Optimize single-character alternatives

[RunDevelopment]

Is your feature request related to a problem? Please describe.

BNF grammars (including dialects) commonly denote large character sets like this

letter = "A" | "B" | "C" | "D" | "E" | "F" | "G"
       | "H" | "I" | "J" | "K" | "L" | "M" | "N"
       | "O" | "P" | "Q" | "R" | "S" | "T" | "U"
       | "V" | "W" | "X" | "Y" | "Z" | "a" | "b"
       | "c" | "d" | "e" | "f" | "g" | "h" | "i"
       | "j" | "k" | "l" | "m" | "n" | "o" | "p"
       | "q" | "r" | "s" | "t" | "u" | "v" | "w"
       | "x" | "y" | "z" ;
digit = "0" | "1" | "2" | "3" | "4" | "5" | "6" | "7" | "8" | "9" ;
symbol = "[" | "]" | "{" | "}" | "(" | ")" | "<" | ">"
       | "'" | '"' | "=" | "|" | "." | "," | ";" ;
character = letter | digit | symbol | "_" ;

(More examples on Wikipedia)

While there are of course better ways to denote character ranges in pomsky, the union of character sets and list of characters (as in character in the above example) is quite common. However, pomsky currently produces quite suboptimal regexes for this pattern.

Example

Describe the solution you'd like

Please merge adjacent single-character alternatives into one character set. E.g. a|b|c -> [abc].

This optimization is particularly useful because it enabled further optimizations within character sets.

Additional context

For a reference implementation of this optimization, checkout the regexp/prefer-character-class rule. Note that this rule also does some interesting analysis to merge non-adjacent single-character alternatives.

[Aloso]

Thanks for the feature request!

[black7375]

Here are some trasform rules & implements

• https://github.com/DmitrySoshnikov/regexp-tree/blob/master/src/optimizer/README.md#transforms
• https://bisqwit.iki.fi/source/regexopt.html
• https://metacpan.org/release/DANKOGAI/Regexp-Optimizer-0.15/view/lib/Regexp/Optimizer.pm