@cj-tech-master/excelts 7.6.0 → 8.0.0

package/dist/esm/modules/archive/compression/deflate-fallback.js

@@ -351,15 +351,33 @@ export function deflateRawStore(data) {
 // while providing a good distribution for the 3-byte hash.
 const HASH_SIZE = 32768;
 const HASH_MASK = HASH_SIZE - 1;
-// Maximum hash chain length to walk per position. Longer chains find better
-// matches at the cost of speed. 64 is a good balance (~zlib level 5-6).
-const MAX_CHAIN_LEN = 64;
 // Minimum match length for LZ77 (RFC 1951 minimum).
 const MIN_MATCH = 3;
 // Maximum match length (RFC 1951 maximum).
 const MAX_MATCH = 258;
 // Maximum back-reference distance (RFC 1951 / 32 KB sliding window).
 const MAX_DIST = 32768;
+/**
+ * Get LZ77 configuration for the given compression level (1-9).
+ * Modelled after zlib's configuration_table.
+ */
+function getLZ77Config(level) {
+    // Level 0 should be handled by the caller (store mode).
+    if (level <= 1) {
+        return { maxChainLen: 4, goodLen: 4, niceLen: 8, lazy: false };
+    }
+    if (level <= 3) {
+        return { maxChainLen: 8, goodLen: 8, niceLen: 32, lazy: true };
+    }
+    if (level <= 5) {
+        return { maxChainLen: 32, goodLen: 16, niceLen: 128, lazy: true };
+    }
+    if (level <= 7) {
+        return { maxChainLen: 64, goodLen: 32, niceLen: 258, lazy: true };
+    }
+    // level 8-9
+    return { maxChainLen: 128, goodLen: 64, niceLen: 258, lazy: true };
+}
 /**
  * Hash function for 3-byte sequences.
  * Uses a multiplicative hash for better distribution than the naive
@@ -370,16 +388,16 @@ function hash3(a, b, c) {
     return ((((a << 16) | (b << 8) | c) * 0x1e35a7bd) >>> 17) & HASH_MASK;
 }
 /**
- * Compress data using DEFLATE with fixed Huffman codes.
+ * Compress data using DEFLATE with Dynamic Huffman codes (BTYPE=2).
  *
- * Uses LZ77 with hash chains and lazy matching for significantly better
- * compression than a single-entry hash table. The algorithm is modelled
- * after zlib's "fast" and "slow" deflate strategies.
+ * Uses LZ77 with hash chains and lazy matching for match finding, then builds
+ * optimal Huffman trees from the symbol frequencies for entropy coding.
  *
  * @param data - Data to compress
+ * @param level - Compression level (1-9, default 6)
  * @returns Compressed data in deflate-raw format
  */
-export function deflateRawCompressed(data) {
+export function deflateRawCompressed(data, level = 6) {
     if (data.length === 0) {
         // Empty input: single final block with just end-of-block symbol
         return new Uint8Array([0x03, 0x00]);
@@ -388,112 +406,12 @@ export function deflateRawCompressed(data) {
     if (data.length < 100) {
         return deflateRawStore(data);
     }
+    const config = getLZ77Config(level);
+    // --- Phase 1: LZ77 match finding → collect symbols ---
+    const lz77Symbols = lz77Compress(data, 0, data.length, config, null);
+    // --- Phase 2: Encode as a single final DEFLATE block ---
     const output = new BitWriter();
-    // Write final block header with fixed Huffman (BFINAL=1, BTYPE=01)
-    output.writeBits(1, 1); // BFINAL
-    output.writeBits(1, 2); // BTYPE = 01 (fixed Huffman)
-    // --- Hash chain tables (typed arrays for performance) ---
-    // head[h]: most recent position with hash h (0 = unused, positions are 1-based internally)
-    // prev[pos & (MAX_DIST-1)]: previous position in the chain for the same hash
-    const head = new Int32Array(HASH_SIZE); // filled with 0 (no match)
-    const prev = new Int32Array(MAX_DIST);
-    let pos = 0;
-    // State for lazy matching:
-    // When we find a match at position N, we check position N+1 too.
-    // If N+1 has a longer match we emit a literal for N and use the N+1 match.
-    let prevMatchLen = 0;
-    let prevMatchDist = 0;
-    let prevLiteral = 0;
-    let hasPrevMatch = false;
-    while (pos < data.length) {
-        let bestLen = 0;
-        let bestDist = 0;
-        if (pos + 2 < data.length) {
-            const h = hash3(data[pos], data[pos + 1], data[pos + 2]);
-            // Walk the hash chain to find the best (longest) match
-            let chainLen = MAX_CHAIN_LEN;
-            let matchHead = head[h];
-            while (matchHead > 0 && chainLen-- > 0) {
-                const mPos = matchHead - 1; // convert from 1-based to 0-based
-                const dist = pos - mPos;
-                if (dist > MAX_DIST || dist <= 0) {
-                    break;
-                }
-                // Quick check: compare the byte just beyond current best length first
-                // to skip obviously shorter matches early.
