Composition
Two module functions build one expression from several. Both return an expression, with all its variations.
Regexes.sequence(*parts)
Each part, one after another. It reads as it would written out:
Regexes.sequence(Regexes.lowercase_letters, Regexes.word_characters.between(0, 23))
# => /[a-z][A-Za-z0-9_]{0,23}/An anchored part stays anchored inside the sequence.
Regexes.either(*alternatives)
Any one of the alternatives, as one unit:
ends = Regexes.either(Regexes.hyphen.one_or_more.leading, Regexes.hyphen.one_or_more.trailing)
'--acme-mail-'.gsub(ends, '') # => "acme-mail"Because it is one unit, repeating it repeats the choice, and anchoring it anchors the whole choice:
digit_or_letter = Regexes.either(Regexes.digits, Regexes.letters).whole
%w[1 a 1a].map { digit_or_letter.match?(it) } # => [true, true, false]Regexes.literal(text)
Text matched character for character, every character standing for itself:
Regexes.literal('b+') # => /b\+/
Regexes.literal('.ns.cloudflare.com').trailing # a Cloudflare nameserver, dots and allIt's the safe way to put a value into a pattern. Interpolating one (/#{suffix}\z/) lets its dots, pluses and brackets match more than they say; a literal can't. An empty literal raises ArgumentError, since it would match everywhere.
Options travel with their part
A case-insensitive part stays case-insensitive inside a case-sensitive whole. It is embedded with its options carried along ((?i-mx:…)), while a part with no options is embedded as its plain source:
link = Regexes.sequence(Regexes.http_scheme.case_insensitive, Regexes.lowercase_letters.one_or_more)
link.match?('HTTP://acme') # => true
link.match?('http://ACME') # => falseWhere your own formats go
The library holds shapes that recur. The grammar of one format — a bounce address, a log line, a handle — belongs in the code that owns it, as a named constant built from these:
class Handle
FORMAT = Regexes.sequence(Regexes.lowercase_letters, Regexes.word_characters.between(0, 23)).whole
end