[Haskell-cafe] parsing machine-generated natural text

[Jason Dagit]

On 5/20/06, Udo Stenzel <u.stenzel at web.de> wrote:
>
> do power
>    colon
>    integer
>    reserved "Supply centers,"
>    integer
>    reserved "Units:"
>    ((reserved "Builds" >> return id) <|>
>         (reserved "Disbands" >> return negate))
>         `ap` integer
>    reserved "units." <|> reserved "unit."
>

I always struggle with when I need to use 'try' with parsec.  I tend
to over use it because I've had unexpected results when I leave it
out.  In your other email you say that 'reserved "Movement"' expands
to 'lexeme
(string "Movement")'.  So I would think that,

reserved "units." <|> reserved "unit."

would need to be wrapped in a 'try'.  Something like,

try (reserved "units.") <|> reserved "unit."

My understanding is that if 'unit.' appears in the input the first
parser will parse up to the '.' and then fail and consume the input up
to that point, leaving the alternative with only the period as input
so it will also fail.

So I'm wondering if someone could explain to me what is wrong with my
understanding of parsec or point me to a resource that explains this
(probably common) misunderstanding.  Specifically, I would like to
develop a better understanding of when 'try' is needed, and when it is
not needed.

Thanks,
Jason