-                if (bestLen >= MIN_MATCH && data[mPos + bestLen] !== data[pos + bestLen]) {
-                    matchHead = prev[mPos & (MAX_DIST - 1)];
-                    continue;
-                }
-                // Full scan
-                let len = 0;
-                const maxLen = Math.min(MAX_MATCH, data.length - pos);
-                while (len < maxLen && data[mPos + len] === data[pos + len]) {
-                    len++;
-                }
-                if (len > bestLen) {
-                    bestLen = len;
-                    bestDist = dist;
-                    if (len >= MAX_MATCH) {
-                        break; // can't do better
-                    }
-                }
-                matchHead = prev[mPos & (MAX_DIST - 1)];
-            }
-            // Insert current position into the hash chain
-            prev[pos & (MAX_DIST - 1)] = head[h];
-            head[h] = pos + 1; // 1-based
-        }
-        // --- Lazy matching logic ---
-        if (hasPrevMatch) {
-            if (bestLen > prevMatchLen) {
-                // Current position has a better match; emit previous as literal
-                writeLiteralCode(output, prevLiteral);
-                // Now adopt current match as the pending one
-                prevMatchLen = bestLen;
-                prevMatchDist = bestDist;
-                prevLiteral = data[pos];
-                pos++;
-            }
-            else {
-                // Previous match is at least as good; emit it
-                writeLengthCode(output, prevMatchLen);
-                writeDistanceCode(output, prevMatchDist);
-                // Insert hash entries for the skipped bytes (positions inside the match)
-                // so future matches can find them. We already inserted pos-1 (the match
-                // start); now insert pos through pos + prevMatchLen - 2.
-                const matchEnd = pos - 1 + prevMatchLen;
-                for (let i = pos; i < matchEnd && i + 2 < data.length; i++) {
-                    const h = hash3(data[i], data[i + 1], data[i + 2]);
-                    prev[i & (MAX_DIST - 1)] = head[h];
-                    head[h] = i + 1;
-                }
-                pos = matchEnd;
-                hasPrevMatch = false;
-                prevMatchLen = 0;
-            }
-        }
-        else if (bestLen >= MIN_MATCH) {
-            // We have a match; hold it and try the next position (lazy evaluation)
-            hasPrevMatch = true;
-            prevMatchLen = bestLen;
-            prevMatchDist = bestDist;
-            prevLiteral = data[pos];
-            pos++;
-        }
-        else {
-            // No match — emit literal
-            writeLiteralCode(output, data[pos]);
-            pos++;
-        }
-    }
-    // Flush any pending lazy match
-    if (hasPrevMatch) {
-        writeLengthCode(output, prevMatchLen);
-        writeDistanceCode(output, prevMatchDist);
-    }
-    // Write end-of-block symbol (256)
-    writeLiteralCode(output, 256);
+    emitDynamicBlock(output, lz77Symbols, true);
     return output.finish();
 }
 /**
@@ -506,6 +424,14 @@ class BitWriter {
         this.bitBuf = 0;
         this.bitCount = 0;
     }
+    /**
+     * Align to the next byte boundary by padding with zero bits.
+     */
+    alignToByte() {
+        if (this.bitCount > 0) {
+            this.writeBits(0, 8 - this.bitCount);
+        }
+    }
     writeBits(value, count) {
         this.bitBuf |= value << this.bitCount;
         this.bitCount += count;
@@ -599,258 +525,512 @@ function writeLiteralCode(output, symbol) {
     output.writeBitsReverse(code, len);
 }
 /**
- * Write a length code (257-285)
+ * Compute the DEFLATE length code (257..285) and extra bits for a given
+ * match length (3..258).
  */
-function writeLengthCode(output, length) {
-    let code;
-    let extraBits;
-    let extraValue;
-    if (length <= 10) {
-        code = 257 + length - 3;
-        extraBits = 0;
-        extraValue = 0;
-    }
-    else if (length <= 18) {
-        const base = length - 11;
-        code = 265 + Math.floor(base / 2);
-        extraBits = 1;
-        extraValue = base % 2;
-    }
-    else if (length <= 34) {
-        const base = length - 19;
-        code = 269 + Math.floor(base / 4);
-        extraBits = 2;
-        extraValue = base % 4;
-    }
-    else if (length <= 66) {
-        const base = length - 35;
-        code = 273 + Math.floor(base / 8);
-        extraBits = 3;
-        extraValue = base % 8;
-    }
-    else if (length <= 130) {
-        const base = length - 67;
-        code = 277 + Math.floor(base / 16);
-        extraBits = 4;
-        extraValue = base % 16;
-    }
-    else if (length <= 257) {
-        const base = length - 131;
-        code = 281 + Math.floor(base / 32);
-        extraBits = 5;
-        extraValue = base % 32;
+function getLengthSymbol(length) {
+    for (let i = 0; i < LENGTH_BASE.length; i++) {
+        if (i === LENGTH_BASE.length - 1 || length < LENGTH_BASE[i + 1]) {
+            return {
+                code: 257 + i,
+                extra: length - LENGTH_BASE[i],
+                extraBits: LENGTH_EXTRA[i]
+            };
+        }
     }
-    else {
-        code = 285;
-        extraBits = 0;
-        extraValue = 0;
+    return { code: 285, extra: 0, extraBits: 0 };
+}
+/**
+ * Compute the DEFLATE distance code (0..29) and extra bits for a given
+ * distance (1..32768).
+ */
+function getDistSymbol(distance) {
+    for (let i = 0; i < DIST_TABLE.length; i++) {
+        const [maxDist, c, extraBitsCount] = DIST_TABLE[i];
+        if (distance <= maxDist) {
+            const baseVal = i === 0 ? 1 : DIST_TABLE[i - 1][0] + 1;
+            return { code: c, extra: distance - baseVal, extraBits: extraBitsCount };
+        }
+    }
+    // Fallback (should not reach for valid distances)
+    return { code: 29, extra: 0, extraBits: 13 };
+}
+/**
+ * Build canonical Huffman code lengths from symbol frequencies.
+ * Uses a bottom-up approach: build a Huffman tree from a priority queue,
+ * then extract depths. Limits maximum code length to maxBits using
+ * the algorithm from zlib's build_tree() / gen_bitlen().
