--- parser3/src/classes/op.C 2002/03/27 15:30:34 1.78 +++ parser3/src/classes/op.C 2016/11/01 23:10:40 1.243 @@ -1,660 +1,977 @@ /** @file Parser: parser @b operators. - Copyright (c) 2001, 2002 ArtLebedev Group (http://www.artlebedev.com) + Copyright (c) 2001-2015 Art. Lebedev Studio (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) - - $Id: op.C,v 1.78 2002/03/27 15:30:34 paf Exp $ */ #include "classes.h" +#include "pa_vmethod_frame.h" + #include "pa_common.h" +#include "pa_os.h" #include "pa_request.h" #include "pa_vint.h" #include "pa_sql_connection.h" +#include "pa_vdate.h" +#include "pa_vmethod_frame.h" +#include "pa_vclass.h" +#include "pa_charset.h" -// limits - -#define MAX_LOOPS 10000 +volatile const char * IDENT_OP_C="$Id: op.C,v 1.243 2016/11/01 23:10:40 moko Exp $"; // defines -#define OP_CLASS_NAME "OP" +#define CASE_DEFAULT_VALUE "DEFAULT" // class -class MOP : public Methoded { +class VClassMAIN: public VClass { public: - MOP(Pool& pool); -public: // Methoded - bool used_directly() { return true; } - void configure_user(Request& r); - -private: - String main_sql_name; - String main_sql_drivers_name; + VClassMAIN(); }; -// methods +// defines for statics -static void _if(Request& r, const String&, MethodParams *params) { - Value& condition_code=params->as_junction(0, "condition must be expression"); +#define SWITCH_DATA_NAME "SWITCH-DATA" +#define CACHE_DATA_NAME "CACHE-DATA" - bool condition=r.process(condition_code, - 0/*no name*/, - false/*don't intercept string*/).as_bool(); - if(condition) - r.write_pass_lang(r.process(params->as_junction(1, "'then' parameter must be code"))); - else if(params->size()>2) - r.write_pass_lang(r.process(params->as_junction(2, "'else' parameter must be code"))); -} - -static void _untaint(Request& r, const String& method_name, MethodParams *params) { - Pool& pool=r.pool(); - - uchar lang; - if(params->size()==1) - lang=String::UL_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_int(lang_name); - if(!lang) - throw Exception(0, - &lang_name, - "invalid taint language"); +#define EXCEPTION_VAR_NAME "exception" + +// statics + +//^switch ^case +static const String switch_data_name(SWITCH_DATA_NAME); +//^cache +static const String cache_data_name(CACHE_DATA_NAME); + +static const String exception_var_name(EXCEPTION_VAR_NAME); + + +// local defines + +#define CACHE_EXCEPTION_HANDLED_CACHE_NAME "cache" + +// helpers + +class Untaint_lang_name2enum: public HashString { +public: + Untaint_lang_name2enum() { + #define ULN(name, LANG) put(name, (value_type)(String::L_##LANG)); + ULN("clean", CLEAN); + ULN("as-is", AS_IS); + ULN("optimized-as-is", AS_IS|String::L_OPTIMIZE_BIT); + ULN("file-spec", FILE_SPEC); + ULN("http-header", HTTP_HEADER); + ULN("mail-header", MAIL_HEADER); + ULN("uri", URI); + ULN("sql", SQL); + ULN("js", JS); + ULN("xml", XML); + ULN("optimized-xml", XML|String::L_OPTIMIZE_BIT); + ULN("html", HTML); + ULN("optimized-html", HTML|String::L_OPTIMIZE_BIT); + ULN("regex", REGEX); + ULN("parser-code", PARSER_CODE); + ULN("json", JSON); + #undef ULN } +} untaint_lang_name2enum; + +// methods + +static void _if(Request& r, MethodParams& params) { + size_t max_param=params.count()-1; + size_t i=0; + do { + bool condition=params.as_bool(i, "condition must be expression", r); + if(condition) { + r.process_write(params[i+1]); + return; + } + i+=2; + } while (i < max_param); + + if(i == max_param) + r.process_write(params[i]); +} + +String::Language get_untaint_lang(const String& lang_name){ + 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 'as-is'. useful in html from sql + else + lang=get_untaint_lang(params.as_string(0, "lang must be string")); + + Value& vbody=params.as_junction(params.count()-1, "body must be code"); + Value& result=r.process(vbody); + + if(const String* string=result.get_string()){ + String &untainted=*new String(); + string->append_to(untainted, lang); // mark