akr / abnf Goto Github PK

print "Regexp for A to C: "
pp ABNF.regexp_tree(<<-'End').to_s
 a-to-c="abc"
End

Expected output: Regexp for A to C: "[abc]" or maybe ... "/[abc]/"

Observed output: Regexp for A to C: "(?i-m:abc)"

If I define two times the same entity, it extends the previous entity.

For example:

a = "a"
a = "b"

Expected result: "b", "a" or (preferably) error.

Result:

%r{[ab]}xi

Maybe this is correct or documented, however it doesn't match my expectation. (I was under the impression in the RFC you have to use a special operator for extension, not = but rather *= or something similar.)

Thanks heaps for this library by the way, it is gold!

ABNF is usually case sensitive, right?

While the standard suggests it is illegal to use string literals "a" / "b" and instead to escape them in to unreadable muck, many people use them anyway.

In these (common) cases, it is often correct to respect the input case.

.regex always seems to return /something/i.

I suppose this may be a workaround for bad input, but at least an option to stop it would be cute.