+ *
+ * Returns an array of code lengths indexed by symbol.
+ */
+function buildCodeLengths(freqs, maxBits) {
+    const n = freqs.length;
+    const codeLens = new Uint8Array(n);
+    // Count symbols with non-zero frequency
+    const activeSymbols = [];
+    for (let i = 0; i < n; i++) {
+        if (freqs[i] > 0) {
+            activeSymbols.push({ sym: i, freq: freqs[i] });
+        }
+    }
+    if (activeSymbols.length === 0) {
+        return codeLens;
+    }
+    // RFC 1951 requires a complete prefix code. For a single symbol, we need
+    // at least 2 entries to form a valid tree. We assign code length 1 to the
+    // symbol — the decoder uses only 1 bit but the tree is valid because
+    // DEFLATE decoders handle this as per the spec (the other 1-bit code is
+    // simply unused). This matches zlib's behavior.
+    if (activeSymbols.length === 1) {
+        codeLens[activeSymbols[0].sym] = 1;
+        return codeLens;
+    }
+    // Sort by frequency (ascending), then by symbol (ascending) for stability
+    activeSymbols.sort((a, b) => a.freq - b.freq || a.sym - b.sym);
+    const nodes = activeSymbols.map(s => ({
+        freq: s.freq,
+        sym: s.sym,
+        left: null,
+        right: null
+    }));
+    let leafIdx = 0;
+    let intIdx = 0;
+    const intNodes = [];
+    function getMin() {
+        const hasLeaf = leafIdx < nodes.length;
+        const hasInt = intIdx < intNodes.length;
+        if (hasLeaf && hasInt) {
+            if (nodes[leafIdx].freq <= intNodes[intIdx].freq) {
+                return nodes[leafIdx++];
+            }
+            return intNodes[intIdx++];
+        }
+        if (hasLeaf) {
+            return nodes[leafIdx++];
+        }
+        return intNodes[intIdx++];
+    }
+    const totalNodes = activeSymbols.length;
+    for (let i = 0; i < totalNodes - 1; i++) {
+        const a = getMin();
+        const b = getMin();
+        const merged = {
+            freq: a.freq + b.freq,
+            sym: -1,
+            left: a,
+            right: b
+        };
+        intNodes.push(merged);
+    }
+    // Extract depths from the root (last internal node)
+    const root = intNodes[intNodes.length - 1];
+    function extractDepths(node, depth) {
+        if (node.sym >= 0) {
+            codeLens[node.sym] = depth;
+            return;
+        }
+        if (node.left) {
+            extractDepths(node.left, depth + 1);
+        }
+        if (node.right) {
+            extractDepths(node.right, depth + 1);
+        }
+    }
+    extractDepths(root, 0);
+    // --- Length limiting using the zlib bl_count redistribution algorithm ---
+    // Count code lengths at each bit depth
+    const blCount = new Uint16Array(maxBits + 1);
+    for (let i = 0; i < n; i++) {
+        if (codeLens[i] > 0) {
+            if (codeLens[i] > maxBits) {
+                blCount[maxBits]++;
+                codeLens[i] = maxBits;
+            }
+            else {
+                blCount[codeLens[i]]++;
+            }
+        }
+    }
+    // Check Kraft inequality: sum of 2^(maxBits - len) must equal 2^maxBits
+    let kraft = 0;
+    for (let bits = 1; bits <= maxBits; bits++) {
+        kraft += blCount[bits] << (maxBits - bits);
+    }
+    const target = 1 << maxBits;
+    if (kraft === target) {
+        return codeLens; // Already valid
+    }
+    // Redistribute to satisfy Kraft's inequality.
+    // Strategy: move symbols from shorter lengths to maxBits until balanced.
+    // Each symbol moved from length `bits` to `maxBits` reduces kraft by
+    // (2^(maxBits-bits) - 1) — we remove a large weight and add a weight of 1.
+    while (kraft > target) {
+        // Find a code length < maxBits that has symbols we can push down.
+        // Start from maxBits-1 to minimize the damage per move.
+        let bits = maxBits - 1;
+        while (bits > 0 && blCount[bits] === 0) {
+            bits--;
+        }
+        if (bits === 0) {
+            break; // Can't redistribute further
+        }
+        // Move one symbol from length `bits` to length `maxBits`
+        blCount[bits]--;
+        blCount[maxBits]++;
+        // Kraft change: removed 2^(maxBits-bits), added 2^0 = 1
+        kraft -= (1 << (maxBits - bits)) - 1;
     }
-    writeLiteralCode(output, code);
-    if (extraBits > 0) {
-        output.writeBits(extraValue, extraBits);
+    // If kraft < target (under-allocated), add dummy codes at maxBits.
+    // This can happen when we overshoot during redistribution.
+    while (kraft < target) {
+        blCount[maxBits]++;
+        kraft++;
     }
+    // Reassign code lengths to symbols (preserve relative order: longer
+    // codes go to less frequent symbols, matching the Huffman property).
+    // Sort symbols by their original code length (longest first), then by
+    // frequency (rarest first) for same length.
+    const symbolsByLen = [];
+    for (let i = 0; i < n; i++) {
+        if (codeLens[i] > 0) {
+            symbolsByLen.push({ sym: i, origLen: codeLens[i], freq: freqs[i] });
+        }
+    }
+    symbolsByLen.sort((a, b) => b.origLen - a.origLen || a.freq - b.freq);
+    // Assign new lengths from the bl_count distribution
+    codeLens.fill(0);
+    let symIdx = 0;
+    for (let bits = maxBits; bits >= 1; bits--) {
+        for (let count = blCount[bits]; count > 0; count--) {
+            if (symIdx < symbolsByLen.length) {
+                codeLens[symbolsByLen[symIdx].sym] = bits;
+                symIdx++;
+            }
+        }
+    }
+    return codeLens;
 }
 /**
- * Write a distance code
+ * Build canonical Huffman codes from code lengths (RFC 1951 §3.2.2).