all tainted to specified language + r.write(untainted); + } else + r.write(result); // this is not normal, just backward compatibility +} + +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:create + else + lang=get_untaint_lang(params.as_string(0, "lang must be string")); { - Value& vbody=params->as_junction(params->size()-1, "body must be code"); + Value& vbody=params.as_no_junction(params.count()-1, "body must not be code"); - Temp_lang temp_lang(r, lang); // set temporarily specified ^untaint[language; - r.write_pass_lang(r.process(vbody)); // process marking tainted with that lang + String result(vbody.as_string(), lang); // force result language to specified + r.write(result); } } -static void _taint(Request& r, const String&, MethodParams *params) { - Pool& pool=r.pool(); +static void _apply_taint(Request& r, MethodParams& params) { + String::Language lang=params.count()==1 ? String::L_AS_IS : get_untaint_lang(params.as_string(0, "lang must be string")); + const String &sbody=params.as_string(params.count()-1, "body must be string"); + String::Body result_body=sbody.cstr_to_string_body_untaint(lang, r.connection(false), &r.charsets); + r.write(*new String(result_body, String::L_AS_IS)); +} + +static void _process(Request& r, MethodParams& params) { + Method* main_method; - uchar lang; - if(params->size()==1) - lang=String::UL_TAINTED; // mark as simply 'tainted'. useful in table:set + size_t index=0; + Value* target_self; + Value& maybe_target_self=params[index]; + if(maybe_target_self.get_string() || maybe_target_self.get_junction()) + target_self=&r.get_method_frame()->caller()->self(); else { - const String& lang_name=params->as_string(0, "lang must be string"); - lang=untaint_lang_name2enum->get_int(lang_name); - if(!lang) - throw Exception(0, - &lang_name, - "invalid taint language"); + target_self=&maybe_target_self; + if(params.count()==1) + throw Exception(PARSER_RUNTIME, 0, "no body specified"); + index++; } { - Value& vbody=params->as_no_junction(params->size()-1, "body must not be code"); - - String result(r.pool()); - result.append( - vbody.as_string(), // process marking tainted with that lang - lang, true); // force result language to specified - r.write_pass_lang(result); - } -} + VStateless_class *target_class=target_self->get_class(); + if(!target_class) + throw Exception(PARSER_RUNTIME, 0, "no target class"); -static void _process(Request& r, const String& method_name, MethodParams *params) { - // calculate pseudo file name of processed chars - // would be something like "/some/file(4) process" - char place[MAX_STRING]; -#ifndef NO_STRING_ORIGIN - const Origin& origin=method_name.origin(); - snprintf(place, MAX_STRING, "%s(%d) %s", - origin.file, 1+origin.line, - method_name.cstr()); -#else - strncpy(place, method_name.cstr(), MAX_STRING-1); place[MAX_STRING-1]=0; -#endif - - VStateless_class& self_class=*r.self->get_class(); - const Method *main_method; - { - // temporary remove language change - Temp_lang temp_lang(r, String::UL_PASS_APPENDED); // temporary zero @main so to maybe-replace it in processed code - Temp_method temp_method_main(self_class, *main_method_name, 0); - // temporary zero @auto so it wouldn't be auto-called in Request::use_buf - Temp_method temp_method_auto(self_class, *auto_method_name, 0); - + Temp_method temp_method_main(*target_class, main_method_name, 0); + + const String* main_alias=0; + const String* file_alias=0; + int line_no_alias_offset=0; + bool allow_class_replace=false; + + size_t options_index=index+1; + if(options_indexas_string(); + } else if(key == "file") { + valid_options++; + file_alias=&value->as_string(); + } else if(key == "lineno") { + valid_options++; + line_no_alias_offset=value->as_int(); + } else if(key == "replace") { + valid_options++; + allow_class_replace=r.process(*value).as_bool(); + } + } + + if(valid_options!