[spirit2] compile time

Joel de Guzman — 2008-12-01T15:50:38

Hi, I'm sure you all know that I am tweaking spirit2 inside out to squeeze more compile time performance. Here are some numbers so far. The most complex example in the spirit2 examples folder is the mini-C compiler (compiles a very simple C like language to a virtual machine). It is as close to a real world application while still maintaining simplicity for the purpose of instruction. Some numbers: spirit2 (in boost trunk): 2 minutes, 37.9 seconds spirit2x (experimental in spirit svn): 1 minute, 46.2 seconds What's interesting about the mini-C compiler is that I tried as much as possible to apply all tricks (***) to make it build faster, like e.g. composing separate grammars, separating the grammar definitions (the constructors) in separate cpp files, explicit instantiation of the templates, etc. So, while the original 2 minutes, 38 seconds is not bad at all to begin with (considering the /relative/ complexity of the grammar(s)), now we have a definite speedup. I'm quite happy. Let's hope w

[karma] streaming empty containers

Matthias Vallentin — 2008-12-01T02:02:07

Hi there, how do I reliably stream containers regardless of their size? The following example only works with non-empty containers: std::vector v; std::cout << boost::spirit::karma::format( '{' << (stream % ", ") << '}', v ) << std::endl; If the container is empty the output is {% and does not include a closing bracket. In fact, the entire stream is cut off at the point of the first '{'. Of course, I could manually check if the container is empty but it would be sweet if karma can handle empty containers as well (output: "{}"). Could it be that I am misusing karma here? Matthias

Simple suggestion for BOOST_FUSION_ADAPT_STRUCT

Raindog — 2008-12-01T00:10:11

Perhaps for structures whose main purpose is to be used w/ fusion, a macro can be used that will both define the structure and adapt it so that: struct foo { int var1; int var2; std::string name; std::string name2; }; BOOST_FUSION_ADAPT_STRUCT( foo, (int, var1) (int, var2) (std::string, name) (std::string, name2) ) Can be shorted to something like: BOOST_FUSION_DEFINE_AND_ADAPT_STRUCT( foo, (int, var1) (int, var2) (std::string, name) (std::string, name2) ) ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

[karma] streaming a std::map

Matthias Vallentin — 2008-11-30T18:54:14

Hi, could someone give me a short example how to stream std::map? I understand how to stream a std::vector with karma: out << karma::format( '{' << (stream % ", ") << '}', container ); If container is a vector, T needs operator<<. Would it make sense to provide an operator<< for std::map::value_type in order make the above code snippet work? Matthias

problem with simple grammar action.

Raindog — 2008-11-30T09:57:22

Hello, I've just created my first grammar and it worked on line at a time usage, but I wanted it to be able to operate on whole files at once and give me back a vector of what I was parsing. The following is working for a file that contains only parsable lines of text: struct foo_parser: boost::spirit::qi::grammar(), space_type> { foo_parser() : foo_parser::base_type(signatures) { signatures %= *(signature_r) ; signature_r %= '{' >> hex_number >> ',' >> hex_number >> ',' >> hex_number >> ',' >> hex_number >> ',' >> quoted_string >> '}' >> ',' >> comment ; quoted_string %= lexeme [ '"' >> +(char_ - '"') >> '"' ] ; comment %= le

[spirit2]"small" mod to mini_xml1.cpp gives error: invalid use of incomplete type 'struct boost::spirit::qi::expr_transform