+ * Returns [code, length] pairs indexed by symbol.
  */
-function writeDistanceCode(output, distance) {
-    // Find the appropriate distance code
+function buildCanonicalCodes(codeLens) {
+    const n = codeLens.length;
+    const codes = new Array(n);
+    const blCount = new Uint16Array(16);
+    for (let i = 0; i < n; i++) {
+        if (codeLens[i] > 0) {
+            blCount[codeLens[i]]++;
+        }
+    }
+    const nextCode = new Uint16Array(16);
     let code = 0;
-    let extraBits = 0;
-    let baseDistance = 1;
-    for (let i = 0; i < DIST_TABLE.length; i++) {
-        const [maxDist, c, extra] = DIST_TABLE[i];
-        if (distance <= maxDist) {
-            code = c;
-            extraBits = extra;
+    for (let bits = 1; bits <= 15; bits++) {
+        code = (code + blCount[bits - 1]) << 1;
+        nextCode[bits] = code;
+    }
+    for (let i = 0; i < n; i++) {
+        const len = codeLens[i];
+        if (len > 0) {
+            codes[i] = [nextCode[len]++, len];
+        }
+        else {
+            codes[i] = [0, 0];
+        }
+    }
+    return codes;
+}
+/**
+ * Emit a Dynamic Huffman DEFLATE block (BTYPE=2).
+ *
+ * Takes the LZ77 symbol sequence, builds optimal Huffman trees,
+ * encodes the tree descriptions, then encodes the symbols.
+ */
+function emitDynamicBlock(out, symbols, isFinal) {
+    // --- Step 1: Collect frequencies ---
+    const litLenFreqs = new Uint32Array(286);
+    const distFreqs = new Uint32Array(30);
+    // Always include EOB
+    litLenFreqs[256] = 1;
+    for (const sym of symbols) {
+        if (sym.dist === 0) {
+            litLenFreqs[sym.litOrLen]++;
+        }
+        else {
+            const ls = getLengthSymbol(sym.litOrLen);
+            litLenFreqs[ls.code]++;
+            const ds = getDistSymbol(sym.dist);
+            distFreqs[ds.code]++;
+        }
+    }
+    // --- Step 2: Build Huffman trees ---
+    const litLenLens = buildCodeLengths(litLenFreqs, 15);
+    let distLens = buildCodeLengths(distFreqs, 15);
+    // DEFLATE requires at least 1 distance code even if unused.
+    // Assign two codes at length 1 to form a complete prefix code.
+    let hasDistCodes = false;
+    for (let i = 0; i < distLens.length; i++) {
+        if (distLens[i] > 0) {
+            hasDistCodes = true;
             break;
         }
-        baseDistance = maxDist + 1;
     }
-    const extraValue = distance - baseDistance;
-    // Distance codes use 5-bit fixed code (reversed for Huffman)
-    output.writeBitsReverse(code, 5);
-    if (extraBits > 0) {
-        output.writeBits(extraValue, extraBits);
+    if (!hasDistCodes) {
+        distLens = new Uint8Array(30);
+        distLens[0] = 1;
+        distLens[1] = 1;
+    }
+    const litLenCodes = buildCanonicalCodes(litLenLens);
+    const distCodes = buildCanonicalCodes(distLens);
+    // --- Step 3: Determine HLIT and HDIST ---
+    let hlit = 286;
+    while (hlit > 257 && litLenLens[hlit - 1] === 0) {
+        hlit--;
+    }
+    let hdist = 30;
+    while (hdist > 1 && distLens[hdist - 1] === 0) {
+        hdist--;
+    }
+    // --- Step 4: Run-length encode the code lengths ---
+    const combined = new Uint8Array(hlit + hdist);
+    combined.set(litLenLens.subarray(0, hlit));
+    combined.set(distLens.subarray(0, hdist), hlit);
+    const clSymbols = [];
+    const clFreqs = new Uint32Array(19);
+    for (let i = 0; i < combined.length;) {
+        const val = combined[i];
+        if (val === 0) {
+            let run = 1;
+            while (i + run < combined.length && combined[i + run] === 0) {
+                run++;
+            }
+            while (run > 0) {
+                if (run >= 11) {
+                    const repeat = Math.min(run, 138);
+                    clSymbols.push({ sym: 18, extra: repeat - 11, extraBits: 7 });
+                    clFreqs[18]++;
+                    run -= repeat;
+                    i += repeat;
+                }
+                else if (run >= 3) {
+                    const repeat = Math.min(run, 10);
+                    clSymbols.push({ sym: 17, extra: repeat - 3, extraBits: 3 });
+                    clFreqs[17]++;
+                    run -= repeat;
+                    i += repeat;
+                }
+                else {
+                    clSymbols.push({ sym: 0, extra: 0, extraBits: 0 });
+                    clFreqs[0]++;
+                    run--;
+                    i++;
+                }
+            }
+        }
+        else {
+            clSymbols.push({ sym: val, extra: 0, extraBits: 0 });
+            clFreqs[val]++;
+            i++;
+            let run = 0;
+            while (i + run < combined.length && combined[i + run] === val) {
+                run++;
+            }
+            while (run >= 3) {
+                const repeat = Math.min(run, 6);
+                clSymbols.push({ sym: 16, extra: repeat - 3, extraBits: 2 });
+                clFreqs[16]++;
+                run -= repeat;
+                i += repeat;
+            }
+            while (run > 0) {
+                clSymbols.push({ sym: val, extra: 0, extraBits: 0 });
