[Haskell-cafe] Parsec on TeX
Ross Boylan
to follow the example of how to use Parsec to parse TeX begin/end groups,
but can't get it to run. I'm using HUGS -98 on Debian.
When I copied the code I got errors about unknown terms (reserved and
braces). I've tried to get them from the lexer, but now get this error
:load grammar.hsl
ERROR "grammar.hsl":21 - Type error in explicitly typed binding
*** Term : envEnd
*** Type : String -> GenParser Char a [Char]
*** Does not match : String -> Parser ()
Can anyone help me understand what the problem is?
Here's the code the caused the above error; I believe the part after --TeX
example is verbatim from the Parsec documentation. I picked haskell as the
language for to lexer "arbitrarily."
import Text.ParserCombinators.Parsec
import qualified Text.ParserCombinators.Parsec.Token as P
import Text.ParserCombinators.Parsec.Language(haskell)
reserved = P.reserved haskell
braces = P.braces haskell
-- TeX example
environment = do{ name <- envBegin
; environment
; envEnd name
}
<|> return ()
envBegin :: Parser String
envBegin = do{ reserved "\\begin"
; braces (many1 letter)
}
envEnd :: String -> Parser ()
envEnd name = do{ reserved "\\end"
; braces (string name)
}
_______________________________________________
Haskell-Cafe mailing list
Haskel...@haskell.org
http://www.haskell.org/mailman/listinfo/haskell-cafe
Brandon S. Allbery KF8NH
On 2008 May 4, at 23:40, Ross Boylan wrote:
> ERROR "grammar.hsl":21 - Type error in explicitly typed binding
> *** Term : envEnd
> *** Type : String -> GenParser Char a [Char]
> *** Does not match : String -> Parser ()
Hugs is prone to error messages that obscure the problem. The trick
here is to realize that the type "Parser ()" is the same as "GenParser
Char a ()"; this then tells you that you have used a function that
returns a [Char] (aka String) where a type () (Haskell's version of
(void)) is expected.
> envEnd :: String -> Parser ()
> envEnd name = do{ reserved "\\end"
> ; braces (string name)
> }
Line 21 is "; braces (string name)"; it is producing a String, when
you need a (). One fix is to add one more line:
> envEnd :: String -> Parser ()
> envEnd name = do reserved "\\end"
> braces (string name)
> return ()
Another possible fix is to change the type of "envEnd" to "String ->
Parser String"; this may depend on how it's used.
--
brandon s. allbery [solaris,freebsd,perl,pugs,haskell] all...@kf8nh.com
system administrator [openafs,heimdal,too many hats] all...@ece.cmu.edu
electrical and computer engineering, carnegie mellon university KF8NH
Don Stewart
>
> On 2008 May 4, at 23:40, Ross Boylan wrote:
>
> >ERROR "grammar.hsl":21 - Type error in explicitly typed binding
> >*** Term : envEnd
> >*** Type : String -> GenParser Char a [Char]
> >*** Does not match : String -> Parser ()
>
> Hugs is prone to error messages that obscure the problem. The trick
> here is to realize that the type "Parser ()" is the same as "GenParser
> Char a ()"; this then tells you that you have used a function that
> returns a [Char] (aka String) where a type () (Haskell's version of
> (void)) is expected.
Yeah, if you're on Debian it would make sense to install GHC -- its much
more active, much faster, and supports more things.
It's well maintained on Debian too.
-- Don
Derek Elkins
braces returns, in this case, a string, so the type of envEnd is
String -> Parser String. You can either change the type and add a
return () to environment after envEnd or add a return () to envEnd.
Ross Boylan
> On 2008 May 4, at 23:40, Ross Boylan wrote:
>
> > ERROR "grammar.hsl":21 - Type error in explicitly typed binding
> > *** Term : envEnd
> > *** Type : String -> GenParser Char a [Char]
> > *** Does not match : String -> Parser ()
>
> Hugs is prone to error messages that obscure the problem. The trick
> here is to realize that the type "Parser ()" is the same as "GenParser
> Char a ()"; this then tells you that you have used a function that
> returns a [Char] (aka String) where a type () (Haskell's version of
> (void)) is expected.
>
> > envEnd :: String -> Parser ()
> > envEnd name = do{ reserved "\\end"
> > ; braces (string name)
> > }
>
>
> Line 21 is "; braces (string name)"; it is producing a String, when
> you need a (). One fix is to add one more line:
>
> > envEnd :: String -> Parser ()
> > envEnd name = do reserved "\\end"
> > braces (string name)
> > return ()
>
> Another possible fix is to change the type of "envEnd" to "String ->
> Parser String"; this may depend on how it's used.
