Bug? - surroundedBy with optional whitespace fails to parse about cats-parse HOT 5 CLOSED

actually... .soft.surroundedBy(a) seems like a good idea to have, which would be a.soft *> p.soft <* a

What do you think of that API?

from cats-parse.

See this comment:

#188 (comment)

surroundedBy with whitespace does not work great. The problem is lack of backtracking. Your design is assuming backtracking. As soon as you see whitespace, the first item will start to parse, if you don't hit the surrounded thing, you will get an error (again, an intentional goal is opt-in backtracking).

So, I think instead you want something like: whitespaces0.soft *> comma *> whitespaces0

The reason your final example fails is because with you hit y ) the space starts looking like a listSeparator, but then you don't see a , but the parser expects that.

Altenatively, you could make listSeparator backtrack by doing:

val listSeparator = comma.surroundedBy(whitespaces0).backtrack

from cats-parse.

Thanks for the explanation. It pointed me to a reasonable solution.

Rather than introducing backtracking just to eliminate whitespace, I'm removing whitespace after each token, rather than both before and after. I created a ".strip" method to skip over trailing whitespace.

extension[T](self:Parser[T]) def strip = self <* (whitespace.rep0.void)

I fell into my way of errors by following the README example. Given the pitfalls, it probably isn't good practice to use surroundedBy to remove whitespace.

from cats-parse.

you are right. We need to improve the readme... it compiles, but it actually isn't great for the same reason.

from cats-parse.

done in #289 #290

from cats-parse.