Skip to content

Apache DataFusion SQL Query Engine

chore: fix `datafusion-spark` substring (#21963)

Oleks Vba038e99c861
diff --git a/datafusion/spark/src/function/string/substring.rs b/datafusion/spark/src/function/string/substring.rs
index ee6919dc0d..21b2242329 100644
--- a/datafusion/spark/src/function/string/substring.rs
+++ b/datafusion/spark/src/function/string/substring.rs
@@ -159,15 +159,17 @@ fn spark_substring(args: &[ArrayRef]) -> Result<ArrayRef> {
/// - Zero start: treated as 1
/// - Negative start: counts from end of string
///
-/// Returns the converted 1-based start position for use with `get_true_start_end`.
+/// The result may be `<= 0` when a negative start lands before the string
+/// (e.g. `start=-10` on a 3-char string gives `-6`). Such values are passed
+/// through to `get_true_start_end`, which clamps them and yields an empty
+/// slice — matching Spark's behavior for out-of-range negative positions.
#[inline]
fn spark_start_to_datafusion_start(start: i64, len: usize) -> i64 {
if start >= 0 {
start.max(1)
} else {
let len_i64 = i64::try_from(len).unwrap_or(i64::MAX);
- let start = start.saturating_add(len_i64).saturating_add(1);
- start.max(1)
+ start + len_i64 + 1
}
}
diff --git a/datafusion/sqllogictest/test_files/spark/string/substring.slt b/datafusion/sqllogictest/test_files/spark/string/substring.slt
index 5bf2fdf2fb..9189ce50c2 100644
--- a/datafusion/sqllogictest/test_files/spark/string/substring.slt
+++ b/datafusion/sqllogictest/test_files/spark/string/substring.slt
@@ -146,7 +146,7 @@ ark SQL
(empty)
(empty)
(empty)
-Spa
+(empty)
NULL
NULL
NULL
@@ -197,7 +197,14 @@ ark SQL
(empty)
(empty)
(empty)
-Spa
+(empty)
NULL
NULL
NULL
+
+query T
+SELECT substr(column1, -10, 3)
+FROM VALUES
+('abc'::string)
+----
+(empty)
\ No newline at end of file