Note

• This is a consolidated version of compress/lzw that supports GIF, TIFF and PDF.
• Please refer to this golang proposal for details.
• github.com/hhrutter/tiff uses this package to extend x/image/tiff.
• pdfcpu uses this package for processing PDFs with embedded TIFF images.

Background

• PDF's LZWDecode filter comes with the optional parameter EarlyChange.
• The type of this parameter is int and the defined values are 0 and 1.
• The default value is 1.

This parameter implies two variants of lzw. (See the PDF spec).

compress/lzw:

• the algorithm implied by EarlyChange value 1
• provides both Reader and Writer.

x/image/tiff/lzw:

• the algorithm implied by EarlyChange value 0
• provides a Reader, lacks a Writer

In addition PDF expects a leading clear_table marker right at the beginning which is not something compress/lzw takes into account.

There are numerous PDF Writers out there and for arbitrary PDF files using the LZWDecode filter the following can be observed:

• Some PDF writers do not write the EOD (end of data) marker.
• Some PDF writers do not write the final bits after the EOD marker.

Goal

An extended version of compress/lzw with reliable support for GIF, TIFF and PDF.