{\"range\":{\"birthdate\":{\"from\":null,\"to\"\
+          :1541721600000,\"include_lower\":true,\"include_upper\":false,\"boost\"\
+          :1.0}}},{\"range\":{\"birthdate\":{\"from\":1481155200000,\"to\":null,\"\
+          include_lower\":false,\"include_upper\":true,\"boost\":1.0}}}],\"adjust_pure_negative\"\
+          :true,\"boost\":1.0}},\"_source\":{\"includes\":[\"account_number\",\"firstname\"\
+          ,\"address\",\"birthdate\",\"gender\",\"city\",\"lastname\",\"balance\"\
+          ,\"employer\",\"state\",\"age\",\"email\",\"male\"]}}, pitId=*,\
+          \ cursorKeepAlive=1m, searchAfter=null, searchResponse=null)"
+      children: []

integ-test/src/test/resources/expectedOutput/ppl/explain_filter_push_timestamp_wrapped_field.yaml

@@ -0,0 +1,19 @@
+root:
+  name: ProjectOperator
+  description:
+    fields: "[account_number, firstname, address, birthdate, gender, city, lastname,\
+      \ balance, employer, state, age, email, male]"
+  children:
+    - name: OpenSearchIndexScan
+      description:
+        request: "OpenSearchQueryRequest(indexName=opensearch-sql_test_index_bank,\
+          \ sourceBuilder={\"from\":0,\"size\":10000,\"timeout\":\"1m\",\"query\"\
+          :{\"bool\":{\"filter\":[{\"range\":{\"birthdate\":{\"from\":null,\"to\"\
+          :1541721600000,\"include_lower\":true,\"include_upper\":false,\"boost\"\
+          :1.0}}},{\"range\":{\"birthdate\":{\"from\":1481155200000,\"to\":null,\"\
+          include_lower\":false,\"include_upper\":true,\"boost\":1.0}}}],\"adjust_pure_negative\"\
+          :true,\"boost\":1.0}},\"_source\":{\"includes\":[\"account_number\",\"firstname\"\
+          ,\"address\",\"birthdate\",\"gender\",\"city\",\"lastname\",\"balance\"\
+          ,\"employer\",\"state\",\"age\",\"email\",\"male\"]}}, pitId=*,\
+          \ cursorKeepAlive=1m, searchAfter=null, searchResponse=null)"
+      children: []

opensearch/src/main/java/org/opensearch/sql/opensearch/request/PredicateAnalyzer.java

@@ -249,6 +249,13 @@ public static QueryExpression analyzeExpression(
     }
   }
 
+  /**
+   * Date/time conversion operators, i.e. the ones that only reinterpret the value. A conversion to
+   * the type the field already has is a no-op and can be folded away; see {@link
+   * Visitor#isRedundantDateCastOverField}.
+   */
+  private static final Set<String> DATE_CONVERSION_OPERATORS = Set.of("TIMESTAMP", "DATE", "TIME");
+
   /** Traverses {@link RexNode} tree and builds OpenSearch query. */
   static class Visitor extends RexVisitorImpl<Expression> {
 
@@ -365,6 +372,14 @@ static RexUnknownAs getNullAsForSearch(RexCall search) {
 
     @Override
     public Expression visitCall(RexCall call) {
+      // Fold a redundant date/time cast over a field of the same date/time type -- e.g.
+      // timestamp(<timestamp field>) or CAST(<timestamp field> AS TIMESTAMP) -- to the bare field
+      // reference. Such a wrap is a no-op, so the enclosing comparison can push down to a native
+      // range query instead of falling back to a per-document script.
+      if (isRedundantDateCastOverField(call)) {
+        return visitInputRef((RexInputRef) call.getOperands().get(0));
+      }
+
       SqlSyntax syntax = call.getOperator().getSyntax();
       if (!supportedRexCall(call)) {
         String message = format(Locale.ROOT, "Unsupported call: [%s]", call);
@@ -1000,6 +1015,47 @@ private CastExpression toCastExpression(RexCall call) {
       return new CastExpression(call.getType(), argument);
     }
 
+    /**
+     * True if {@code call} is a date/time cast — a {@code CAST(... AS TIMESTAMP/DATE/TIME)} or the
+     * {@code timestamp()}/{@code date()}/{@code time()} builtins, both of which yield a date/time
+     * UDT — applied to a single field reference whose own type is <b>the same</b> date/time type.
+     * Such a wrap is a no-op, so it is redundant for a comparison and can be unwrapped to the bare
+     * field, letting the predicate push down instead of falling back to a per-document script.
+     *
+     * <p>The target type must match the field type exactly. A cast that changes the date/time type
+     * is a real conversion and must not be folded: {@code date(<timestamp field>)} truncates the
+     * time component (so {@code date(ts) <= '2024-01-15'} is not {@code ts <= '2024-01-15'}), and
+     * {@code time(<timestamp field>)} extracts the time of day, which is not even monotonic with
+     * respect to the timestamp.
+     */
+    private boolean isRedundantDateCastOverField(RexCall call) {
+      if (call.getOperands().size() != 1) {
+        return false;
+      }
+      if (!(call.getOperands().get(0) instanceof RexInputRef inputRef)) {
+        return false;
+      }
+      // Only a cast or a date/time conversion operator can be a no-op. The result type alone is not
+      // sufficient: other single-argument functions also return a date/time type while changing the
+      // value (LAST_DAY being the clearest example), and folding those away would be wrong.
+      if (call.getKind() != SqlKind.CAST
+          && !DATE_CONVERSION_OPERATORS.contains(
+              call.getOperator().getName().toUpperCase(Locale.ROOT))) {
+        return false;
+      }
+      // Date/time values are modelled as UDTs (EXPR_DATE/EXPR_TIME/EXPR_TIMESTAMP) whose backing
+      // SqlTypeName is VARCHAR, so the UDT identifies the cast target -- not getSqlTypeName().
+      if (!(call.getType() instanceof ExprSqlType exprSqlType)) {
+        return false;
+      }
+      ExprUDT udt = exprSqlType.getUdt();
+      if (udt != ExprUDT.EXPR_TIMESTAMP && udt != ExprUDT.EXPR_DATE && udt != ExprUDT.EXPR_TIME) {
+        return false;
+      }
+      ExprType fieldType = new NamedFieldExpression(inputRef, schema, fieldTypes).getCoreExprType();
+      return udt.getExprCoreType().equals(fieldType);
+    }
+
     private static NamedFieldExpression toNamedField(RexLiteral literal) {
       return new NamedFieldExpression(literal);
     }
@@ -1871,6 +1927,17 @@ boolean isTimeStampType() {
                   : type.getOriginalExprType());
     }
 
+    /** The field's underlying core type, unwrapping {@link OpenSearchDataType} if present. */
+    @Nullable
+    ExprType getCoreExprType() {
+      if (type == null) {
+        return null;
+      }
+      return type.getOriginalExprType() instanceof OpenSearchDataType osType
+          ? osType.getExprCoreType()
+          : type.getOriginalExprType();
+    }
+
     boolean isTextType() {
       return type != null && type.getOriginalExprType() instanceof OpenSearchTextType;
     }