parse until a string is found

[pth14]

Hi,
I am still learning the lib.
I am trying to do something simple but I failed to do it so far: I am looking for a parser like noneof but for a string, not for a char or some chars.
In other words, I would like to skip all chars until a string is matched (this string should not be consumed, using attempt probably).

[louthy]

This suggests a poorly structured parser. You really should be moving forwards if you are able to parse something expected. Usually, you'd only backtrack when there is something unexpected in the stream and you need to try alternatives.

Perhaps if you expand your question to some more specifics, I can give you some better advice?

But, if you reaally want to do it, you can use manyUntil:

var p = manyUntil(anyChar, str("THING"));
var r = p.Parse("****THING").ToEither();          // [ '*', '*', '*', '*', '*' ]

Be warned though, it's a recursive operator - which means if the amount of text being skipped is too large, you'll probably get a stack overflow. You can see the implementation here.

[pth14]

I really have a hard time composing with "many" and "either".
I try to parse code like the following (I added line numbers):

1. /* SystemInterface.h
2. *      SystemInterface
3. *       
4. */
5.
6. #ifndef SYSTEMTYPEDEFS_H_
7. #define SYSTEMTYPEDEFS_H_
8.
9. #include <stdint.h>
10.
11. /* Project Parameters
12.  * *********************************************************************************************************************************************************/
13.
14. enum { e_1_1 = 0 , e_1_2 = 1,}; typedef uint16_t e1;
15. enum { e_2_1 = 0, e_2_2 = 1,}; typedef uint16_t e2;

I wrote 1 parser to skip comments, 1 to skip pre processor statements and 1 to parse enums.
So far so good.
If I applied it sequentially (i.e. comments parser then pre processor parser then enums parser ) and in right order, it works.
The problem is I don't know the order and how many times it repeats(like in lines 11-12 where comments repeat).
It can repeat anywhere so I wrote a parser like this:

EnumsParser =
  from e in many(either(spacesP, either(multiLinesCommentP, either(preProcessorP, EnumParser))))
  select e.Where(er => er.IsValid).AsEnumerable();

parsers ending in "P" are just parsers that return all the same type to allow to use "either" (IsValid property allow me to filter, getting only enum data).
I get this error: "unexpected many: combinator 'many' is applied to a parser that accepts an empty string."

[louthy]

spacesP presumably allows zero or more spaces, that means it will parse forever if you put it in a many (because parsing zero spaces will always succeed) - that's why you get the error. If you put something in a many it must parse something or fail, it can't successfully parse nothing. For example, using something like many1(satisfy(Char.IsWhiteSpace)) for spacesP, would only succeed when it parses one or more space.

You can also use choice instead of either when you have more than two options:

    many(choice(spacesP, multiLinesCommentP, preProcessorP, EnumParser))

[louthy]

btw, if you're looking for something to reference. I wrote a C# parser a few years back with LanguageExt.Parsec, for my documentation generator. It also tries to extract comments (for the documentation). Note, it will skip over the code blocks as it's mostly trying to extract type and member definitions (and their comments).

I deprecated it and switched to using Roslyn (as C# kept changing!), so this is a link to an old commit has.

[pth14]

spacesP presumably allows zero or more spaces, that means it will parse forever if you put it in a many (because parsing zero spaces will always succeed) - that's why you get the error. If you put something in a many it must parse something or fail, it can't successfully parse nothing. For example, using something like many1(satisfy(Char.IsWhiteSpace)) for spacesP, would only succeed when it parses one or more space.

thank you for your help, I really appreciate.
yes it is just a wrapper call to "spaces" that select a common type for using it with "either" (or "choice" as suggested):

var spacesP =
  from _s in spaces
  select new EnumRecord { IsValid = false };

Then I am confused, how should I skip lines 1 to 12 in my example, knowing that I should keep comments in lines like this, I mean I can't just ignore all comments, only line or multi lines comments:

static INTERFACE_t p_list [] =
{
    {ID_E_P_P_M  , (void*)&e_p_p_m,   sizeof(e_p_p_m)   , ID_FLOAT32, PROD}, //TEXT=""E: IPM"" DU=""ppm""
};

[louthy]

As I mentioned earlier, you should build a general case token parser. Something that takes a Parser<A> p, parses the p, and then strips the spaces after. This will allow you to automatically remove the gaps between tokens.

    Parser<A> token<A>(Parser<A> p) =>
        from r in p
        from _ in spaces
        select r;

You need to make sure any parsers you build, that could have spaces after, are wrapped in token(p). If you use the makeTokenParser function, then you need to make sure you build a token parser like so:

    Parser<A> token<A>(Parser<A> p) =>
        from r in p
        from _ in lexer.WhiteSpace
        select r;

Stop manually dealing with spaces, it's a fools errand, treat your source-code as a sequence of tokens and you'll find it much, much easier.