Recently I wrote a CSS tokenizer based on the css-syntax spec (the same standard that Blink now uses to parse CSS). The detailed description of my motivation is in the readme, but if I had to describe it in a sentence: The existing Haskell parsers were too high-level for me needs, they provided too much abstraction. The package does not implement the full parser, at least one token type is not implemented. But it should be able to tokenize most CSS files. I copied the tests from the Blink source. All but one pass (and of course the tests for the token types that are not implemented) While the spec describes the algorithm in a way that pretty much prescribes the implementation, I chose to use attoparsec instead. Not sure about the speed implications, but speed is not that important for me. The spec module also describes a parser. I did not have time to implement it, and am currently not planning to. But it would be nice if that was part of the package as well. I had hoped to be able to write a minimizer as well, but I'm not sure if it's possible to write a minimizer just on the token stream, maybe the full parser is needed for that. However, consecutive whitespace is collapsed so tokenizing and then serializing the file will in most cases lead to a smaller file. Hackage: http://hackage.haskell.org/package/css-syntax GitHub repo: https://github.com/wereHamster/haskell-css-syntax Spec: https://drafts.csswg.org/css-syntax