=options->count()) + throw Exception(PARSER_RUNTIME, 0, CALLED_WITH_INVALID_OPTION); + } + + uint processe_file_no=file_alias ? r.register_file(r.absolute(*file_alias)) : pseudo_file_no__process; + // process...{string} + Value& vjunction=params.as_junction(index, "body must be code"); // evaluate source to process - const String& source= - r.process(params->as_junction(0, "body must be code")).as_string(); + const String& source=r.process_to_string(vjunction); + + Temp_class_replace class_replace(r, allow_class_replace); + + r.use_buf(*target_class, source.untaint_cstr(String::L_PARSER_CODE, r.connection(false)), main_alias, processe_file_no, line_no_alias_offset); - // process source code, append processed methods to 'self' class - // maybe-define new @main - r.use_buf( - source.cstr(String::UL_UNSPECIFIED, r.connection(0)), - place, - &self_class); - // main_method - main_method=self_class.get_method(*main_method_name); + main_method=target_class->get_method(main_method_name); } // after restoring current-request-lang // maybe-execute @main[] if(main_method) { - // execute! - r.execute(*main_method->parser_code); + VMethodFrame frame(*main_method, r.get_method_frame()->caller(), *target_self); + frame.empty_params(); + r.call(frame); + r.write(frame.result()); } } -static void _rem(Request& r, const String&, MethodParams *params) { - params->as_junction(0, "body must be code"); +static void _rem(Request&, MethodParams& params) { + params.as_junction(0, "body must be code"); } -static void _while(Request& r, const String& method_name, MethodParams *params) { - Pool& pool=r.pool(); +static void _while(Request& r, MethodParams& params) { + InCycle temp(r); + + Value& vcondition=params.as_expression(0, "condition must be number, bool or expression"); - Value& vcondition=params->as_junction(0, "condition must be expression"); - Value& body=params->as_junction(1, "body must be code"); + 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; - while(true) { - if(++endless_loop_count>=MAX_LOOPS) // endless loop? - throw Exception("parser.runtime", - &method_name, - "endless loop detected"); - - bool condition= - r.process( - vcondition, - 0/*no name*/, - false/*don't intercept string*/).as_bool(); - if(!condition) // ...condition is true - break; + if(delim_maybe_code){ // delimiter set + bool need_delim=false; + while(true) { + if(++endless_loop_count>=pa_loop_limit) // endless loop? + throw Exception(PARSER_RUNTIME, 0, "endless loop detected"); + + if(!r.process(vcondition).as_bool()) + break; + + Value& 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->is_empty()) { // we have body + if(need_delim) // need delim & iteration produced string? + r.write(r.process(*delim_maybe_code)); + else + need_delim=true; + } + r.write(sv_processed); + + if(lskip==Request::SKIP_BREAK) + break; + } + } else { + while(true) { + if(++endless_loop_count>=pa_loop_limit) // endless loop? + throw Exception(PARSER_RUNTIME, 0, "endless loop detected"); + + if(!r.process(vcondition).as_bool()) + break; + + r.process_write(body_code); + Request::Skip lskip=r.get_skip(); r.set_skip(Request::SKIP_NOTHING); - // write processed body - r.write_pass_lang(r.process(body)); + if(lskip==Request::SKIP_BREAK) + break; + } } } -static void _use(Request& r, const String& method_name, MethodParams *params) { - Value& vfile=params->as_no_junction(0, "file name must not be code"); - r.use_file(vfile.as_string()); +static void _use(Request& r, MethodParams& params) { + Value& vfile=params.as_no_junction(0, FILE_NAME_MUST_NOT_BE_CODE); + + bool allow_class_replace=false; + + if(params.count()==2) + if(HashStringValue* options=params.as_hash(1)) { + int valid_options=0; + for(HashStringValue::Iterator i(*options); i; i.next() ){ + + String::Body key=i.key(); + Value* value=i.value(); + + if(key == "replace") { + valid_options++; + allow_class_replace=r.process(*value).as_bool(); + } + + if(valid_options!