+                clFreqs[val]++;
+                run--;
+                i++;
+            }
+        }
+    }
+    // --- Step 5: Build code-length Huffman tree ---
+    const clLens = buildCodeLengths(clFreqs, 7);
+    const clCodes = buildCanonicalCodes(clLens);
+    let hclen = 19;
+    while (hclen > 4 && clLens[CODE_LENGTH_ORDER[hclen - 1]] === 0) {
+        hclen--;
+    }
+    // --- Step 6: Write block header ---
+    out.writeBits(isFinal ? 1 : 0, 1); // BFINAL
+    out.writeBits(2, 2); // BTYPE = 10 (dynamic Huffman)
+    out.writeBits(hlit - 257, 5);
+    out.writeBits(hdist - 1, 5);
+    out.writeBits(hclen - 4, 4);
+    for (let i = 0; i < hclen; i++) {
+        out.writeBits(clLens[CODE_LENGTH_ORDER[i]], 3);
+    }
+    for (const cls of clSymbols) {
+        const [clCode, clLen] = clCodes[cls.sym];
+        out.writeBitsReverse(clCode, clLen);
+        if (cls.extraBits > 0) {
+            out.writeBits(cls.extra, cls.extraBits);
+        }
+    }
+    // --- Step 7: Write compressed data ---
+    for (const sym of symbols) {
+        if (sym.dist === 0) {
+            const [lCode, lLen] = litLenCodes[sym.litOrLen];
+            out.writeBitsReverse(lCode, lLen);
+        }
+        else {
+            const ls = getLengthSymbol(sym.litOrLen);
+            const [lCode, lLen] = litLenCodes[ls.code];
+            out.writeBitsReverse(lCode, lLen);
+            if (ls.extraBits > 0) {
+                out.writeBits(ls.extra, ls.extraBits);
+            }
+            const ds = getDistSymbol(sym.dist);
+            const [dCode, dLen] = distCodes[ds.code];
+            out.writeBitsReverse(dCode, dLen);
+            if (ds.extraBits > 0) {
+                out.writeBits(ds.extra, ds.extraBits);
+            }
+        }
     }
+    // End of block
+    const [eobCode, eobLen] = litLenCodes[256];
+    out.writeBitsReverse(eobCode, eobLen);
 }
-// ============================================================================
-// Stateful Streaming Deflater
-// ============================================================================
-/** Maximum LZ77 sliding window size (32 KB per RFC 1951). */
-const WINDOW_SIZE = 32768;
 /**
- * Stateful synchronous DEFLATE compressor.
- *
- * Unlike `deflateRawCompressed` (which is a one-shot function), this class
- * maintains state across multiple `write()` calls:
- *
- *  - **LZ77 sliding window**: back-references can span across chunks.
- *  - **Hash chains**: match positions persist across chunks with typed-array
- *    hash tables for fast lookup.
- *  - **Lazy matching**: each match is compared with the next position's match
- *    to pick the longer one.
- *  - **Bit writer**: bit position is preserved, so consecutive blocks form
- *    a single valid DEFLATE bit-stream without alignment issues.
+ * Run LZ77 match-finding on `data[start..end)`.
  *
- * Each `write()` emits one non-final fixed-Huffman block (BFINAL=0).
- * `finish()` emits a final empty block (BFINAL=1) and returns the tail bytes.
+ * When `state` is null, performs one-shot compression with fresh hash tables.
+ * When `state` is provided, maintains sliding window and hash chains across calls.
  *
- * This is the pure-JS equivalent of Node.js `zlib.deflateRawSync` with
- * `Z_SYNC_FLUSH`, used by the streaming ZIP writer (`pushSync`) to achieve
- * constant-memory streaming in both Node.js and browsers.
+ * Returns an array of LZ77 symbols (literals + length/distance pairs).
  */
-export class SyncDeflater {
-    constructor() {
-        this._output = new BitWriter();
-        // Hash chain tables — shared across chunks for cross-chunk matching.
-        this._head = new Int32Array(HASH_SIZE);
-        this._prev = new Int32Array(MAX_DIST);
-        /** Sliding window: the last WINDOW_SIZE bytes of uncompressed data. */
-        this._window = new Uint8Array(WINDOW_SIZE);
-        /** Number of valid bytes currently in the window. */
-        this._windowLen = 0;
-        /** Total bytes written so far (monotonically increasing; used for hash offsets). */
-        this._totalIn = 0;
-        // Lazy matching state that may span across chunks.
-        this._hasPrevMatch = false;
-        this._prevMatchLen = 0;
-        this._prevMatchDist = 0;
-        this._prevLiteral = 0;
+function lz77Compress(data, start, end, config, state) {
+    const symbols = [];
+    const maxChainLen = config.maxChainLen;
+    const goodLen = config.goodLen;
+    const niceLen = config.niceLen;
+    const useLazy = config.lazy;
+    let head;
+    let prevArr;
+    let window;
+    let wLen;
+    let totalIn;
+    let hasPrevMatch;
+    let prevMatchLen;
+    let prevMatchDist;
+    let prevLiteral;
+    if (state) {
+        head = state.head;
+        prevArr = state.prev;
+        window = state.window;
+        wLen = state.windowLen;
+        totalIn = state.totalIn;
+        hasPrevMatch = state.hasPrevMatch;
+        prevMatchLen = state.prevMatchLen;
+        prevMatchDist = state.prevMatchDist;
+        prevLiteral = state.prevLiteral;
     }
-    /**
-     * Compress a chunk and return the compressed bytes produced so far.
