An ergonomic "reverse identifier"

[amake]

The readme has the example of a parser that matches an identifier:

final id = letter() & (letter() | digit()).star();

I happened to need something equivalent to a reverse identifier: something that can have foos and bars up until the end, where it must end with a foo.

I tried to do something like

final id = (letter() | digit()).star() & letter();

But I understand that this doesn't work because there is no backtracking, correct?

My instinct was to use starLazy but the best limit condition I could come up with is pretty gnarly:

    final id =
        (letter() | digit()).starLazy(letter() & ((letter() | digit()).not())) &
            letter();

Is there an idiomatic or egonomic way to write this? I did come up with the following, but I always feel like where is cheating :)

    final id = [letter(), digit()]
        .toChoiceParser()
        .plus()
        .where((values) => letter().accept(values.last));

(Using toChoiceParser instead of | so that the type of values.last is known.)

[renggli]

Your example with starLazy is what I would have suggested. Only thing I would do is to extract a helper method, to make it more readable, i.e.

Parser reverseId(Parser inner, Parser terminator) => 
    inner.starLazy(terminator & inner.not()) & terminator;

final id = reverseId(letter() | digit(), letter());

[amake]

Thanks, the refactor does make it a lot more readable!

[amake]

I have a new twist on this: Is it possible to match a "reverse ID" up to a trailing "inner" character? In other words:

• Reverse ID: 123a
• Reverse ID with trailing inner character: 123a0 (desire to match 123a and let 0 be matched by another, non-reverse-ID parser)

I am reimplementing a feature that was originally using regex and this case was handled by backtracking. I can't see a way to fix this without breaking other edge cases.

[amake]

(This is pretty abstract as written; if it would help I can share my real use case, which is not really a "reverse ID" but is isomorphic to it.)

[renggli]

Yeah, I am a bit confused with the non-reverse-ID parser. Maybe if you could share the regex, and/or provide some more examples of what inputs are accepted/rejected?