=options->count()) + throw Exception(PARSER_RUNTIME, 0, CALLED_WITH_INVALID_OPTION); + } + } + + Temp_class_replace class_replace(r, allow_class_replace); + + // _use could be called from the parser3 method only, so caller is always defined + r.use_file(r.main_class, vfile.as_string(), r.get_method_filename(&r.get_method_frame()->caller()->method)); +} + +static void set_skip(Request& r, Request::Skip askip) { + if(!r.get_in_cycle()) + throw Exception(askip==Request::SKIP_BREAK ? "parser.break" : "parser.continue", 0, "without cycle"); + r.set_skip(askip); } -static void _for(Request& r, const String& method_name, MethodParams *params) { - Pool& pool=r.pool(); - const String& var_name=params->as_string(0, "var name must be string"); - int from=params->as_int(1, "from must be int", r); - int to=params->as_int(2, "to must be int", r); - Value& body_code=params->as_junction(3, "body must be code"); - Value *delim_maybe_code=params->size()>4?¶ms->get(4):0; +static void _break(Request& r, MethodParams& params) { + if(!params.count() || params.as_bool(0, "condition must be expression", r)) set_skip(r, Request::SKIP_BREAK); +} + +static void _continue(Request& r, MethodParams& params) { + if(!params.count() || params.as_bool(0, "condition must be expression", r)) set_skip(r, Request::SKIP_CONTINUE); +} - if(to-from>=MAX_LOOPS) // too long loop? - throw Exception("parser.runtime", - &method_name, - "endless loop detected"); +static void _for(Request& r, MethodParams& params) { + InCycle temp(r); - bool need_delim=false; - VInt *vint=new(pool) VInt(pool, 0); - for(int i=from; i<=to; i++) { - vint->set_int(i); - r.root->put_element(var_name, vint); + const String& var_name=params.as_string(0, "var name must be string"); + int from=params.as_int(1, "from must be int", r); + int to=params.as_int(2, "to must be int", r); + Value& body_code=params.as_junction(3, "body must be code"); + Value* delim_maybe_code=params.count()>4?¶ms[4]:0; - Value& processed_body=r.process(body_code); - if(delim_maybe_code) { // delimiter set? - const String *string=processed_body.get_string(); - if(need_delim && string && string->size()) // need delim & iteration produced string? - r.write_pass_lang(r.process(*delim_maybe_code)); - need_delim=true; + if(to-from>=pa_loop_limit) // too long loop? + throw Exception(PARSER_RUNTIME, 0, "endless loop detected"); + + VInt* vint=new VInt(0); + + VMethodFrame& caller=*r.get_method_frame()->caller(); + r.put_element(caller, var_name, vint); + if(delim_maybe_code){ // delimiter set + bool need_delim=false; + + for(int i=from; i<=to; i++) { + vint->set_int(i); + + Value& 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->is_empty()) { // we have body + if(need_delim) // need delim & iteration produced string? + r.write(r.process(*delim_maybe_code)); + else + need_delim=true; + } + r.write(sv_processed); + + if(lskip==Request::SKIP_BREAK) + break; + } + } 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; } - r.write_pass_lang(processed_body); } } -static void _eval(Request& r, const String& method_name, MethodParams *params) { - Value& expr=params->as_junction(0, "need expression"); +static void _eval(Request& r, MethodParams& params) { + Value& expr=params.as_junction(0, "need expression"); // evaluate expresion - Value *result=r.process(expr, - 0/*no name YET*/, - true/*don't intercept string*/).as_expr_result(); - if(params->size()>1) { - Value& fmt=params->as_no_junction(1, "fmt must not be code"); - - Pool& pool=r.pool(); - String& string=*new(pool) String(pool); - string.APPEND_CONST(format(pool, result->as_double(), fmt.as_string().cstr())); - result=new(pool) VString(string); - } - result->set_name(method_name); - r.write_no_lang(*result); -} - -static void _connect(Request& r, const String& method_name, MethodParams *params) { - Pool& pool=r.pool(); -#ifdef RESOURCES_DEBUG -struct timeval mt[2]; -#endif - Value& url=params->as_no_junction(0, "url