On 8/3/10 10:22 AM, Lukas Renggli wrote:
Hi Bijan,
Please ask further questions in the Moose mailing-list, other people might be interested too.
Thanks for sharing it: I would be definitively happy to read any discussion related to PetitParser! :)
Btw, it seems that the Java grammar (i.e. PetitJava) is coming.
Just to share my experience, I first tried to implement it following the Java Language Specification book, Chapter 18. Even though such chapter should sum up the whole syntax, it is not completely consistent with the rest of the book (which details all the rules with explanations and examples), and presents some errors. Thus, after a brief trial, I moved to implement the grammar according to the more detailed version. Alas, such chapters are not perfect, too! I found out a "corrected" version in a web site [1]. When finally I managed to implement it, it turned out that such specifications contain various cycles. I tried to fix them, but that would have required much time. So, I decided to take the ANTLR version of the java grammar (which has been tested successfully over a comprehensive regression test suite) that would contain no dangerous cycles (since it is LL(1)) for PetitParser [2]. Now that grammar is implemented, and it correctly parses the compilation units I tried. It still needs work (first of all, more tests. Then, a parser), but it's coming.
From this experience, I would say that the first translator from grammar-in-another-specification to PetitParser, should be from ANTLR grammars. The fact that they are LL(1) should avoid cycles (please correct me, if I am wrong), and there is a very big number of grammars written for ANTLR.
Ciao, Alberto
[1] http://www.cmis.brighton.ac.uk/staff/rnb/bosware/javaSyntax/Java1_6GrammarCo... [2] http://openjdk.java.net/projects/compiler-grammar/antlrworks/Java.g
Hi Alberto,
This is great news. Maybe you also want to post the repository URL? :-)
For what you noticed on bogus grammar specifications check out the work of Vadim Zaytsev (http://mobile.twitter.com/grammarware). Among others he analized varous Java Language Specifications.
Note that cycles are present in most grammars and perfecly fine. The only problematic cycles are the left-recursive ones for recursive descent parsers like ANTLR or PetitParser. Even then you can simply fix the problem by adding a memoizer into the offending cycle, but that yields a very slow grammar. You can find offending cycles with the message #cycleSets from the PetitAnalyzer package.
Lukas
On Tuesday, August 3, 2010, Alberto Bacchelli alberto.bacchelli@usi.ch wrote:
On 8/3/10 10:22 AM, Lukas Renggli wrote:
Hi Bijan,
Please ask further questions in the Moose mailing-list, other people might be interested too.
Thanks for sharing it: I would be definitively happy to read any discussion related to PetitParser! :)
Btw, it seems that the Java grammar (i.e. PetitJava) is coming.
Just to share my experience, I first tried to implement it following the Java Language Specification book, Chapter 18. Even though such chapter should sum up the whole syntax, it is not completely consistent with the rest of the book (which details all the rules with explanations and examples), and presents some errors. Thus, after a brief trial, I moved to implement the grammar according to the more detailed version. Alas, such chapters are not perfect, too! I found out a "corrected" version in a web site [1]. When finally I managed to implement it, it turned out that such specifications contain various cycles. I tried to fix them, but that would have required much time. So, I decided to take the ANTLR version of the java grammar (which has been tested successfully over a comprehensive regression test suite) that would contain no dangerous cycles (since it is LL(1)) for PetitParser [2]. Now that grammar is implemented, and it correctly parses the compilation units I tried. It still needs work (first of all, more tests. Then, a parser), but it's coming.
From this experience, I would say that the first translator from grammar-in-another-specification to PetitParser, should be from ANTLR grammars. The fact that they are LL(1) should avoid cycles (please correct me, if I am wrong), and there is a very big number of grammars written for ANTLR.
Ciao, Alberto
[1] http://www.cmis.brighton.ac.uk/staff/rnb/bosware/javaSyntax/Java1_6GrammarCo... [2] http://openjdk.java.net/projects/compiler-grammar/antlrworks/Java.g _______________________________________________ Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
-
Alberto Bacchelli -
Lukas Renggli