-     * The output is a valid prefix of a DEFLATE stream (one or more non-final blocks).
-     */
-    write(data) {
-        if (data.length === 0) {
-            return new Uint8Array(0);
+    else {
+        head = new Int32Array(HASH_SIZE);
+        prevArr = new Int32Array(MAX_DIST);
+        window = null;
+        wLen = 0;
+        totalIn = 0;
+        hasPrevMatch = false;
+        prevMatchLen = 0;
+        prevMatchDist = 0;
+        prevLiteral = 0;
+    }
+    const getByte = state
+        ? (globalPos) => {
+            const localPos = globalPos - totalIn;
+            if (localPos >= start && localPos < end) {
+                return data[localPos];
+            }
+            return window[globalPos & (MAX_DIST - 1)];
         }
-        const out = this._output;
-        // Start a non-final fixed-Huffman block
-        out.writeBits(0, 1); // BFINAL = 0
-        out.writeBits(1, 2); // BTYPE  = 01 (fixed Huffman)
-        const window = this._window;
-        let wLen = this._windowLen;
-        const head = this._head;
-        const prevArr = this._prev;
-        const totalIn = this._totalIn;
-        let hasPrevMatch = this._hasPrevMatch;
-        let prevMatchLen = this._prevMatchLen;
-        let prevMatchDist = this._prevMatchDist;
-        let prevLiteral = this._prevLiteral;
-        /**
-         * Insert a global position into the hash chain and the sliding window.
-         */
-        const insertHash = (localPos) => {
-            if (localPos + 2 >= data.length) {
+        : (globalPos) => data[globalPos];
+    const insertHash = state
+        ? (localPos) => {
+            if (localPos + 2 >= end) {
                 return;
             }
             const h = hash3(data[localPos], data[localPos + 1], data[localPos + 2]);
-            const globalPos = totalIn + localPos;
-            prevArr[globalPos & (MAX_DIST - 1)] = head[h];
-            head[h] = globalPos + 1; // 1-based
+            const gp = totalIn + localPos;
+            prevArr[gp & (MAX_DIST - 1)] = head[h];
+            head[h] = gp + 1;
+        }
+        : (localPos) => {
+            if (localPos + 2 >= end) {
+                return;
+            }
+            const h = hash3(data[localPos], data[localPos + 1], data[localPos + 2]);
+            prevArr[localPos & (MAX_DIST - 1)] = head[h];
+            head[h] = localPos + 1;
         };
-        const insertWindow = (localPos, count) => {
+    const insertWindow = state
+        ? (localPos, count) => {
             for (let i = 0; i < count; i++) {
-                window[(wLen + i) & (WINDOW_SIZE - 1)] = data[localPos + i];
+                window[(wLen + i) & (MAX_DIST - 1)] = data[localPos + i];
             }
             wLen += count;
-        };
-        let pos = 0;
-        for (; pos < data.length;) {
-            let bestLen = 0;
-            let bestDist = 0;
-            if (pos + 2 < data.length) {
-                const h = hash3(data[pos], data[pos + 1], data[pos + 2]);
-                const globalPos = totalIn + pos;
-                // Walk the hash chain
-                let chainLen = MAX_CHAIN_LEN;
-                let matchHead = head[h];
-                while (matchHead > 0 && chainLen-- > 0) {
-                    const mGlobalPos = matchHead - 1;
-                    const dist = globalPos - mGlobalPos;
-                    if (dist > MAX_DIST || dist <= 0) {
-                        break;
-                    }
-                    // Compare bytes through the sliding window + current chunk
-                    const maxLen = Math.min(MAX_MATCH, data.length - pos);
-                    let len = 0;
-                    // Quick reject on the byte beyond current bestLen
-                    if (bestLen >= MIN_MATCH) {
-                        const checkOffset = mGlobalPos + bestLen;
-                        // Determine the byte at checkOffset
-                        let checkByte;
-                        const checkLocal = checkOffset - totalIn;
-                        if (checkLocal >= 0 && checkLocal < data.length) {
-                            checkByte = data[checkLocal];
-                        }
-                        else {
-                            checkByte = window[checkOffset & (WINDOW_SIZE - 1)];
-                        }
-                        if (checkByte !== data[pos + bestLen]) {
-                            matchHead = prevArr[mGlobalPos & (MAX_DIST - 1)];
-                            continue;
-                        }
+        }
+        : (_localPos, _count) => { };
+    let pos = start;
+    for (; pos < end;) {
+        let bestLen = 0;
+        let bestDist = 0;
+        if (pos + 2 < end) {
+            const h = hash3(data[pos], data[pos + 1], data[pos + 2]);
+            const globalPos = state ? totalIn + pos : pos;
+            // When we already have a good match from a previous lazy evaluation,
+            // reduce the chain search length (matching zlib's good_length behavior).