must not be code"); - Value& body_code=params->as_junction(1, "body must be code"); + Value& value_result=r.process(expr).as_expr_result(); + if(params.count()>1) { + const String& fmt=params.as_string(1, "fmt must be string").trim(); + if(fmt.is_empty()){ + r.write(value_result); + } else { + r.write(String(format(value_result.as_double(), fmt.cstrm()))); + } + } else + r.write(value_result); +} - Table *protocol2driver_and_client= - static_cast(r.classes_conf.get(r.OP.name())); +static void _connect(Request& r, MethodParams& params) { + Value& url=params.as_no_junction(0, "url must not be code"); + Value& body_code=params.as_junction(1, "body must be code"); + + Table* protocol2driver_and_client=0; + if(Value* sql=r.main_class.get_element(String(MAIN_SQL_NAME))) { + if(Value* element=sql->get_element(String(MAIN_SQL_DRIVERS_NAME))) { + protocol2driver_and_client=element->get_table(); + } + } -#ifdef RESOURCES_DEBUG -//measure:before -gettimeofday(&mt[0],NULL); -#endif // connect - SQL_Connection_ptr connection=SQL_driver_manager->get_connection( - url.as_string(), method_name, protocol2driver_and_client); + SQL_Connection* connection=SQL_driver_manager->get_connection(url.as_string(), + protocol2driver_and_client, + r.charsets.source().NAME().cstr(), + r.request_info.document_root); -#ifdef RESOURCES_DEBUG -//measure:after connect -gettimeofday(&mt[1],NULL); - -double t[2]; -for(int i=0;i<2;i++) - t[i]=mt[i].tv_sec+mt[i].tv_usec/1000000.0; + Temp_connection temp_connection(r, connection); -r.sql_connect_time+=t[1]-t[0]; -#endif - Temp_connection temp_connection(r, connection.get()); // execute body try { - r.write_assign_lang(r.process(body_code)); + r.process_write(body_code); + connection->commit(); + connection->close(); } catch(...) { // process problem - connection->mark_to_rollback(); - /*re*/throw; + connection->rollback(); + connection->close(); + rethrow; } } #ifndef DOXYGEN -struct Switch_data { - Value *searching; - Value *found; - Value *_default; +class Switch_data: public PA_Object { +public: + Request& r; + const String* searching_string; + double searching_double; + bool searching_bool; + Value* found; + Value* _default; +public: + Switch_data(Request& ar, Value& asearching): + r(ar) + { + if(asearching.is_string() || asearching.is_void()){ + searching_string=&asearching.as_string(); + searching_double=0; + searching_bool=false; + } else { + searching_string=0; + searching_double=asearching.as_double(); + searching_bool=asearching.is_bool(); + } + } }; #endif -static void _switch(Request& r, const String&, MethodParams *params) { - void *backup=r.classes_conf.get(*switch_data_name); - Switch_data data={&r.process(params->get(0))}; - r.classes_conf.put(*switch_data_name, &data); - - r.process(params->as_junction(1, "switch cases must be code")); // and ignore result - - r.classes_conf.put(*switch_data_name, backup); - - if(Value *code=data.found ? data.found : data._default) - r.write_pass_lang(r.process(*code)); +static void _switch(Request& r, MethodParams& params) { + Switch_data* data=new Switch_data(r, r.process(params[0])); + Temp_hash_value, void*> switch_data_setter(&r.classes_conf, switch_data_name, data); + + Value& cases_code=params.as_junction(1, "switch cases must be code"); + // execution of found ^case[...]{code} must be in context of ^switch[...]