ESQL: Change "substring" function to not return null on empty string (e…

…lastic#109174)

Currently, `substring("", 1, 1)` returns null instead of an empty string. This PR changes the behaviour to the expected one, as in other languages and databases.

Unrelated to this PR's intent, returning null is causing the function to fail with a NPE, so this is currently raising an error, and doesn't affect existing queries.

Related issue: elastic#109095

docs/changelog/109174.yaml

@@ -0,0 +1,5 @@
+pr: 109174
+summary: "ESQL: Change \"substring\" function to not return null on empty string"
+area: ES|QL
+type: bug
+issues: []

x-pack/plugin/esql/qa/testFixtures/src/main/java/org/elasticsearch/xpack/esql/CsvTestUtils.java

@@ -354,7 +354,8 @@ public static ExpectedResults loadCsvSpecValues(String csv) {
                 for (int i = 0; i < row.size(); i++) {
                     String value = row.get(i);
                     if (value == null) {
-                        rowValues.add(null);
+                        // Empty cells are converted to null by SuperCSV. We convert them back to empty strings.
+                        rowValues.add("");
                         continue;
                     }
 

x-pack/plugin/esql/qa/testFixtures/src/main/resources/string.csv-spec

@@ -195,6 +195,14 @@ emp_no:integer | last_name:keyword | x:keyword | z:keyword
 10010 | Piveteau  | P | a
 ;
 
+substring empty string
+required_capability: fn_substring_empty_null
+row sub = substring("", 1, 3);
+
+sub:keyword
+""
+;
+
 substring Emoji#[skip:-8.13.99,reason:bug fix in 8.14]
 row a = "🐱Meow!🐶Woof!" | eval sub1 = substring(a, 2) | eval sub2 = substring(a, 2, 100);
 

x-pack/plugin/esql/src/main/java/org/elasticsearch/xpack/esql/action/EsqlCapabilities.java

@@ -32,6 +32,11 @@ public class EsqlCapabilities {
      */
     private static final String FN_IP_PREFIX = "fn_ip_prefix";
 
+    /**
+     * Fix on function {@code SUBSTRING} that makes it not return null on empty strings.
+     */
+    private static final String FN_SUBSTRING_EMPTY_NULL = "fn_substring_empty_null";
+
     /**
      * Optimization for ST_CENTROID changed some results in cartesian data. #108713
      */
@@ -53,6 +58,7 @@ private static Set<String> capabilities() {
         List<String> caps = new ArrayList<>();
         caps.add(FN_CBRT);
         caps.add(FN_IP_PREFIX);
+        caps.add(FN_SUBSTRING_EMPTY_NULL);
         caps.add(ST_CENTROID_AGG_OPTIMIZED);
         caps.add(METADATA_IGNORED_FIELD);
 

x-pack/plugin/esql/src/main/java/org/elasticsearch/xpack/esql/expression/function/scalar/string/Substring.java

@@ -111,12 +111,12 @@ static BytesRef process(BytesRef str, int start) {
 
     @Evaluator
     static BytesRef process(BytesRef str, int start, int length) {
-        if (str.length == 0) {
-            return null;
-        }
         if (length < 0) {
             throw new IllegalArgumentException("Length parameter cannot be negative, found [" + length + "]");
         }
+        if (str.length == 0) {
+            return str;
+        }
         int codePointCount = UnicodeUtil.codePointCount(str);
         int indexStart = indexStart(codePointCount, start);
         int indexEnd = Math.min(codePointCount, indexStart + length);

x-pack/plugin/esql/src/test/java/org/elasticsearch/xpack/esql/expression/function/scalar/string/SubstringTests.java

@@ -78,6 +78,24 @@ public static Iterable<Object[]> parameters() {
                                     equalTo(new BytesRef(text.substring(start - 1, start + length - 1)))
                                 );
                             }
+                        ),
+                        new TestCaseSupplier(
+                            "Substring empty string",
+                            List.of(DataTypes.TEXT, DataTypes.INTEGER, DataTypes.INTEGER),
+                            () -> {
+                                int start = between(1, 8);
+                                int length = between(1, 10 - start);
+                                return new TestCaseSupplier.TestCase(
+                                    List.of(
+                                        new TestCaseSupplier.TypedData(new BytesRef(""), DataTypes.TEXT, "str"),
+                                        new TestCaseSupplier.TypedData(start, DataTypes.INTEGER, "start"),
+                                        new TestCaseSupplier.TypedData(length, DataTypes.INTEGER, "end")
+                                    ),
+                                    "SubstringEvaluator[str=Attribute[channel=0], start=Attribute[channel=1], length=Attribute[channel=2]]",
+                                    DataTypes.KEYWORD,
+                                    equalTo(new BytesRef(""))
+                                );
+                            }
                         )
                     )
                 )