feat: add fallback font loading and handling for various languages

src/js/utils/font-loader.ts

@@ -3,6 +3,7 @@ import {
   getFontUrlForFamily,
   languageToFontFamily,
 } from '../config/font-mappings.js';
+import { needsFontEmbedding } from './freetext-script.js';
 
 const fontCache: Map<string, ArrayBuffer> = new Map();
 
@@ -134,6 +135,56 @@ export async function getFontForLanguage(lang: string): Promise<ArrayBuffer> {
   }
 }
 
+const fallbackFontLoads = new Map<string, Promise<boolean>>();
+
+const fallbackFontFamily = (lang: string): string =>
+  languageToFontFamily[lang] || 'Noto Sans';
+
+export function fallbackFontLanguages(text: string): Set<string> {
+  const langs = new Set<string>();
+  for (const ch of text) {
+    if (needsFontEmbedding(ch)) langs.add(getLanguageForChar(ch));
+  }
+  return langs;
+}
+
+export function hasFallbackFonts(
+  loaded: Map<string, Uint8Array>,
+  langs: Iterable<string>
+): boolean {
+  return [...langs].every((lang) => loaded.has(fallbackFontFamily(lang)));
+}
+
+export async function loadFallbackFonts(
+  into: Map<string, Uint8Array>,
+  langs: Iterable<string>
+): Promise<boolean> {
+  const results = await Promise.all(
+    [...langs].map((lang) => {
+      const family = fallbackFontFamily(lang);
+      let load = fallbackFontLoads.get(family);
+      if (!load) {
+        load = getFontForLanguage(lang)
+          .then((buffer) => {
+            if (!fontCache.has(family)) {
+              throw new Error('Only a substitute font was available');
+            }
+            into.set(family, new Uint8Array(buffer));
+            return true;
+          })
+          .catch((error: unknown) => {
+            fallbackFontLoads.delete(family);
+            console.warn(`Fallback font unavailable for ${family}`, error);
+            return false;
+          });
+        fallbackFontLoads.set(family, load);
+      }
+      return load;
+    })
+  );
+  return results.every(Boolean);
+}
+
 export function detectScripts(text: string): string[] {
   const scripts = new Set<string>();
 

src/js/utils/freetext-flatten.ts

@@ -1,5 +1,6 @@
 import type { FreeTextSystemFontAnnotation } from '@/types';
 import { needsFontEmbedding } from './freetext-script.js';
+import { fallbackFontLanguages, loadFallbackFonts } from './font-loader.js';
 
 export const FPDF_ANNOT_FREETEXT = 3;
 
@@ -153,7 +154,7 @@ function buildRuns(
   return [
     {
       text: contents,
-      family: 'Helvetica',
+      family: needsFontEmbedding(contents) ? 'Noto Sans' : 'Helvetica',
       size,
       rgba: hexToRgba(style?.fontColor ?? '#000000'),
       bold: false,
@@ -196,6 +197,7 @@ export async function flattenFreeTextToPageText(
     PdfEngine: {
       create: () => Promise<EngineLike>;
       localFonts: Map<string, Uint8Array>;
+      fallbackFonts: Map<string, Uint8Array>;
     };
     scriptFallbackFamilies: (cps: number[]) => string[];
   };
@@ -245,6 +247,10 @@ export async function flattenFreeTextToPageText(
     return { bytes, flattened: 0 };
   }
   await primeScriptFonts(mod, found);
+  await loadFallbackFonts(
+    mod.PdfEngine.fallbackFonts,
+    fallbackFontLanguages(found.join('\n'))
+  );
 
   for (const pageIndex of [...pages].sort((x, y) => x - y)) {
     if (pageIndex < 0 || pageIndex >= eng.pageCount) continue;