{code} + // because of stacked WWrapper used there as wcontext + r.process(cases_code); + if(Value* selected_code=data->found? data->found: data->_default) + r.write(r.process(*selected_code)); } -static void _case(Request& r, const String& method_name, MethodParams *params) { - Pool& pool=r.pool(); - - Switch_data *data=static_cast(r.classes_conf.get(*switch_data_name)); +static void _case(Request& r, MethodParams& params) { + Switch_data* data=static_cast(r.classes_conf.get(switch_data_name)); if(!data) - throw Exception("parser.runtime", - &method_name, + throw Exception(PARSER_RUNTIME, + 0, "without switch"); - int count=params->size(); - Value *code=¶ms->as_junction(--count, "case result must be code"); - for(int i=0; iget(i)); + if(data->found) // matches already was found + return; + + int count=params.count(); + Value* code=¶ms.as_expression(--count, "case result must be code"); - if(value.as_string() == *case_default_value) { +#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 + + for(int i=0; i_default=code; - break; + continue; } bool matches; - if(data->searching->is_string()) - matches=data->searching->as_string() == value.as_string(); + if(data->searching_string) + matches=(*data->searching_string) == value.as_string(); + else if(data->searching_bool || value.is_bool()) + matches=(data->searching_double != 0) == value.as_bool(); else - matches=data->searching->as_double() == value.as_double(); + matches=data->searching_double == value.as_double(); - if(matches) { + if(matches){ data->found=code; break; } } } +#ifndef DOXYGEN +struct Try_catch_result { + ValueRef processed_code; + const String* exception_should_be_handled; + + Try_catch_result(): exception_should_be_handled(0) {} +}; + +/// Auto-object used for temporary changing Request::skip. +class Temp_skip { + Request& frequest; + Request::Skip saved_skip; +public: + Temp_skip(Request& arequest) : frequest(arequest), saved_skip(arequest.get_skip()) { + arequest.set_skip(Request::SKIP_NOTHING); + } + ~Temp_skip() { + if(frequest.get_skip() == Request::SKIP_NOTHING) + frequest.set_skip(saved_skip); + } +}; +#endif + +/// used by ^try and ^cache, @returns $exception.handled[string] if any +template +static Try_catch_result try_catch(Request& r, Value& body_code(Request&, I), I info, Value* catch_code, bool could_be_handled_by_caller=false) { + Try_catch_result result; + + // minor bug: context not restored if only finally code is present, see #1062 + if(!catch_code) { + result.processed_code=body_code(r, info); + return result; + } + + // taking snapshot of try-context + Request_context_saver try_context(r); + try { + result.processed_code=body_code(r, info); + } catch(const Exception& e) { + Request_context_saver throw_context(r); // remembering exception stack trace + + Request::Exception_details details=r.get_details(e); + + try_context.restore(); // restoring try-context for code after try and catch-code + + { + Temp_value_element temp(r, *catch_code->get_junction()->method_frame, exception_var_name, &details.vhash); + Temp_skip temp_skip(r); + result.processed_code=r.process(*catch_code); + } + + // retriving $exception.handled + Value* vhandled=details.vhash.hash().get(exception_handled_part_name); + + bool bhandled=false; + if(vhandled) { + if(vhandled->is_string()) { // not simple $exception.handled(1/0)? + if(bhandled=could_be_handled_by_caller) { // and we can possibly handle it + result.exception_should_be_handled=vhandled->get_string(); // considering 'recovered' and let the caller recover + } + } else + bhandled=vhandled->as_bool(); + } + + if(!bhandled){ + throw_context.restore(); // restoring exception stack trace creared by try_context.restore() + rethrow; + } + } + + return result; +} + +static Value& process_try_body_code(Request& r, Value* body_code) { + return r.process(*body_code); +} // cache-- // consts -const int DATA_STRING_SERIALIZED_VERSION=0x0001; +const int DATA_STRING_SERIALIZED_VERSION=0x0006; // helper types #ifndef DOXYGEN struct Data_string_serialized_prolog { int version; + time_t expires; }; #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 +struct Cache_scope { +public: + time_t expires; + const String* body_from_disk; +}; struct Locked_process_and_cache_put_action_info { Request *r; - Value *body; + Cache_scope *scope; + Value* body_code; + Value* catch_code; + const String* processed_code; }; #endif + + + +static Value& process_cache_body_code(Request& r, Value* body_code) { + return