From f55f77a63a4e827ec26f4af7b41693eef6e84e06 Mon Sep 17 00:00:00 2001
From: prql-bot <107324867+prql-bot@users.noreply.github.com>
Date: Thu, 23 Jul 2026 07:18:21 +0000
Subject: [PATCH 1/2] fix: correct byte indexing for non-ASCII Jinja
 interpolations

find_span used char indices from chars().enumerate() to slice source by
bytes, corrupting (or panicking on) spans when the source contained
multi-byte characters before an interpolation. Use char_indices() and
len_utf8() so the slice stays on char boundaries.
---
 prqlc/prqlc/src/cli/jinja.rs | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/prqlc/prqlc/src/cli/jinja.rs b/prqlc/prqlc/src/cli/jinja.rs
index 3d8f219a8aaa..6b4722fff4ae 100644
--- a/prqlc/prqlc/src/cli/jinja.rs
+++ b/prqlc/prqlc/src/cli/jinja.rs
@@ -98,7 +98,9 @@ fn find_span(source: &str, spans: Vec<Span>) -> &str {
 
     let mut line = 1;
     let mut col = 0;
-    for (index, char) in source.chars().enumerate() {
+    // `char_indices` yields byte offsets, so the slice below stays on char
+    // boundaries even when `source` contains multi-byte characters.
+    for (index, char) in source.char_indices() {
         if char == '\n' {
             line += 1;
             col = 0;
@@ -111,7 +113,7 @@ fn find_span(source: &str, spans: Vec<Span>) -> &str {
             start_index = index;
         }
         if line == end.end_line && col == end.end_col {
-            end_index = index + 1;
+            end_index = index + char.len_utf8();
         }
     }
     &source[start_index..end_index]
@@ -198,6 +200,17 @@ mod test {
         insta::assert_yaml_snapshot!(post_proc_text, @r#""from in_process =  {{ source('salesforce', 'in_process') }}""#);
     }
 
+    #[test]
+    fn test_non_ascii_before_interpolation() {
+        // Non-ASCII bytes before an interpolation must not cause a panic or a
+        // mis-sliced span. `é` is two bytes but one char, so a char-indexed
+        // slice would land mid-character.
+        let src = r###"from café = {{ source('salesforce', 'in_process') }}"###;
+        let (pre_proc_text, ctx) = pre_process(src).unwrap();
+        insta::assert_yaml_snapshot!(pre_proc_text, @"from café = _jinja_0");
+        insta::assert_yaml_snapshot!(ctx.anchor_map["_jinja_0"], @r#"" {{ source('salesforce', 'in_process') }}""#);
+    }
+
     #[test]
     fn test_config_interpolation() {
         let src = r###"{{ config(materialized = "table") }}\nfrom in_process = {{ source('salesforce', 'in_process') }}"###;

From 7f7f20f2c0c3f6d81b424fa74d3fd9de40e27d99 Mon Sep 17 00:00:00 2001
From: prql-bot <107324867+prql-bot@users.noreply.github.com>
Date: Sat, 25 Jul 2026 20:14:24 +0000
Subject: [PATCH 2/2] fix: reword comment to satisfy typos

---
 prqlc/prqlc/src/cli/jinja.rs | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/prqlc/prqlc/src/cli/jinja.rs b/prqlc/prqlc/src/cli/jinja.rs
index 6b4722fff4ae..4fe23211dd37 100644
--- a/prqlc/prqlc/src/cli/jinja.rs
+++ b/prqlc/prqlc/src/cli/jinja.rs
@@ -203,7 +203,7 @@ mod test {
     #[test]
     fn test_non_ascii_before_interpolation() {
         // Non-ASCII bytes before an interpolation must not cause a panic or a
-        // mis-sliced span. `é` is two bytes but one char, so a char-indexed
+        // corrupted span. `é` is two bytes but one char, so a char-indexed
         // slice would land mid-character.
         let src = r###"from café = {{ source('salesforce', 'in_process') }}"###;
         let (pre_proc_text, ctx) = pre_process(src).unwrap();