work around missing tilde character in huffman tree

2026-08-14 09:57:21 +03:00
parent b12948bbef
commit bdc3c33026
1 changed files with 14 additions and 1 deletions
@@ -263,10 +263,23 @@ export function compress (input, alphabet) {
        huffmanNumber += BigInt(byte);
        huffmanNumber = huffmanEncode(huffmanNumber, pathEncode["%"]);
        i -= 2;
+      } else {
+        if (segment.value[i] === "~") {
+          /**
+           * HACK HACK HACK!!!
+           * Our Huffman tree is missing the tilde character (whoops!)
+           * It's too late to change it now without bumping the version
+           * number, and that currently costs 1 bit. Tildes are so rare
+           * that it makes more sense to %-encode them instead.
+           */
+          huffmanNumber *= 256n;
+          huffmanNumber += BigInt(126);
+          huffmanNumber = huffmanEncode(huffmanNumber, pathEncode["%"]);
        } else {
          huffmanNumber = huffmanEncode(huffmanNumber, pathEncode[segment.value[i]]);
        }
      }
+    }
    // Encode segment variant as 0
    // (We're adding +1 here to introduce 0 as a special value indicating Huffman)
    huffmanNumber *= BigInt(subalphabets.length + 1);