The lexing step is usually sufficiently described by a list of (pattern,
action) pairs (maybe with priorization)
re"if\b": keywordIf
re"[A-Za-z_]+[A-Za-z_0-9]*": identifier
re"\d+": number
Run
So it's not the regexes per se that are the problem. The problem comes from the
fact that the typical regex library has no support for this setup at all and
instead makes you write `if|[A-Za-z_]+[A-Za-z_0-9]*|\d+` with no support to
distinguish between the different cases.
A different but related problem is that regexes are not used only for the
lexing step but also to directly extract data and skipping the parsing step
which is terrible for correctness.
