Hi, I'm having some difficulty using the Parsec library, perhaps you could help. I've reduced my problem as shown below. I would like the 'only_prod' parser to require the reserved string "only", _optionally_ followed by an identifier. As part of 'mytest', this should then be followed by the reserved string "end". mytest = do { only_prod; end } only_prod = do { reserved "only"; try identifier } end = reserved "end" i.e. I'd like both
parseTest mytest "only end"
and
parseTest mytest "only green end"
to parse successfully. As it stands, only the second is successful. The first fails with: parse error at (line 1, column 9): unexpected end of input expecting letter or digit or "end" Does anyone have an idea how I could repair the situation? Thanks in advance, Paul
Paul Keir wrote:
I’m having some difficulty using the Parsec library, perhaps you could help. I’ve reduced my problem as shown below. I would like the ‘only_prod’ parser to require the reserved string “only”, _optionally_ followed by an identifier. As part of ‘mytest’, this should then be followed by the reserved string “end”.
Look into option, optional, and optionMaybe.
On Tue, 2008-03-25 at 11:47 -0400, Albert Y. C. Lai wrote:
Paul Keir wrote:
I’m having some difficulty using the Parsec library, perhaps you could help. I’ve reduced my problem as shown below. I would like the ‘only_prod’ parser to require the reserved string “only”, _optionally_ followed by an identifier. As part of ‘mytest’, this should then be followed by the reserved string “end”.
Look into option, optional, and optionMaybe.
And in particular, try doesn't do what you (Paul) think it does. try however is a rather important combinator of Parsec so I would behoove you to look up it's documentation and the explanations of it in the Parsec letter.
Thanks. I can't find optionMaybe in my version 2.1 of Parsec, but in any case, defining my only_prod as only_prod = do { reserved "only"; option [] identifier } or only_prod = do { reserved "only"; identifier <|> return [] } gives the same error responses as before. I will anyway look closer at option. You're right that I don't understand try, but it's not for lack of trying. My examples' use of try though was just a stab at a readable failure. Maybe I should refactor my example. Paul
On Mar 25, 2008, at 12:43 , Paul Keir wrote:
Thanks. I can't find optionMaybe in my version 2.1 of Parsec, but in any case, defining my only_prod as
only_prod = do { reserved "only"; option [] identifier }
or
only_prod = do { reserved "only"; identifier <|> return [] }
gives the same error responses as before. I will anyway look closer at option.
The other problem here is that just using a given string in "reserved" doesn't prevent it from being parsed elsewhere by "identifier". (Note the character offset of the error was 9, i.e. just past "only end", and it was looking for "end" or more identifier characters.) Are you using the higher level parser facilities from Text.ParserCombinators.Parsec.Token, or rolling your own? If the latter, you will need to modify "identifier" to not accept keywords. -- brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allbery@kf8nh.com system administrator [openafs,heimdal,too many hats] allbery@ece.cmu.edu electrical and computer engineering, carnegie mellon university KF8NH
Thankyou. Yes, I'd also noticed that "only end" could result in the "end" part being taken as an identifier. The language I'm parsing actually doesn't have reserved words though; so "end" and "only" are both possible valid identifiers. I should then probably replace my use of say, reserved "only", with string "only"; whiteSpace; for clarity. Still stuck though... P -----Original Message----- From: Brandon S. Allbery KF8NH [mailto:allbery@ece.cmu.edu] Sent: Tue 25/03/2008 19:58 To: Paul Keir; haskell-cafe@haskell.org Cafe Subject: Re: [Haskell-cafe] Parsec (Zero or One of) On Mar 25, 2008, at 12:43 , Paul Keir wrote:
Thanks. I can't find optionMaybe in my version 2.1 of Parsec, but in any case, defining my only_prod as
only_prod = do { reserved "only"; option [] identifier }
or
only_prod = do { reserved "only"; identifier <|> return [] }
gives the same error responses as before. I will anyway look closer at option.
The other problem here is that just using a given string in "reserved" doesn't prevent it from being parsed elsewhere by "identifier". (Note the character offset of the error was 9, i.e. just past "only end", and it was looking for "end" or more identifier characters.) Are you using the higher level parser facilities from Text.ParserCombinators.Parsec.Token, or rolling your own? If the latter, you will need to modify "identifier" to not accept keywords. -- brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allbery@kf8nh.com system administrator [openafs,heimdal,too many hats] allbery@ece.cmu.edu electrical and computer engineering, carnegie mellon university KF8NH
On Mar 25, 2008, at 16:26 , Paul Keir wrote:
Thankyou. Yes, I'd also noticed that "only end" could result in the "end" part being taken as an identifier. The language I'm parsing actually doesn't have reserved words though; so "end" and "only" are both possible valid identifiers. I should then probably replace my use of say, reserved "only", with string "only"; whiteSpace; for clarity. Still stuck though...
But now you have an ambiguity in your language, which is exactly why the parse is failing: "only end" could be waiting for "end", or for end of file / whatever tokens might follow this clause. In the worst case, the latter might lead to a situation where an unambiguous parse is impossible. You might want to provide a better description of the full language --- and think about how it would need to be implemented to avoid ambiguity. -- brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allbery@kf8nh.com system administrator [openafs,heimdal,too many hats] allbery@ece.cmu.edu electrical and computer engineering, carnegie mellon university KF8NH
Many thanks. -----Original Message----- From: Brandon S. Allbery KF8NH [mailto:allbery@ece.cmu.edu] Sent: Tue 25/03/2008 20:29 To: Paul Keir; haskell-cafe@haskell.org Cafe Subject: Re: [Haskell-cafe] Parsec (Zero or One of) On Mar 25, 2008, at 16:26 , Paul Keir wrote:
Thankyou. Yes, I'd also noticed that "only end" could result in the "end" part being taken as an identifier. The language I'm parsing actually doesn't have reserved words though; so "end" and "only" are both possible valid identifiers. I should then probably replace my use of say, reserved "only", with string "only"; whiteSpace; for clarity. Still stuck though...
But now you have an ambiguity in your language, which is exactly why the parse is failing: "only end" could be waiting for "end", or for end of file / whatever tokens might follow this clause. In the worst case, the latter might lead to a situation where an unambiguous parse is impossible. You might want to provide a better description of the full language --- and think about how it would need to be implemented to avoid ambiguity. -- brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allbery@kf8nh.com system administrator [openafs,heimdal,too many hats] allbery@ece.cmu.edu electrical and computer engineering, carnegie mellon university KF8NH
participants (4)
-
Albert Y. C. Lai -
Brandon S. Allbery KF8NH -
Derek Elkins -
Paul Keir