Hawkynt

LZW (Lempel-Ziv-Welch)

Dictionary-based compression algorithm that builds a table of frequently occurring strings, starting from a dictionary of all single bytes and adding new patterns dynamically. Emits a pure variable-width (9-16 bit) LZW bitstream, LSB-first packed, with clear and stop codes and no length header - matching CompressionWorkbench’s BB_Lzw building block. Used in GIF/TIFF formats.

Properties

Property	Value
Category	Compression Algorithms
Sub-category	Dictionary
Security status	Not classified
Complexity	Advanced
Inventor	Terry Welch
Year	1984
Origin	🇺🇸 United States
Source	algorithms/compression/lzw.js

Security

Status: not classified — treat as unverified.

No vulnerabilities are recorded for this implementation.

Documentation

• A Technique for High-Performance Data Compression
• LZW - Wikipedia
• GIF Format Specification

References

• Original IEEE Paper by Terry Welch
• TIFF LZW Reference
• Educational LZW Implementation

Test vectors

6 vectors ship with this algorithm and run in the test suite. Byte values are hexadecimal.

Vector 1 — Empty input - still emits clear code and stop code

Field	Value
input	(empty)
expected	000302

Vector 2 — Single byte - no dictionary matches possible

Field	Value
input	41
expected	00830404

Vector 3 — Repeated two-character pattern - Wikipedia example

Field	Value
input	41424142414241424142
expected	008308114870a09080

Vector 4 — Classic Shakespeare-inspired LZW test string - Wikipedia

Field	Value
input	544f42454f524e4f54544f42454f52544f42454f524e4f54
expected	00a93c1152e48914274fa80824687061c183090302

Vector 5 — Maximum redundancy test (all identical characters)

Source: Edge case - maximum compression ratio

Field	Value
input	41414141414141414141
expected	0083081c483020

Vector 6 — All 256 byte values 0..255 once each - no repetition

Source: Edge case - minimal compression benefit

Field	Value
input	000102030405060708090a0b0c0d0e0f 101112131415161718191a1b1c1d1e1f 202122232425262728292a2b2c2d2e2f 303132333435363738393a3b3c3d3e3f 404142434445464748494a4b4c4d4e4f 505152535455565758595a5b5c5d5e5f 606162636465666768696a6b6c6d6e6f 707172737475767778797a7b7c7d7e7f 808182838485868788898a8b8c8d8e8f 909192939495969798999a9b9c9d9e9f a0a1a2a3a4a5a6a7a8a9aaabacadaeaf b0b1b2b3b4b5b6b7b8b9babbbcbdbebf c0c1c2c3c4c5c6c7c8c9cacbcccdcecf d0d1d2d3d4d5d6d7d8d9dadbdcdddedf e0e1e2e3e4e5e6e7e8e9eaebecedeeef f0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
expected	00010410308040010307102450b08041 03070f204490308142050b173064d0b0 8143070f1f40841031824409132750a4 50b182450b172f60c4903183460d1b37 … (291 bytes; the full value is in the source)

---

← All algorithms

This site is open source. Improve this page.