Larry Evans — 2008-11-29T21:18:11

The code in 1st attachement gives error in 2nd attachment. The error is not much help to me. Hopefully someone more experienced in spirit2 could give me some hint about what's wrong. Any help is appreciated. -Larry -*- mode: compilation; default-directory: "~/prog_dev/boost-svn/ro/boost_1_37_0/sandbox/lje/libs/utility/example/" -*- Compilation started at Sat Nov 29 15:05:43 bjam spirit2.open_close_delim.proto ...patience... ...patience... ...found 1462 targets... ...updating 3 targets... gcc.compile.c++ ../../../../../bin.v2/sandbox/lje/libs/utility/example/gcc-4.2/debug/spirit2.open_close_delim.proto.o ../../../../../boost/spirit/home/support/nonterminal/detail/nonterminal_fcall.hpp: In member function 'typename boost::lazy_enable_if_c<(boost::mpl::size > >::type>::value == 1), boost::spirit::make_nonterminal_holder

to lex or not to lex

Jürgen Hofmann — 2008-11-28T19:09:38

Hi, can someone give me a hint, if I want to use a lexer or not? ;-) I was playing a bit with qi and and lex; so for a very simple highly specialised scripting language, what would you recommend? The approach using a lexer seemed to be a bit cleaner and more reusable to me although it seems to require more work to get nice error messages. On the other hand qi seems to be so powerful, that I don't see, why I should need a lexer. For my co-workers I think it's easer to understand something like char_(';') than token(T_SEMICOLON) (although I'm not sure about it). And then again the lexer makes it much easier to "block reserved keywords"... I'm lost. Any recommendations? Thanks, Juergen ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://mobli

Phoenix lazy class 2008?

Toby Walsh — 2008-11-27T03:15:52

------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/_______________________________________________ Spirit-general mailing list Spirit-general< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/spirit-general

Lexer Questions will be answered

Ben Hanson — 2008-11-25T16:41:37

test

Ben Hanson — 2008-11-25T16:27:39

[Spirit2]UTF-8

OvermindDL1 — 2008-11-25T07:48:59

So, I have lots of little files I need to parse. However, they are all in UTF-8 and do include characters that span up to 4 bytes in length. Since Spirit2 (or Spirit2x) does not include a char page set for UTF-8 (or Unicode in general), what would be the recommended way to parse this? ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

[spirit 2] spirit-lex continuing testing

Carl Barron — 2008-11-25T03:56:03

I forget the exact flax syntax but this should b close enough to explain the problem: /* %% "[ \t\n]+ {} "[^ \t\n]"{return *yytext;} { "[ \t]{} "\n+"{BEGIN BOL_1;} "[^ \t\n]" {return '\0';} // some char that will fail the parse } { "[ \t]"{BEGIN BOL;} .{BEGIN INIYIAL; return *yytext;}// no newlines here!! } */ in english in the initial state it merely eats [skips] [ \t\n] and returns the matched character, in BOL it skips [ \y\n]+ where the last char is a \n. In other words in the defalt state passed " \ta" it returns 'a' in state BOL it returns something the grammar will not parse such as '\0'. " \na" returns "a" in both states " \n a" returns a in initial state and fails in state BOL [returns '\0' for instance then a lex_and_parse()'s grammar can get a bel parser just changing the state of the lexer where a bol is nmeded and it will go back to initial state after the lexer returns its token. I am confused . is this possible, fairly easily?

A small problem with the kleene star

Kisso Bajslovski — 2008-11-24T14:02:23

Permutation parser with non mandatory perm.items

Ladislav — 2008-11-24T08:03:44

Hi, I tryed make my self permutation parser with possibility to have non obligatory permutation item parsers. Evrithing goes well for me, but i cant compile if i use main parse function with Skipper parser. Evidently i miss somthing in my templates definition. Here is source code for my pmut_p and pi_p types. http://www.nabble.com/file/p20625402/permutation.hpp permutation.hpp If some one have time and understund more then me of templates pls help me :) (Im a template begginer) Here is some grammar for testing: struct test : public grammar { template struct definition { definition(calculator const& ) { a = !ch_p('a'); b = !ch_p('b'); c = ch_p('c'); file = !ch_p('>') >> pmut_p( (pi_p(a), pi_p(b), pi_p(c)) ); BOOST_SPIRIT_DEBUG_RULE(a); BOOST_SPIRIT_DEBUG_RULE(b); BOOST_SPIRIT_DEBUG_RULE(c); BOOST_SPIRIT_DEBUG_RULE(file); } rule a, b, c, file; rule const&

