Add a repSplit(p) function to Parser

[nicolas-zozol]

A simple exemple would be for bullet points

* bullet 1
* bullet 2
* bullet 3

If we use bullet().then(eol()).rep(), then the last eol() is eaten by the parser, so we can't separate easily bullet block by at least two eol().

So the idea would be bullet().repSplit(eol()) and would return a TupleParser<Bullet>.

Other option maybe more general but maybe less in the Parsec philosohy is to make

bullet().then(eol())
.rep() // eat last '\n'
.backdrop(1); // go back 1 character

It could be use in many more situations (including bad ones)

[d-plaindoux]

This should be written:

bullet.then(eol().then(bullet()).rep()

Indeed the rep parser can be generalized with start, join and end parsers.

[nicolas-zozol]

The back() function would not help, because for example if the document finishes at the * bullet3 without last \n, applying the back() function would move the cursor before the 3.

[nicolas-zozol]

For the record, here is a way to go back if ever needed.

const parser = C.string('this is wanted').then(C.char(';'));

const backParser = new Parser(function (input, index) {
    return parser.parse(input, index)    // parsing with regular parser
        .fold(    // then modifing the response
            accept =>
                response.accept(
                    accept.value,
                    accept.input,
                    accept.offset - 1,   // moving back of one offset
                    accept.consumed
                ),
            reject =>
                response.reject(
                    reject.input, reject.offset, reject.consumed
                )
        );
});

const resp = backParser.parse(Streams.ofString(line));