Thank you the question is forwarded to the spirit ML. Regards Austina 2013/5/27 Igor R

...

I would like to choose a parser.

I would like to use a parser for text processing (natural language text). Which parser is suited in this case. In general which are the benefit to use boost spirit instead of ANTLR or ANTLR instead of boost spirit. Thank you.

I guess you'd get more answers on Spirit General ML: https://lists.sourceforge.net/lists/listinfo/spirit-general _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

-- Regards Austina

ANTLR is LR, Spirit is LL. Spirit is embedded in C++, ANTLR is a separate preprocessor.

Spirit is slow to compile and isn't very efficient at runtime, but it's fairly nice to use once you get used to it.

As for runtime performance, it depends on the use-case. http://alexott.blogspot.co.il/2010/01/boostspirit2-vs-atoi.html

On 05/29/13 07:59, salvatore dario minonne wrote:

ATLR is LL as well More specifically, ANTLR is LL(k) for some k>0:

http://en.wikipedia.org/wiki/LL_parser Spirit is a PEG: http://www.boost.org/doc/libs/1_43_0/libs/spirit/doc/html/spirit/abstracts/p... Both PEG and LL)k) are recursive descent; however, LL(k) means the parser can lookahead K tokens to decide which alternative to parse. I think spirit's expect operator means it can lookahead 1 token. Also, spirit will try alternatives in order until it finds a match; whereas LL{k) will not try to parse each alternative because it looks ahead k tokens to decide which alternative is the only one possible(as mentioned in the LL_parser page mentioned above). Also, w.r.t. natural language parsing, wikipedia says that's probably not a good idea ( as mentioned in the parsing_expression_grammar.html page). HTH.

On Wed, May 29, 2013 at 2:32 PM, Mathias Gaunard mailto:mathias.gaunard@ens-lyon.org> wrote:

On 27/05/13 20:06, Olivier Austina wrote:

Hi,

I would like to choose a parser.

I would like to use a parser for text processing (natural language text). Which parser is suited in this case. In general which are the benefit to use boost spirit instead of ANTLR or ANTLR instead of boost spirit. Thank you.

ANTLR is LR, Spirit is LL. Spirit is embedded in C++, ANTLR is a separate preprocessor.

Spirit is slow to compile and isn't very efficient at runtime, but it's fairly nice to use once you get used to it.

_________________________________________________ Boost-users mailing list Boost-users@lists.boost.org mailto:Boost-users@lists.boost.org http://lists.boost.org/__mailman/listinfo.cgi/boost-__users http://lists.boost.org/mailman/listinfo.cgi/boost-users

-- SDM

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

I'd like to apologize for my previous email which was somehow laconic. My goal was to stress the fact that ANTLR does not generate LR parser. To add something to original question I'd like to add that TTBOMK the latest version of ANTLR (v4) doesn't generate C/C++ code and the v3 generates only C. Hope it useful for Olivier On Wed, May 29, 2013 at 6:05 PM, Evan Driscoll wrote:

On 05/29/2013 09:51 AM, Larry Evans wrote:

...

On 05/29/13 07:59, salvatore dario minonne wrote:

...

ATLR is LL as well More specifically, ANTLR is LL(k) for some k>0:

http://en.wikipedia.org/wiki/LL_parser

Actually *that's* not true either. Antlr generates LL(*) parsers. With some restrictions, the generated parsers have infinite lookahead.

For instance,

nonterm1 : (term1)* term2 (term1)* term3

is not LL(k) for any k as if you give me a k I give you the string "term1^k term2" and you can't decide between those alternatives, but that grammar is is LL(*).

I'm not sure how much this actually affects real usage, just trying to be accurate. There is at least one example of something you may want to do which is LL(*) but not LL(k)

classMember: modifier* type ident SEMI | modifier* type ident LPAREN arglist RPAREN body

modifier: PUBLIC | PRIVATE | PROTECTED | STATIC | ...

(The description of what makes a grammar LL(*) requires more than I feel like explaining now.)

Evan

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

-- SDM

On 29/05/13 14:59, salvatore dario minonne wrote:

ATLR is LL as well

Sorry for the huge mistake, I must have confused it with other preprocessors that are usually LR.

On 29/05/13 17:48, Michael Caisse wrote:

On 05/29/2013 05:32 AM, Mathias Gaunard wrote:

...

Spirit is slow to compile and isn't very efficient at runtime, but it's fairly nice to use once you get used to it.

Mathias,

"isn't very efficient at runtime" ... are you concerned about the processing speed, or something else?

I'm concerned by the massive copying of attributes going on. Building a deep AST will result in a huge number of copies.

On 5/30/13 8:27 PM, Mathias Gaunard wrote:

On 29/05/13 17:48, Michael Caisse wrote:

...

On 05/29/2013 05:32 AM, Mathias Gaunard wrote:

...

Spirit is slow to compile and isn't very efficient at runtime, but it's fairly nice to use once you get used to it.

Mathias,

"isn't very efficient at runtime" ... are you concerned about the processing speed, or something else?

I'm concerned by the massive copying of attributes going on. Building a deep AST will result in a huge number of copies.

FWIW, Spirit is faster than ANTLR. People report Spirit to be 2.5 to 7 times faster than ANTLR. http://article.gmane.org/gmane.comp.parsers.spirit.general/16459/match=boost... http://article.gmane.org/gmane.comp.parsers.spirit.general/16463/match=spiri... Regards, -- Joel de Guzman http://www.ciere.com http://boost-spirit.com http://www.cycfi.com/