--- parser3/src/classes/op.C 2007/04/23 10:30:09 1.166 +++ parser3/src/classes/op.C 2009/05/04 09:26:19 1.185 @@ -1,11 +1,11 @@ /** @file Parser: parser @b operators. - Copyright (c) 2001-2005 ArtLebedev Group (http://www.artlebedev.com) + Copyright (c) 2001-2009 ArtLebedev Group (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) */ -static const char * const IDENT_OP_C="$Date: 2007/04/23 10:30:09 $"; +static const char * const IDENT_OP_C="$Date: 2009/05/04 09:26:19 $"; #include "classes.h" #include "pa_vmethod_frame.h" @@ -97,24 +97,27 @@ public: static void _if(Request& r, MethodParams& params) { bool condition=params.as_bool(0, "condition must be expression", r); if(condition) - r.write_pass_lang(r.process(*params.get(1))); + r.process_write(*params.get(1)); else if(params.count()>2) - r.write_pass_lang(r.process(*params.get(2))); + r.process_write(*params.get(2)); } -static void _untaint(Request& r, MethodParams& params) { +static String::Language get_untaint_lang(MethodParams& params, int index){ + const String& lang_name=params.as_string(index, "lang must be string"); + String::Language lang=untaint_lang_name2enum.get(lang_name); + if(!lang) + throw Exception(PARSER_RUNTIME, + &lang_name, + "invalid taint language"); + return lang; +} +static void _untaint(Request& r, MethodParams& params) { String::Language lang; if(params.count()==1) - lang=String::L_AS_IS; // mark as simply 'tainted'. useful in html from sql - else { - const String& lang_name=params.as_string(0, "lang must be string"); - lang=untaint_lang_name2enum.get(lang_name); - if(!lang) - throw Exception(0, - &lang_name, - "invalid taint language"); - } + lang=String::L_AS_IS; // mark as simply 'as-is'. useful in html from sql + else + lang=get_untaint_lang(params, 0); { Value& vbody=params.as_junction(params.count()-1, "body must be code"); @@ -127,15 +130,9 @@ static void _untaint(Request& r, MethodP static void _taint(Request& r, MethodParams& params) { String::Language lang; if(params.count()==1) - lang=String::L_TAINTED; // mark as simply 'tainted'. useful in table:set - else { - const String& lang_name=params.as_string(0, "lang must be string"); - lang=untaint_lang_name2enum.get(lang_name); - if(!lang) - throw Exception(0, - &lang_name, - "invalid taint language"); - } + lang=String::L_TAINTED; // mark as simply 'tainted'. useful in table:create + else + lang=get_untaint_lang(params, 0); { Value& vbody=params.as_no_junction(params.count()-1, "body must not be code"); @@ -243,41 +240,59 @@ static void _while(Request& r, MethodPar Temp_hash_value cycle_data_setter(r.classes_conf, cycle_data_name, /*any not null flag*/&r); - Value& vcondition=params.as_junction(0, "condition must be expression"); + Value& vcondition=params.as_expression(0, "condition must be number, bool or expression"); + Value& body_code=params.as_junction(1, "body must be code"); Value* delim_maybe_code=params.count()>2?¶ms[2]:0; // while... int endless_loop_count=0; - bool need_delim=false; - while(true) { - if(++endless_loop_count>=MAX_LOOPS) // endless loop? - throw Exception(PARSER_RUNTIME, - 0, - "endless loop detected"); + if(delim_maybe_code){ + bool need_delim=false; + while(true) { + if(++endless_loop_count>=MAX_LOOPS) // endless loop? + throw Exception(PARSER_RUNTIME, + 0, + "endless loop detected"); - bool condition=r.process_to_value(vcondition, - false/*don't intercept string*/).as_bool(); - if(!condition) // ...condition is true - break; + if(!r.process_to_value(vcondition, false/*don't intercept string*/).as_bool()) + break; + + StringOrValue sv_processed=r.process(body_code); + Request::Skip lskip=r.get_skip(); r.set_skip(Request::SKIP_NOTHING); + const String* s_processed=sv_processed.get_string(); + if(delim_maybe_code && s_processed && s_processed->length()) { // delimiter set and we have body + if(need_delim) // need delim & iteration produced string? + r.write_pass_lang(r.process(*delim_maybe_code)); + else + need_delim=true; + } + r.write_pass_lang(sv_processed); - StringOrValue sv_processed=r.process(body_code); - Request::Skip lskip=r.get_skip(); r.set_skip(Request::SKIP_NOTHING); - const String* s_processed=sv_processed.get_string(); - if(delim_maybe_code && s_processed && s_processed->length()) { // delimiter set and we have body - if(need_delim) // need delim & iteration produced string? - r.write_pass_lang(r.process(*delim_maybe_code)); - need_delim=true; + if(lskip==Request::SKIP_BREAK) + break; } - r.write_pass_lang(sv_processed); + } else { + while(true) { + if(++endless_loop_count>=MAX_LOOPS) // endless loop? + throw Exception(PARSER_RUNTIME, + 0, + "endless loop detected"); - if(lskip==Request::SKIP_BREAK) - break; + if(!r.process_to_value(vcondition, false/*don't intercept string*/).as_bool()) + break; + + r.process_write(body_code); + Request::Skip lskip=r.get_skip(); r.set_skip(Request::SKIP_NOTHING); + + if(lskip==Request::SKIP_BREAK) + break; + } } } static void _use(Request& r, MethodParams& params) { - Value& vfile=params.as_no_junction(0, "file name must not be code"); + Value& vfile=params.as_no_junction(0, FILE_NAME_MUST_NOT_BE_CODE); r.use_file(r.main_class, vfile.as_string()); } @@ -314,26 +329,40 @@ static void _for(Request& r, MethodParam 0, "endless loop detected"); - bool need_delim=false; VInt* vint=new VInt(0); VMethodFrame& caller=*r.get_method_frame()->caller(); caller.put_element(caller, var_name, vint, false); - for(int i=from; i<=to; i++) { - vint->set_int(i); + if(delim_maybe_code){ + bool need_delim=false; + + for(int i=from; i<=to; i++) { + vint->set_int(i); - StringOrValue sv_processed=r.process(body_code); - Request::Skip lskip=r.get_skip(); r.set_skip(Request::SKIP_NOTHING); - const String* s_processed=sv_processed.get_string(); - if(delim_maybe_code && s_processed && s_processed->length()) { // delimiter set and we have body - if(need_delim) // need delim & iteration produced string? - r.write_pass_lang(r.process(*delim_maybe_code)); - need_delim=true; + StringOrValue sv_processed=r.process(body_code); + Request::Skip lskip=r.get_skip(); r.set_skip(Request::SKIP_NOTHING); + const String* s_processed=sv_processed.get_string(); + if(s_processed && s_processed->length()) { // delimiter set and we have body + if(need_delim) // need delim & iteration produced string? + r.write_pass_lang(r.process(*delim_maybe_code)); + else + need_delim=true; + } + r.write_pass_lang(sv_processed); + + if(lskip==Request::SKIP_BREAK) + break; } - r.write_pass_lang(sv_processed); + } else { + for(int i=from; i<=to; i++) { + vint->set_int(i); + + r.process_write(body_code); + Request::Skip lskip=r.get_skip(); r.set_skip(Request::SKIP_NOTHING); - if(lskip==Request::SKIP_BREAK) - break; + if(lskip==Request::SKIP_BREAK) + break; + } } } @@ -343,8 +372,12 @@ static void _eval(Request& r, MethodPara Value& value_result=r.process_to_value(expr, false/*don't intercept string*/).as_expr_result(); if(params.count()>1) { - Value& fmt=params.as_no_junction(1, "fmt must not be code"); - r.write_no_lang(String(format(value_result.as_double(), fmt.as_string().cstrm()))); + const String& fmt=params.as_string(1, "fmt must be string").trim(); + if(fmt.is_empty()){ + r.write_no_lang(value_result); + } else { + r.write_no_lang(String(format(value_result.as_double(), fmt.cstrm()))); + } } else r.write_no_lang(value_result); } @@ -370,7 +403,8 @@ gettimeofday(&mt[0],NULL); // connect SQL_Connection* connection=SQL_driver_manager->get_connection(url.as_string(), protocol2driver_and_client, - r.charsets.source().NAME().cstr()); + r.charsets.source().NAME().cstr(), + r.request_info.document_root); #ifdef RESOURCES_DEBUG //measure:after connect @@ -399,12 +433,22 @@ r.sql_connect_time+=t[1]-t[0]; class Switch_data: public PA_Object { public: Request& r; - Value& searching; + const String* searching_string; + double searching_double; Value* found; Value* _default; public: Switch_data(Request& ar, Value& asearching): - r(ar), searching(asearching) {} + r(ar) + { + if(asearching.is_string() || asearching.is_void()){ + searching_string=&asearching.as_string(); + searching_double=0; + } else { + searching_string=0; + searching_double=asearching.as_double(); + } + } }; #endif static void _switch(Request& r, MethodParams& params) { @@ -427,32 +471,36 @@ static void _case(Request& r, MethodPara 0, "without switch"); + if(data->found) // matches