First, I'm really impressed with the fast and helpful responses from
several people!
So the example is wrong? What inference should I draw about the state
of Parsec and its documentation? I was thinking of trying Frost et al's
X-SAIGA, but that the better documentation for parsec would be a plus.
I had thought HUGS made more sense for fiddling around, but it seems all
I'm doing is loading files anyway. What is the style people use for
this exploratory work? I've already installed haskell-mode for emacs
and ghc6, but it's not clear to me to what extent the former servers as
a development environment rather than just a language formatter.
Thanks.
Don Stewart
> I had thought HUGS made more sense for fiddling around, but it seems all
> I'm doing is loading files anyway. What is the style people use for
> this exploratory work? I've already installed haskell-mode for emacs
> and ghc6, but it's not clear to me to what extent the former servers as
> a development environment rather than just a language formatter.
GHCi makes a great exploratory tool (and then you have the option of
compiling the code as well).
There are some nice integration tools with emacs and vim, but its also
good to gain familiarity with ghci, the online documentation on
haskell.org, and perhaps even drop by the #haskell IRC channel.
Cheers,
Don
Brandon S. Allbery KF8NH
On 2008 May 5, at 0:14, Ross Boylan wrote:
> So the example is wrong? What inference should I draw about the state
> of Parsec and its documentation? I was thinking of trying Frost et
> al's
> X-SAIGA, but that the better documentation for parsec would be a plus.
The original documentation which you will probably find via Google is
several years out of date; unfortunately, nobody has updated it and
the newer API documentation lacks most of the actual details of how to
*use* Parsec --- so really you need to have daan's Parsec paper and
the Haskell library API documentation ( http://www.haskell.org/ghc/docs/latest/html/libraries/parsec/Text-ParserCombinators-Parsec.html
) open at the same time to see how things have changed.
I will note that a newer version of Parsec is in development; I am
hoping it will have somewhat better documentation (preferably an
updated version of the original Parsec paper).
> I had thought HUGS made more sense for fiddling around, but it seems
> all
> I'm doing is loading files anyway. What is the style people use for
> this exploratory work? I've already installed haskell-mode for emacs
> and ghc6, but it's not clear to me to what extent the former servers
> as
> a development environment rather than just a language formatter.
ghci (the interpreter component of GHC) works much like hugs but is
much more up to date; there are a number of useful libraries which
either haven't been ported to or simply don't work with Hugs. (The
new Parsec is one of these, as it works with the ByteString library
(much faster than standard Strings, which are naive linked lists of
characters); ByteString really needs to be compiled and Hugs is only
an interpreter.)
--
brandon s. allbery [solaris,freebsd,perl,pugs,haskell] all...@kf8nh.com
system administrator [openafs,heimdal,too many hats] all...@ece.cmu.edu
electrical and computer engineering, carnegie mellon university KF8NH
Brandon S.Allbery KF8NH
going for it has nothing to do with the language itself; as you
noticed, it's the community. One might expect to find crusty
mathematicians and dubiously social geeks(*), but in fact the Haskell
community is one of the friendliest and most helpful I've encountered
anywhere.
(*) ironically, the person saying this is both dubiously social and an
even crustier system administrator :)
Derek Elkins
Parsec is currently maintained, but wasn't maintained for quite a while
(admittedly, it didn't really require much maintenance; it was certainly
rampantly used during that time.) Daan's paper on Parsec that you are
reading/read is probably the best tutorial-like introduction. The
reference documentation has been Haddock-ed for the newer version of
Parsec (which you are almost certainly not using, but it is still very
similar.)
http://hackage.haskell.org/packages/archive/parsec/3.0.0/doc/html/Text-Parsec.html
Unfortunately, Daan seems to have disappeared off of the face of the
internet so there isn't anything that can be done about fixing that
documentation (and likely that is in large part why this error is still
there.)
Luke Palmer
My guess is the following:
string :: String -> Parser String
braces :: Parser a -> Parser a
Meaning
braces (string name) :: Parser String
Which is not the same as your declared return type Parser (). Add a
return () at the end of envEnd.
Ross Boylan
> Unfortunately, Daan seems to have disappeared off of the face of the
> internet so there isn't anything that can be done about fixing that
> documentation (and likely that is in large part why this error is
> still
> there.)
>
Andrew Coppin
> First, I'm really impressed with the fast and helpful responses from
> several people!
