parse

• Add many hashes for CSS
• Add amp-boilerplate hash for HTML
• NewError now has fmt-style function signature
• Append 0x00 to signal EOF even if source already ends in 0x00 (which may be valid)

CSS: add Invert and Solid hashes

• XML and HTML: Text() []byte returns nil for start tag closers
• XML and HTML: Text() []byte returns textual content for TextToken, CommentToken, ...
• Added Offset() int for all lexers and parsers that returns the current character offset in the input stream
• CSS: EOF after \ now properly handled as DelimToken and not EscapeToken in some cases

Bugfix: bad URL encoding resulted in no decoding at all, the new EncodeURL and DecodeURL are faster and never fail
Bugfix: get correct Position if it is in the middle of a unicode codepoint

• Re-parse input immediately when encountering parse error. Previously this was only done if the error was actually read which would save us from re-parsing the file (i.e. better performance). However, an error is (a) rare and (b) happens only once per file. Re-parsing on errors does not impact the performance of well formatted files, and a reliable error message is valued more.
• Bugfix: prevent infinite loop on error on input file that contains unicode code points (i.e. any character bigger than 0xC0 usually followed by more bytes).
• Improve error messages for parsers to include parser name and print offending byte(s)

• Improve performance of ReplaceMultipleWhitespace by 20%--25%
• Add ReplaceEntities and ReplaceMultipleWhitespaceAndEntities to replace QuoteEntity. These allow to do replacements such as: " => ", " => ", " => ", ’ => ’, ' => '.
• Update list of HTML entity rewrites

Revert v2.3.11 and readd html.Keygen hash.

Add ReplaceEntities and remove entity replacement from EscapeAttrVal

Add and remove hashes for CSS and HTML

CSS: keep whitespace in unknown at rules

• Remove import comments
• Fix bug in CSS parsing when encountering } in CSS inline

• Fix position information on errors
• CSS: fix error position

• Prevent panic when returning Bytes() on empty buffer
• Set proper context in errors
• Limit the length in context in errors
• In EqualFold only match upper/lower case on alphabet characters
• Add option to EscapeAttrVal to handle attribute values for XML (use double quotes)
• HTML: add hashes for RDFa attribute names
• JS: accept NULL characters as code
• JSON: report errors NULL characters

• CSS: fix parsing of bad declarations
• CSS: handle comments within declarations
• CSS: reintroduce IE star hack *color: white
• CSS: add keywords to hash table

Add Go modules support by using the github.com/tdewolff/parse/v2 import path.

This fixes a couple of bugs in the HTML, XML and JS implementations that caused out-of-bounds panics. The HTML parser now allows NULL bytes to be in the source.

• Use a new buffer that holds all data in memory, giving significant speed-ups for parsing
• Speed-up JSON consumeString
• Break CSS selectors at the commas, this reduces memory requirements for (mostly) SASS/LESS generated CSS
• Add position tracking upon errors, showing line number and column in the error message

This v2.2 release is synchronous with tdewolff/minify v2.2

Changelog:

• JS: improved parsing of edge cases
• JS: allow HTML inline comments
• CSS: improved parsing of bad declarations
• CSS: parse custom properties

Thanks to @RReverser for improvements on the JS parser.