Changelog

v1.0.1 (2026-09-21)

Bug fixes

• Premailex.HTMLParser.Meeseeks.parse/1 no longer raises MatchError for document fragments that contains head level content such as <style>, <meta>, <link>, or <title>
• Premailex.DOM.traverse_with_matching_items/3 now preserves the order of appearance for matching selector items

v1.0.0 (2026-05-24)

Requires Elixir 1.15 or higher.

This release contains significant performance improvements, with typical HTML emails seeing a 30x+ speedup when inlining styles. Premailex is also now zero dependency thanks to the new fallback HTML parser Premailex.HTMLParser.Xmerl.

Architecture changes

The layers between parser, DOM operations, and the top-level API have been reshaped:

• Premailex.HTMLParser is now a thin behaviour with only Premailex.HTMLParser.parse/1 and Premailex.HTMLParser.to_html/1 callbacks
• Premailex.DOM is a new module that handles all selector matching, traversal, and tree manipulation
• Premailex.Util has been removed with functions moved into Premailex.DOM
• Premailex.HTMLInlineStyles.process/2 is now a pure tree to tree transformation that accepts an explicit list of CSS rules

Breaking changes

• Premailex.HTMLInlineStyles.process/3 no longer exposed, use Premailex.HTMLInlineStyles.process/2
• Premailex.HTMLToPlainText.process/1 no longer accepts HTML string
• Premailex.HTMLParser behaviour callback to_string renamed to to_html
• Premailex.HTMLParser behaviour no longer requires all, filter, or text callbacks
• Premailex.HTMLParser no longer exposes parse, use Premailex.parse/2 instead
• Premailex.HTMLParser no longer exposes to_string, use Premailex.to_html/2 instead
• Premailex.HTMLParser no longer exposes all, use Premailex.DOM.all/2 instead
• Premailex.HTMLParser no longer exposes filter, use Premailex.DOM.reject/2 instead
• Premailex.HTMLParser no longer exposes text, use Premailex.DOM.text_content/1 instead
• Premailex.Util has been removed:
  • Premailex.Util.traverse/3 is now Premailex.DOM.replace_all_matches/3
  • Premailex.Util.traverse_until_first/3 is now Premailex.DOM.replace_first_match/3
  • Premailex.Util.traverse_reduce/3 is removed, use Premailex.DOM.traverse_with_matching_items/3 for indexed single-walk updates
• Renamed Premailex.CSSParser.parse_rules/1 to Premailex.CSSParser.parse_declaration_block/1
• Renamed Premailex.CSSParser.merge/1 to Premailex.CSSParser.cascade/1
• Premailex.HTTPAdapter.request/5 callback no longer requires the Premailex.HTTPAdapter.HTTPResponse struct in favor of using a map
• Premailex.parse/2 now always returns a list
• Floki minimum version bumped from ~> 0.19 to ~> 0.24
• Premailex.to_inline_css/2 :optimize option has been replaced by a single boolean option :remove_style_tags

Additions

• Added Premailex.parse/2 and Premailex.to_html/2
• Added support for LazyHTML
• Added fallback support for :xmerl
• Added Premailex.CSSParser.split_selector_groups/1 for selector group splitting
• Added Premailex.DOM.traverse_with_matching_items/3 for indexed single-walk tree updates
• Added support for structural pseudo-classes: :first-child, :last-child, :only-child, :last-of-type, :only-of-type, :nth-child, :nth-of-type, :nth-last-child, :nth-last-of-type, :empty, :root (An+B, odd, and even arguments supported)

Other

• Floki is now optional
• Premailex.CSSParser rewritten and no longer uses regular expressions to parse CSS
• Premailex.CSSParser.parse_declaration_block/1 now does case insensitive, terminal !important detection
• Premailex.HTMLInlineStyles.process/2 tree traversal performance changed from O(N^2) to O(N)
• Fixed compiler warnings in Premailex.HTMLParser.Meeseeks