>
Ask an easy question and you'll get at least a dozen people tripping
over each other in their rush to offer you useful advice. Ask a hard
question, or an uninteresting question, and enjoy the utter silence...
:-( Still, I suppose that's not *too* surprising.
> I had thought HUGS made more sense for fiddling around, but it seems all
> I'm doing is loading files anyway. What is the style people use for
> this exploratory work? I've already installed haskell-mode for emacs
> and ghc6, but it's not clear to me to what extent the former servers as
> a development environment rather than just a language formatter.
>
I used to use Hugs as well. Then I updated to the newest version, and
found it horrifically buggy. That's when I moved over to GHC. As Don
said, GHCi doesn't let you define new functions like Hugs does, but
other than that it's far more useful. (And actually, you *can* define
small functions using 'let'. It's just awkward.) Best thing is, you can
compile your code with GHC, and then load that compiled code into GHCi
so it runs at near-native speed. And of course, the latest version of
GHCi now comes with an actual debugger. And GHC-compiled binaries have a
profiler. And...
[This was a while ago; maybe Hugs has been updated again since I last
looked. Certainly Hugs is smaller to download, and more friendly to use
than GHC. But it just doesn't have the same power...]
apfelmus
> Yeah, if you're on Debian it would make sense to install GHC -- its much
Except for compile and load time, Hugs is really fast there.
Regards,
apfelmus
Andrew Coppin
> Don Stewart wrote:
>> Yeah, if you're on Debian it would make sense to install GHC -- its much
>> more active, much faster, and supports more things. [than Hugs]
>
> Except for compile and load time, Hugs is really fast there.
Mmm, possibly. I seem to recall it being instantaneous on my old AMD K6
500MHz with both Hugs and GHCi. But it was a while ago...
Certainly only Hugs as the nice feature of automatically reloading your
source every time you hit save. I really miss that in GHCi. [OTOH, if
you're going to do this, expect lots of "hey, this doesn't parse!" type
messages. undefined is your friend!]
Derek Elkins
> On Sun, 2008-05-04 at 23:59 -0500, Derek Elkins wrote:
> > Unfortunately, Daan seems to have disappeared off of the face of the
> > internet so there isn't anything that can be done about fixing that
> > documentation (and likely that is in large part why this error is
> > still
> > there.)
> >
> http://research.microsoft.com/users/daan/index.html ?
I'm not sure that he is still at Microsoft; I've heard tell that he is
off on another pursuit. At any rate, I've tried three different email
addresses including the one listed there and have not gotten a response
and others have had trouble getting in touch with him as well. I guess
I could try calling that number and see what happens...
Ross Boylan
> On Sun, 2008-05-04 at 23:57 -0700, Ross Boylan wrote:
> > On Sun, 2008-05-04 at 23:59 -0500, Derek Elkins wrote:
> > > Unfortunately, Daan seems to have disappeared off of the face of the
> > > internet so there isn't anything that can be done about fixing that
> > > documentation (and likely that is in large part why this error is
> > > still
> > > there.)
> > >
> > http://research.microsoft.com/users/daan/index.html ?
>
> I'm not sure that he is still at Microsoft; I've heard tell that he is
> off on another pursuit. At any rate, I've tried three different email
> addresses including the one listed there and have not gotten a response
> and others have had trouble getting in touch with him as well. I guess
> I could try calling that number and see what happens...
>
Ross
Don Stewart
> Ross Boylan wrote:
> >First, I'm really impressed with the fast and helpful responses from
> >several people!
> >
>
> Ask an easy question and you'll get at least a dozen people tripping
> over each other in their rush to offer you useful advice. Ask a hard
> question, or an uninteresting question, and enjoy the utter silence...
> :-( Still, I suppose that's not *too* surprising.
>
> >I had thought HUGS made more sense for fiddling around, but it seems all
> >I'm doing is loading files anyway. What is the style people use for
> >this exploratory work? I've already installed haskell-mode for emacs
> >and ghc6, but it's not clear to me to what extent the former servers as
> >a development environment rather than just a language formatter.
> >
>
> I used to use Hugs as well. Then I updated to the newest version, and
> found it horrifically buggy. That's when I moved over to GHC. As Don
> said, GHCi doesn't let you define new functions like Hugs does, but
Other way around, GHCi lets you define new functions, unlike Hugs :)
$ ghci
Prelude> let f x = x ^ 2
Prelude> f 2
4
And:
$ hugs
Type :? for help
Hugs.Base> let f x = x ^ 2
ERROR - Syntax error in expression (unexpected end of input)
Hugs.Base> [Leaving Hugs]