Help me getting started please

Brett Stottlemyer — 2008-11-22T20:44:50

a new directive similar to lexeme_d that keeps newlines

Behringer Behringer — 2008-11-22T12:53:48

stopping a parser via semantic actions

Carl Barron — 2008-11-17T18:59:20

Ok so the only way I can figure out how to solve the problem I had with lazy , is to use a variable in the parser's struct to hold a temporary symbol table such as an std:;set since I don't need type info until the list is parsed, A phoenix function can attempt to insert a new entry and if it exists indicate an error, using eps semantic predicates I can get the parser to fail and the iterators will be correct , correct? Now are inherited attributes modified in a called rule modified in the calling rule foo = bar(_val) % ','; bar = name [_a = _1] >> eps( insert_and_check_dups(_r1,_a) ) ; name %= lexeme[+alpha]; is foo's _val changed by bar, assuming it matches and insert_and_check_dups(....) changes _r1, further _1,_2,etc are place holders for synthetic attributes of a called parser, but there is only one so what are _2,_3 etc good for?? Is there a way to write a functor_parser with inherited attributes? -----------------------------------------------------

[spirit2] on_error

Jürgen Hofmann — 2008-11-10T20:00:56

In spirit2 we can use the (wonderful) on_error feature like on_error ( xml , std::cout << val("Error! Expecting ") << _4 // what failed? << val(" here: \"") << construct(_3, _2) // iterators to error-pos, end << val("\"") << std::endl ); This works perfect with std::string::iterator etc.. However, if I use lexertl_lexer like : typedef std::string::iterator base_iterator_type; typedef lexertl_token< base_iterator_type, boost::mpl::vector > token_type; typedef lexertl_lexer lexer_type; typedef idl_tokens idl_tokens; typedef lexer::iterator_type iterator_type; I can pass this iterator_type to a grammar keeping the compiler happy, but the construct(_3, _2) will not work. Is there a way to use on_error with this iterator_type, so that I can access the underlying base_iterator? Juergen -------------------

[spirit2] bug in qi::in_state?

Jürgen Hofmann — 2008-11-10T18:54:38

Hi, I don't know about the official way to report bugs... Some of the lex-samples in boost 1.37.0 don't compile (e.g. example3, example5, example6). They all complain about template inline typename proto::terminal >::type in_state(std::basic_string const& s) { typedef std::basic_string string_type; typename proto::terminal >::type that; >> that.s = s; return that; } 's' : is not a member of 'boost::proto::exprns_::expr' Juergen ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

wave and support_option_insert_whitespace

Jürgen Hofmann — 2008-11-10T18:40:26

Hi, probably I miss the point... The wave documentation says: "If the support_option_insert_whitespace flag is set, the Wave library inserts additional whitespace between tokens where the concatenated string representation of the two tokens would form another valid C++ token. This is very importatnt to avoid mis-interpretations of the generated output if it is going to be processed by a C++ compiler afterwards." If I pre-process a file with wave (using the cpp-lexer and support_option_insert_whitespace) containing an erroneous fragment like "int 1abc;" it generates "int 1 abc;" Is this intended (the compiler produces different error messages, which may be confusing)? Juergen ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-co

position_iterator bug (Boost 1.37.0)

Rainer Deyke — 2008-11-08T05:34:50

In boost::spirit::position_iterator::equal: return (_isend && x_is_end) || (this->base() == rhs.base()); This is incorrect. If either iterator is an end iterator created by default construction, it is not safe to compare the underlying iterators because they may be uninitialized. This error has two possible symptoms: - It can cause a crash, if the iterator is of a type that cannot be safely compared when default-constructed. (This is what happened in my case.) - It can cause an end iterator to compare equal to a non-end iterator, if the end iterator happens to have the correct bit pattern in its underlying iterator. Corrected code: return (_isend != x_is_end) && (_isend ? true : this->base() == rhs.base());

gmane.comp.parsers.spirit.general