+            let chainRemaining = useLazy && hasPrevMatch && prevMatchLen >= goodLen ? maxChainLen >> 2 : maxChainLen;
+            let matchHead = head[h];
+            while (matchHead > 0 && chainRemaining-- > 0) {
+                const mGlobalPos = matchHead - 1;
+                const dist = globalPos - mGlobalPos;
+                if (dist > MAX_DIST || dist <= 0) {
+                    break;
+                }
+                if (bestLen >= MIN_MATCH) {
+                    const checkGlobal = mGlobalPos + bestLen;
+                    if (getByte(checkGlobal) !== data[pos + bestLen]) {
+                        matchHead = prevArr[mGlobalPos & (MAX_DIST - 1)];
+                        continue;
                     }
-                    while (len < maxLen) {
-                        const matchOffset = mGlobalPos + len;
-                        // Get byte from window or current data
-                        let matchByte;
-                        const matchLocal = matchOffset - totalIn;
-                        if (matchLocal >= 0 && matchLocal < data.length) {
-                            matchByte = data[matchLocal];
-                        }
-                        else {
-                            matchByte = window[matchOffset & (WINDOW_SIZE - 1)];
-                        }
-                        if (matchByte !== data[pos + len]) {
-                            break;
-                        }
-                        len++;
+                }
+                const maxLen = Math.min(MAX_MATCH, end - pos);
+                let len = 0;
+                while (len < maxLen) {
+                    if (getByte(mGlobalPos + len) !== data[pos + len]) {
+                        break;
                     }
-                    if (len > bestLen) {
-                        bestLen = len;
-                        bestDist = dist;
-                        if (len >= MAX_MATCH) {
-                            break;
-                        }
+                    len++;
+                }
+                if (len > bestLen) {
+                    bestLen = len;
+                    bestDist = dist;
+                    if (len >= niceLen) {
+                        break;
                     }
-                    matchHead = prevArr[mGlobalPos & (MAX_DIST - 1)];
                 }
-                // Insert current position into hash chain
+                matchHead = prevArr[mGlobalPos & (MAX_DIST - 1)];
+            }
+            if (state) {
                 prevArr[globalPos & (MAX_DIST - 1)] = head[h];
                 head[h] = globalPos + 1;
             }
-            // --- Lazy matching logic ---
-            if (hasPrevMatch) {
-                if (bestLen > prevMatchLen) {
-                    // Current position wins — emit previous as literal
-                    writeLiteralCode(out, prevLiteral);
-                    prevMatchLen = bestLen;
-                    prevMatchDist = bestDist;
-                    prevLiteral = data[pos];
-                    insertWindow(pos, 1);
-                    pos++;
-                }
-                else {
-                    // Previous match wins — emit it
-                    writeLengthCode(out, prevMatchLen);
-                    writeDistanceCode(out, prevMatchDist);
-                    // Insert hash entries for skipped positions inside the match
-                    const matchEnd = pos - 1 + prevMatchLen;
-                    const insertEnd = Math.min(matchEnd, data.length);
-                    for (let i = pos; i < insertEnd; i++) {
-                        insertHash(i);
-                    }
-                    insertWindow(pos, insertEnd - pos);
-                    pos = insertEnd;
-                    hasPrevMatch = false;
-                    prevMatchLen = 0;
-                }
+            else {
+                prevArr[pos & (MAX_DIST - 1)] = head[h];
+                head[h] = pos + 1;
             }
-            else if (bestLen >= MIN_MATCH) {
-                hasPrevMatch = true;
+        }
+        if (useLazy && hasPrevMatch) {
+            if (bestLen > prevMatchLen) {
+                symbols.push({ litOrLen: prevLiteral, dist: 0 });
                 prevMatchLen = bestLen;
                 prevMatchDist = bestDist;
                 prevLiteral = data[pos];
@@ -858,38 +1038,132 @@ export class SyncDeflater {
                 pos++;
             }
             else {
-                writeLiteralCode(out, data[pos]);
+                symbols.push({ litOrLen: prevMatchLen, dist: prevMatchDist });
+                const matchEnd = Math.min(pos - 1 + prevMatchLen, end);
+                for (let i = pos; i < matchEnd; i++) {
+                    insertHash(i);
+                }
+                insertWindow(pos, matchEnd - pos);
+                pos = matchEnd;
+                hasPrevMatch = false;
+                prevMatchLen = 0;
+            }
+        }
+        else if (bestLen >= MIN_MATCH) {
+            if (useLazy) {
+                hasPrevMatch = true;
+                prevMatchLen = bestLen;
+                prevMatchDist = bestDist;
+                prevLiteral = data[pos];
                 insertWindow(pos, 1);
                 pos++;
             }
+            else {
+                symbols.push({ litOrLen: bestLen, dist: bestDist });
+                const matchEnd = Math.min(pos + bestLen, end);
+                for (let i = pos + 1; i < matchEnd; i++) {
+                    insertHash(i);
+                }
+                insertWindow(pos, matchEnd - pos);
+                pos = matchEnd;
+            }
         }
-        // If there's a pending lazy match and we're at chunk boundary,
-        // flush it now (the next chunk will start fresh for lazy matching).
-        if (hasPrevMatch) {
-            writeLengthCode(out, prevMatchLen);
-            writeDistanceCode(out, prevMatchDist);
-            // The pending match started at pos-1 and covers prevMatchLen bytes.
-            // pos-1 was already hashed/windowed when it was first encountered;
-            // now insert the remaining positions (pos .. pos-1+prevMatchLen-1)
-            // into hash chains and the sliding window so the next chunk can
-            // reference them.
