perf: reduce Indic character searching

Author: cometkim

.changeset/grumpy-olives-warn.md

@@ -0,0 +1,5 @@
+---
+"unicode-segmenter": patch
+---
+
+grapheme: improve runtime perf by ~9% for most common use cases

README.md

@@ -256,7 +256,7 @@ Since [Hermes doesn't support the `Intl.Segmenter` API](https://github.com/faceb
 
 | Name                         | Unicode® | ESM? |   Size    | Size (min) | Size (min+gzip) | Size (min+br) |
 |------------------------------|----------|------|----------:|-----------:|----------------:|--------------:|
-| `unicode-segmenter/grapheme` |   16.0.0 |    ✔️ |    15,997 |     12,130 |           5,061 |         3,751 |
+| `unicode-segmenter/grapheme` |   16.0.0 |    ✔️ |    16,053 |     12,150 |           5,070 |         3,752 |
 | `graphemer`                  |   15.0.0 |    ✖️ ️|   410,435 |     95,104 |          15,752 |        10,660 |
 | `grapheme-splitter`          |   10.0.0 |    ✖️ |   122,252 |     23,680 |           7,852 |         4,841 |
 | `@formatjs/intl-segmenter`*  |   15.0.0 |    ✖️ |   603,285 |    369,560 |          72,218 |        49,416 |
@@ -272,9 +272,9 @@ Since [Hermes doesn't support the `Intl.Segmenter` API](https://github.com/faceb
 
 | Name                         | Bytecode size | Bytecode size (gzip)* |
 |------------------------------|--------------:|----------------------:|
-| `unicode-segmenter/grapheme` |        22,061 |                11,539 |
-| `graphemer`                  |       133,974 |                31,719 |
-| `grapheme-splitter`          |        63,831 |                19,140 |
+| `unicode-segmenter/grapheme` |        22,110 |                11,566 |
+| `graphemer`                  |       133,978 |                31,713 |
+| `grapheme-splitter`          |        63,835 |                19,137 |
 
 * It would be compressed when included as an app asset.
 

src/grapheme.js

@@ -104,13 +104,15 @@ export function* graphemeSegments(input) {
       catBegin = catBefore;
     }
 
-    // Note: Lazily update `consonant` and `linker` state
-    // which is a extra overhead only for Hindi text.
-    if (!consonant && catBefore === 0) {
-      consonant = isIndicConjunctCosonant(cp);
-    } else if (catBefore === 3 /* Extend */) {
-      // Note: \p{InCB=Linker} is a subset of \p{Extend}
-      linker = isIndicConjunctLinker(cp);
+    if (cp >= 2325) {
+      // Note: Lazily update `consonant` and `linker` state
+      // which is a extra overhead only for Hindi text.
+      if (!consonant && catBefore === 0) {
+        consonant = isIndicConjunctCosonant(cp);
+      } else if (catBefore === 3 /* Extend */) {
+        // Note: \p{InCB=Linker} is a subset of \p{Extend}
+        linker = isIndicConjunctLinker(cp);
+      }
     }
 
     if (cursor < len) {
@@ -138,7 +140,7 @@ export function* graphemeSegments(input) {
       ) {
         emoji = true;
 
-      } else if (catAfter === 0 /* Any */) {
+      } else if (catAfter === 0 /* Any */ && cp >= 2325) {
         // Note: Put GB9c rule checking here to reduce.
         incb = consonant && linker && (consonant = isIndicConjunctCosonant(cp));
         // It cannot be both a linker and a consonant.