r.process(*body_code); +} + +/* @todo maybe network order worth spending some effort? + don't bothering myself with network byte order, + am not planning to be able to move resulting file across platforms +*/ static void locked_process_and_cache_put_action(int f, void *context) { Locked_process_and_cache_put_action_info& info= *static_cast(context); - - // body->process - info.body=&info.r->process(*info.body); - - // result->string - const String& data_string=info.body->as_string(); - // string -serialize> buffer - void *data; size_t data_size; - data_string.serialize( - sizeof(Data_string_serialized_prolog), - data, data_size); - Data_string_serialized_prolog& prolog= - *static_cast(data); - prolog.version=DATA_STRING_SERIALIZED_VERSION; - - // buffer -write> file - write(f, data, data_size); -} -Value *locked_process_and_cache_put(Request& r, - Value& body_code, - const String& file_spec) { - Locked_process_and_cache_put_action_info info={ - &r, - &body_code, - }; + const String* body_from_disk=info.scope->body_from_disk; + // body->process + Try_catch_result result=try_catch(*info.r, + process_cache_body_code, info.body_code, + info.catch_code, body_from_disk!=0 /*we have something old=we can handle=recover later*/); + + if(result.exception_should_be_handled) { + if(*result.exception_should_be_handled==CACHE_EXCEPTION_HANDLED_CACHE_NAME) { + assert(body_from_disk); + info.processed_code=body_from_disk; + } else + throw Exception(PARSER_RUNTIME, + result.exception_should_be_handled, + "$" EXCEPTION_VAR_NAME "." EXCEPTION_HANDLED_PART_NAME " value must be " + "either boolean or string '" CACHE_EXCEPTION_HANDLED_CACHE_NAME "'"); + } else + info.processed_code=&((Value &)result.processed_code).as_string(); + + // expiration time not spoiled by ^cache(0) or something? + if(info.scope->expires > time(0)) { + // string -serialize> buffer + String::Cm serialized=info.processed_code->serialize( + sizeof(Data_string_serialized_prolog)); + Data_string_serialized_prolog& prolog= + *reinterpret_cast(serialized.str); + prolog.version=DATA_STRING_SERIALIZED_VERSION; + prolog.expires=info.scope->expires; + + // buffer -write> file + write(f, serialized.str, serialized.length); + } else // expired! + info.scope->expires=0; // flag it so that could be easily checked by caller +} +const String* locked_process_and_cache_put(Request& r, + Value& body_code, + Value* catch_code, + Cache_scope& scope, + const String& file_spec) +{ + Locked_process_and_cache_put_action_info info={&r, &scope, &body_code, catch_code, 0}; - return file_write_action_under_lock( + 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*/) ? info.body : 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); + return result; } -String *cache_get(Pool& pool, const String& file_spec) { - void* data; size_t data_size; - if(file_read(pool, file_spec, - data, data_size, + +#ifndef DOXYGEN +struct Cache_get_result { + const String* body; + bool expired; +}; +#endif + +static Cache_get_result cache_get(Request_charsets& charsets, const String& file_spec, time_t now) { + Cache_get_result result={0, false}; + + File_read_result file=file_read(charsets, file_spec, false/*as_text*/, - false/*fail_on_read_problem*/) - && data_size/* ignore reads which are empty due to + 0, //no params + false/*fail_on_read_problem*/); + if(file.success && file.length/* ignore reads which are empty due to non-unary open+lockEX conflict with lockSH */) { - + Data_string_serialized_prolog& prolog= - *static_cast(data); + *reinterpret_cast(file.str); - String *result=new(pool) String(pool); + String* body=new String; if( - data_size>=sizeof(Data_string_serialized_prolog) - && prolog.version==DATA_STRING_SERIALIZED_VERSION - && result->deserialize( - sizeof(Data_string_serialized_prolog), data, data_size, file_spec.cstr())) - return