-            const matchEnd = Math.min(pos - 1 + prevMatchLen, data.length);
-            for (let i = pos; i < matchEnd; i++) {
-                insertHash(i);
-            }
-            insertWindow(pos, matchEnd - pos);
-            hasPrevMatch = false;
-            prevMatchLen = 0;
-        }
-        // End-of-block symbol
-        writeLiteralCode(out, 256);
-        this._windowLen = wLen;
-        this._totalIn = totalIn + data.length;
-        this._hasPrevMatch = hasPrevMatch;
-        this._prevMatchLen = prevMatchLen;
-        this._prevMatchDist = prevMatchDist;
-        this._prevLiteral = prevLiteral;
-        // Flush completed bytes from the bit writer
+        else {
+            if (hasPrevMatch) {
+                // Non-lazy mode shouldn't reach here, but handle gracefully
+                symbols.push({ litOrLen: prevMatchLen, dist: prevMatchDist });
+                hasPrevMatch = false;
+                prevMatchLen = 0;
+            }
+            symbols.push({ litOrLen: data[pos], dist: 0 });
+            insertWindow(pos, 1);
+            pos++;
+        }
+    }
+    // Flush pending lazy match
+    if (hasPrevMatch) {
+        symbols.push({ litOrLen: prevMatchLen, dist: prevMatchDist });
+        const matchEnd = Math.min(pos - 1 + prevMatchLen, end);
+        for (let i = pos; i < matchEnd; i++) {
+            insertHash(i);
+        }
+        insertWindow(pos, matchEnd - pos);
+        hasPrevMatch = false;
+        prevMatchLen = 0;
+    }
+    if (state) {
+        state.windowLen = wLen;
+        state.totalIn = totalIn + (end - start);
+        state.hasPrevMatch = hasPrevMatch;
+        state.prevMatchLen = prevMatchLen;
+        state.prevMatchDist = prevMatchDist;
+        state.prevLiteral = prevLiteral;
+    }
+    return symbols;
+}
+// ============================================================================
+// Stateful Streaming Deflater
+// ============================================================================
+/**
+ * Stateful synchronous DEFLATE compressor with Dynamic Huffman encoding.
+ *
+ * Unlike `deflateRawCompressed` (which is a one-shot function), this class
+ * maintains state across multiple `write()` calls:
+ *
+ *  - **LZ77 sliding window**: back-references can span across chunks.
+ *  - **Hash chains**: match positions persist across chunks with typed-array
+ *    hash tables for fast lookup.
+ *  - **Lazy matching**: configurable per compression level.
+ *  - **Dynamic Huffman**: each block builds optimal Huffman trees from
+ *    actual symbol frequencies (BTYPE=2), producing significantly smaller
+ *    output than fixed Huffman (BTYPE=1).
+ *  - **Bit writer**: bit position is preserved, so consecutive blocks form
+ *    a single valid DEFLATE bit-stream without alignment issues.
+ *
+ * Each `write()` emits one non-final Dynamic Huffman block (BFINAL=0).
+ * `finish()` emits a final empty fixed-Huffman block (BFINAL=1).
+ *
+ * This is the pure-JS equivalent of Node.js `zlib.deflateRawSync` with
+ * `Z_SYNC_FLUSH`, used by the streaming ZIP writer (`pushSync`) to achieve
+ * constant-memory streaming in both Node.js and browsers.
+ *
+ * @param level - Compression level (0-9). Level 0 emits STORE blocks.
+ *                Default: 6 (matching zlib default).
+ */
+export class SyncDeflater {
+    constructor(level = 6) {
+        this._output = new BitWriter();
+        this._state = {
+            head: new Int32Array(HASH_SIZE),
+            prev: new Int32Array(MAX_DIST),
+            window: new Uint8Array(MAX_DIST),
+            windowLen: 0,
+            totalIn: 0,
+            hasPrevMatch: false,
+            prevMatchLen: 0,
+            prevMatchDist: 0,
+            prevLiteral: 0
+        };
+        this._level = Math.max(0, Math.min(9, level));
+        this._config = getLZ77Config(this._level);
+    }
+    /**
+     * Compress a chunk and return the compressed bytes produced so far.
+     * The output is a valid prefix of a DEFLATE stream (one or more non-final blocks).
+     */
+    write(data) {
+        if (data.length === 0) {
+            return new Uint8Array(0);
+        }
+        const out = this._output;
+        if (this._level === 0) {
+            // Store mode: emit uncompressed block(s)
+            this._writeStore(data);
+            return out.flushBytes();
+        }
+        // LZ77 + Dynamic Huffman
+        const symbols = lz77Compress(data, 0, data.length, this._config, this._state);
+        emitDynamicBlock(out, symbols, false);
         return out.flushBytes();
     }
     /**
@@ -900,8 +1174,38 @@ export class SyncDeflater {
         const out = this._output;
         // Final block: BFINAL=1, BTYPE=01, immediately followed by EOB (symbol 256)
         out.writeBits(1, 1); // BFINAL = 1
-        out.writeBits(1, 2); // BTYPE  = 01
+        out.writeBits(1, 2); // BTYPE  = 01 (fixed Huffman)
         writeLiteralCode(out, 256);
         return out.finish();
     }
+    /**
+     * Write STORE (uncompressed) blocks for level=0.
+     * Each block is non-final (BFINAL=0); the final block is emitted by finish().
+     */
+    _writeStore(data) {
+        const out = this._output;
+        const MAX_BLOCK_SIZE = 65535;
+        let offset = 0;
+        while (offset < data.length) {
+            const remaining = data.length - offset;
+            const blockSize = Math.min(MAX_BLOCK_SIZE, remaining);
+            // Align to byte boundary before stored block header
+            out.alignToByte();
+            out.writeBits(0, 1); // BFINAL = 0 (never final; finish() handles that)
+            out.writeBits(0, 2); // BTYPE = 00 (stored)
+            // Align to byte boundary after block header (3 bits → pad to 8)
+            out.alignToByte();
+            // LEN
+            out.writeBits(blockSize & 0xff, 8);
+            out.writeBits((blockSize >> 8) & 0xff, 8);
+            // NLEN
+            out.writeBits(~blockSize & 0xff, 8);
+            out.writeBits((~blockSize >> 8) & 0xff, 8);
+            // Data
+            for (let i = 0; i < blockSize; i++) {
+                out.writeBits(data[offset + i], 8);
+            }
+            offset += blockSize;
+        }
+    }
 }