On: Rob Pike on regular expressions in lexing and parsing

[chwahoo]

Interesting that he claims that it's difficult to change a regular expression to cope with new requirements but easy to change a hand-coded loop.

In my experience, the opposite is true (up to the point when your grammar becomes non-regular).

[lukesandberg → chwahoo]

Many regular expression implementations don't deal properly with unicode character classes and collation. So adding something like unicode identifiers to go may be impossible or impractical with regexs whereas with a loop you can rely on more standard collation support.

[fooyc → lukesandberg]

Some libs such as PCRE can match by character properties, for example \pL would match any unicode letter.

Doing that in a loop seems to be much more difficult. And unless you work on utf-32 data, you have to handle variable-size encoding of characters.