already was found + return; + int count=params.count(); - Value& code=params.as_junction(--count, "case result must be code"); + Value* code=¶ms.as_expression(--count, "case result must be code"); + +#ifdef USE_DESTRUCTORS + Junction *j=code->get_junction(); + if (j){ + code=new VJunction(j->self,j->method,j->method_frame,j->rcontext,j->wcontext,j->code); + if (j->wcontext) j->wcontext->attach_junction((VJunction *)code); + } +#endif - Value& searching=data->searching; - bool we_are_searching_string_or_void=searching.is_string() || searching.is_void(); - for(int i=0; i_default=&code; - break; + if(value.is_string() && value.as_string() == CASE_DEFAULT_VALUE){ + data->_default=code; + continue; } bool matches; - if(we_are_searching_string_or_void) - matches=searching.as_string() == value.as_string(); + if(data->searching_string) + matches=(*data->searching_string) == value.as_string(); else - matches=searching.as_double() == value.as_double(); + matches=data->searching_double == value.as_double(); - if(matches) { - if(data->found) - throw Exception(PARSER_RUNTIME, - 0, - "duplicate found"); - - data->found=&code; + if(matches){ + data->found=code; break; } } @@ -470,9 +518,11 @@ struct Try_catch_result { template static Try_catch_result try_catch(Request& r, StringOrValue body_code(Request&, I), I info, - Value* catch_code, bool could_be_handled_by_caller=false) + Value* catch_code, + bool could_be_handled_by_caller=false) { Try_catch_result result; + if(!catch_code) { result.processed_code=body_code(r, info); return result; @@ -492,6 +542,7 @@ static Try_catch_result try_catch(Reques Value* saved_exception_var_value=method_frame->get_element(exception_var_name, *method_frame, false); VMethodFrame& frame=*junction->method_frame; frame.put_element(frame, exception_var_name, &details.vhash, false); + result.processed_code=r.process(*catch_code); // retriving $exception.handled, restoring $exception var @@ -508,7 +559,7 @@ static Try_catch_result try_catch(Reques bhandled=false; } else - bhandled=vhandled->as_bool(); + bhandled=vhandled->as_bool(); } if(!bhandled) { @@ -516,6 +567,7 @@ static Try_catch_result try_catch(Reques rethrow; } } + return result; } @@ -535,7 +587,7 @@ struct Data_string_serialized_prolog { #endif void cache_delete(const String& file_spec) { - file_delete(file_spec, false/*fail_on_read_problem*/); + file_delete(file_spec, false/*fail_on_problem*/); } #ifndef DOXYGEN @@ -547,7 +599,8 @@ public: struct Locked_process_and_cache_put_action_info { Request *r; Cache_scope *scope; - Value* body_code; Value* catch_code; + Value* body_code; + Value* catch_code; const String* processed_code; }; #endif @@ -566,7 +619,6 @@ static void locked_process_and_cache_put Locked_process_and_cache_put_action_info& info= *static_cast(context); - const String* body_from_disk=info.scope->body_from_disk; // body->process Try_catch_result result=try_catch(*info.r, @@ -610,11 +662,14 @@ const String* locked_process_and_cache_p const String* result=file_write_action_under_lock( file_spec, - "cache_put", locked_process_and_cache_put_action, &info, + "cache_put", + locked_process_and_cache_put_action, + &info, false/*as_text*/, false/*do_append*/, - false/*block*/, - false/*fail on lock problem*/) ? info.processed_code: 0; + false/*block == don't wait till other thread release lock*/, + false/*dun throw exception if lock failed*/) ? info.processed_code: 0; + time_t now=time(0); if(scope.expires<=now) cache_delete(file_spec); @@ -666,9 +721,8 @@ static const String& as_file_spec(Reques return r.absolute(params.as_string(index, "filespec must be string")); } static void _cache(Request& r, MethodParams& params) { - if(params.count()==0) - { - // return current expiration time + if(params.count()==0) { + // ^cache[] -- return current expiration time Cache_scope* scope=static_cast(r.classes_conf.get(cache_data_name)); if(!scope) throw Exception(PARSER_RUNTIME, @@ -680,8 +734,8 @@ static void _cache(Request& r, MethodPar time_t now=time(0); - // ^cache[filename] ^cache(seconds) ^cache[expires date] if(params.count()==1) { + // ^cache[filename] ^cache(seconds) ^cache[expires