fix(codegen): consume original code units for length-changing case-insensitive folds

AUTHORS

@@ -40,3 +40,4 @@ Steven Spungin <steven@spungin.tv> (https://github.com/flamenco/)
 XenoS (https://github.com/XenoS-ITA)
 Samuel Bronson (https://github.com/SamB)
 knoan (https://github.com/knoan)
+Yarchik <spoko.dev@gmail.com> (https://github.com/spokodev/)

CHANGELOG.md

@@ -3,6 +3,18 @@ Change Log
 
 This file documents all notable changes to Peggy.
 
+Unreleased
+----------
+
+### Bug fixes
+
+- Fixed case-insensitive string literals whose lower case has a different
+  length than the original (for example `"İ"i`, where U+0130 is one code unit
+  but lower-cases to two). The generated parser now reads and consumes the
+  original code units instead of the lower-cased length, so the match no
+  longer fails.
+  [#662](https://github.com/peggyjs/peggy/pull/662)
+
 5.1.0
 -----
 

lib/compiler/passes/generate-js.js

@@ -737,10 +737,18 @@ function generateJS(ast, options) {
 
           case op.MATCH_STRING_IC: {  // MATCH_STRING_IC s, a, f, ...
             const litNum = bc[ip + 1];
+            // The stored literal is already lower-cased, so its length can
+            // differ from the original literal when `.toLowerCase()` changes
+            // the number of code units (e.g. "İ" -> "i̇"). The input
+            // chunk we read and consume must span the original code units, so
+            // use the paired ACCEPT_N count instead of the lower-cased length.
+            const inputChunkLength = bc[ip + 4] === op.ACCEPT_N
+              ? bc[ip + 5]
+              : literals[litNum].length;
             compileInputChunkCondition(
               inputChunk => `${inputChunk}.toLowerCase() === ${l(litNum)}`,
               1,
-              literals[litNum].length
+              inputChunkLength
             );
             break;
           }

test/behavior/generated-parser-behavior.spec.js

@@ -278,6 +278,15 @@ describe("generated parser behavior", () => {
           expect(parser).to.parse("a");
           expect(parser).to.parse("A");
         });
+
+        it("matches a case-insensitive literal whose lower case changes its length", () => {
+          // "İ" (U+0130) is one UTF-16 code unit, but `.toLowerCase()` yields
+          // two ("i" + U+0307). The case-insensitive match must read and
+          // consume the original code units, not the lower-cased length.
+          const parser = peg.generate("start = \"İ\"i .", options);
+
+          expect(parser).to.parse("İx", ["İ", "x"]);
+        });
       });
 
       describe("when it matches", () => {