date] if(params[0].is_string()) { // filename? cache_delete(as_file_spec(r, params, 0)); return; @@ -705,7 +759,7 @@ static void _cache(Request& r, MethodPar "invalid number of parameters"); // file_spec, expires, body code - const String& file_spec=r.absolute(params.as_string(0, "filespec must be string")); + const String& file_spec=as_file_spec(r, params, 0); Cache_scope scope={as_expires(r, params, 1, now), 0}; @@ -716,63 +770,41 @@ static void _cache(Request& r, MethodPar if(params.count()>3) catch_code=¶ms.as_junction(3, "catch_code must be code"); - if(scope.expires>now) { // valid 'expires' specified? try cached copy... - // hence we don't hope to have unary create/lockEX - // we need some plan to live in a life like that, so... - // worst races plan: - // A B - // open - // |open - // lockSH - // |nonblocking-lockEX fails - // unlockSH - // close, cache_get returns 0 - // open - // nonblocking-lockEX succeeds; process, write, close - // |retry1: open - // ... - // |lockSH succeeds; ... - - for(int retry=0; retry<2; retry++) { - Cache_get_result cached=cache_get(r.charsets, file_spec, now); - if(cached.body) { // have cached copy - if(cached.expired) - scope.body_from_disk=cached.body; // storing for user to retrive it with ^cache[] - else // and it's not expired yet - { - // write it out - r.write_assign_lang(*cached.body); - // happy with it - return; - } - } + if(scope.expires>now) { + Cache_get_result cached=cache_get(r.charsets, file_spec, now); - // non-blocked lock; process; cache it - if(const String* processed_body= - locked_process_and_cache_put(r, body_code, catch_code, scope, file_spec)) { - // write it out - r.write_assign_lang(*processed_body); - // happy with it - return; - } else { // somebody writing result right now - pa_sleep(0, 500000); // waiting half a second - retry=0; // prolonging our wait, than could cache_get it, without processing body_code - } - } - throw Exception(0, - &file_spec, - "locking problem"); + if(cached.body) { // have cached copy + if(cached.expired) { + scope.body_from_disk=cached.body; // storing for user to retrive it with ^cache[] + } else { + // and it's not expired yet write it out + r.write_assign_lang(*cached.body); + // happy with it + return; + } + } + + // no cached info or it's already expired + + // trying to process it under lock and store result in file + const String* processed_body=locked_process_and_cache_put(r, body_code, catch_code, scope, file_spec); + if(processed_body){ + // write it out + r.write_assign_lang(*processed_body); + // happy with it + return; + } else { + // we fail while get exclusive lock. nvm, we just execute body_code a bit later + } } else { // instructed not to cache; forget cached copy cache_delete(file_spec); - // process - const String& processed_body=r.process_to_string(body_code); - // write it out - r.write_assign_lang(processed_body); - // happy with it - return; } - // never reached + + // process without cacheing + const String& processed_body=r.process_to_string(body_code); + // write it out + r.write_assign_lang(processed_body); } static StringOrValue process_try_body_code(Request& r, Value* body_code) { @@ -781,22 +813,37 @@ static StringOrValue process_try_body_co static void _try_operator(Request& r, MethodParams& params) { Value& body_code=params.as_junction(0, "body_code must be code"); Value& catch_code=params.as_junction(1, "catch_code must be code"); + Value* finally_code=(params.count()==3) ? ¶ms.as_junction(2, "finally_code must be code") : 0; - Try_catch_result result=try_catch(r, - process_try_body_code, &body_code, - &catch_code); - - if(result.exception_should_be_handled) - throw Exception(PARSER_RUNTIME, - result.exception_should_be_handled, - "catch block must set $exception.handled to some boolean value, not string"); + Try_catch_result result; + StringOrValue finally_result; + try{ + result=try_catch(r, + process_try_body_code, &body_code, + &catch_code); + if(result.exception_should_be_handled) + throw Exception(PARSER_RUNTIME, + result.exception_should_be_handled, + "catch block must set $exception.handled to some boolean value, not string"); + } catch(...){ + if(finally_code) + finally_result=r.process(*finally_code); + rethrow; + } + + if(finally_code) + finally_result=r.process(*finally_code); // write out processed body_code or catch_code r.write_pass_lang(result.processed_code); + + // write out processed finally code + if(finally_code) + r.write_pass_lang(finally_result); } static void _throw_operator(Request&, MethodParams& params) { - if(params.count()==1) { + if(params.count()==1 && !params[0].is_string()) { if(HashStringValue *hash=params[0].get_hash()) { const char* type=0; if(Value* value=hash->get(exception_type_part_name)) @@ -814,13 +861,12 @@ static void _throw_operator(Request&, Me } else throw Exception(PARSER_RUNTIME, 0, - "one-param version has hash param"); + "one-param version has hash or string param"); } else { const char* type=params.as_string(0, "type must be string").cstr(); - const String& source=params.as_string(1, "source must be string"); - const char* comment=params.count()>2? params.as_string(2, "comment must be string").cstr() - :0; - throw Exception(type, &source, "%s", comment?comment:""); + const String* source=params.count()>1? ¶ms.as_string(1, "source must be string"):0; + const char* comment=params.count()>2? params.as_string(2, "comment must be string").cstr():0; + throw Exception(type, source, "%s", comment?comment:""); } } @@ -848,36 +894,38 @@ VClassMAIN::VClassMAIN(): VClass() { // ^if(condition){code-when-true} // ^if(condition){code-when-true}{code-when-false} - add_native_method("if", Method::CT_ANY, _if, 2, 3); + add_native_method("if", Method::CT_ANY, _if, 2, 3, Method::CO_WITHOUT_FRAME); - // ^untaint[as-is|uri|sql|js|html|html-typo]{code} - add_native_method("untaint", Method::CT_ANY, _untaint, 1, 2); + // ^untaint[as-is|uri|sql|js|html|html-typo|regex]{code} + add_native_method("untaint", Method::CT_ANY, _untaint, 1, 2, Method::CO_NONE); - // ^taint[as-is|uri|sql|js|html|html-typo]{code} - add_native_method("taint", Method::CT_ANY, _taint, 1, 2); + // ^taint[as-is|uri|sql|js|html|html-typo|regex]{code} + add_native_method("taint", Method::CT_ANY, _taint, 1, 2, Method::CO_NONE); // ^process[code] add_native_method("process", Method::CT_ANY, _process, 1, 3); // ^rem{code} - add_native_method("rem", Method::CT_ANY, _rem, 1, 10000); + add_native_method("rem", Method::CT_ANY, _rem, 1, 10000, Method::CO_WITHOUT_FRAME); // ^while(condition){code} - add_native_method("while", Method::CT_ANY, _while, 2, 3); + add_native_method("while", Method::CT_ANY, _while, 2, 3, Method::CO_WITHOUT_FRAME); // ^use[file] add_native_method("use", Method::CT_ANY, _use, 1, 1); // ^break[] - add_native_method("break", Method::CT_ANY, _break, 0, 0); + add_native_method("break", Method::CT_ANY, _break, 0, 0, Method::CO_WITHOUT_FRAME); + // ^continue[] - add_native_method("continue", Method::CT_ANY, _continue, 0, 0); + add_native_method("continue", Method::CT_ANY, _continue, 0, 0, Method::CO_WITHOUT_FRAME); + // ^for[i](from-number;to-number-inclusive){code}[delim] - add_native_method("for", Method::CT_ANY, _for, 3+1, 3+1+1); + add_native_method("for", Method::CT_ANY, _for, 3+1, 3+1+1, Method::CO_WITHOUT_WCONTEXT); // ^eval(expr) // ^eval(expr)[format] - add_native_method("eval", Method::CT_ANY, _eval, 1, 2); + add_native_method("eval", Method::CT_ANY, _eval, 1, 2, Method::CO_WITHOUT_FRAME); // ^connect[protocol://user:pass@host[:port]/database]{code with ^sql-s} add_native_method("connect", Method::CT_ANY, _connect, 2, 2); @@ -891,15 +939,15 @@ VClassMAIN::VClassMAIN(): VClass() { // switch // ^switch[value]{cases} - add_native_method("switch", Method::CT_ANY, _switch, 2, 2); + add_native_method("switch", Method::CT_ANY, _switch, 2, 2, Method::CO_WITHOUT_FRAME); // ^case[value]{code} - add_native_method("case", Method::CT_ANY, _case, 2, 10000); + add_native_method("case", Method::CT_ANY, _case, 2, 10000, Method::CO_WITHOUT_FRAME); // try-catch // ^try{code}{catch code} - add_native_method("try", Method::CT_ANY, _try_operator, 2, 2); + add_native_method("try", Method::CT_ANY, _try_operator, 2, 3, Method::CO_WITHOUT_FRAME); // ^throw[$exception hash] // ^throw[type;source;comment] add_native_method("throw", Method::CT_ANY, _throw_operator, 1, 3);