result; + file.length>=sizeof(Data_string_serialized_prolog) + && prolog.version==DATA_STRING_SERIALIZED_VERSION) { + if(body->deserialize(sizeof(Data_string_serialized_prolog), file.str, file.length)) { + result.body=body; + result.expired=prolog.expires <= now; + } + } } - return 0; + return result; } -static void _cache(Request& r, const String& method_name, MethodParams *params) { - Pool& pool=r.pool(); +static time_t as_expires(Request& r, MethodParams& params, + int index, time_t now) { + time_t result; + if(Value* vdate=params[index].as(VDATE_TYPE)) + result=(time_t)(static_cast(vdate)->get_time()); + else + result=now+(time_t)params.as_double(index, "lifespan must be date or number", r); - // file_spec, expires, body code - const String &file_spec=r.absolute(params->as_string(0, "filespec must be string")); - if(params->size()==1) { // delete - cache_delete(file_spec); + return result; +} + +static const String& as_file_spec(Request& r, MethodParams& params, int index) { + return r.absolute(params.as_string(index, "filespec must be string")); +} + +static void _cache(Request& r, MethodParams& params) { + 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, 0, "expire-time get without cache"); + r.write(*new VDate((pa_time_t)scope->expires)); return; } - time_t lifespan=(time_t)params->as_double(1, "lifespan must be number", r); - Value& body_code=params->as_junction(2, "body must be code"); + time_t now=time(0); - if(lifespan) { // 'lifespan' specified? try cached copy... - size_t size; - time_t atime, mtime, ctime; - - // 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; ... - - // {file_spec} modification time - for(int retry=0; retry<2; retry++) { - if(file_stat(file_spec, size, atime, mtime, ctime, false/*no exception on error*/)) // exists? - if(time(0)-mtime > lifespan) // expired - cache_delete(file_spec); - else // not expired - if(String *cached_body=cache_get(pool, file_spec)) { // have cached copy? - // write it out - r.write_assign_lang(*cached_body); - // happy with it - return; - } + 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; + } - // non-blocked lock; process; cache it - if(Value *processed_body=locked_process_and_cache_put(r, body_code, 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 - } + // secods|expires date + Cache_scope* scope=static_cast(r.classes_conf.get(cache_data_name)); + if(!scope) + throw Exception(PARSER_RUNTIME, + 0, + "expire-time reducing instruction without cache"); + + time_t expires=as_expires(r, params, 0, now); + if(expires < scope->expires) + scope->expires=expires; + + return; + } else if(params.count()<3) + throw Exception(PARSER_RUNTIME, + 0, + "invalid number of parameters"); + + // file_spec, expires, body code + const String& file_spec=as_file_spec(r, params, 0); + + Cache_scope scope={as_expires(r, params, 1, now), 0}; + + Temp_hash_value, void*> cache_scope_setter(&r.classes_conf, cache_data_name, &scope); + Value& body_code=params.as_junction(2, "body_code must be code"); + Value* catch_code=0; + if(params.count()>3) + catch_code=¶ms.as_junction(3, "catch_code must be code"); + + if(scope.expires>now) { + 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(*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(*processed_body); + // happy with it + return; + } else { + // we fail while get exclusive lock. nvm, we just execute body_code a bit later } - throw Exception(0, - &file_spec, - "locking problem"); } else { - // 'lifespan'=0, forget cached copy + // instructed not to cache; forget cached copy cache_delete(file_spec); - // process - Value& processed_body=r.process(body_code); - // write it out - r.write_assign_lang(processed_body); - // happy with it - return; } - // never reached + + // process without caching + if(catch_code){ + Try_catch_result result=try_catch(r, process_try_body_code, &body_code, catch_code); + r.write(result.processed_code); + } else { + r.write(r.process_to_string(body_code)); + } } -// also used in pa_request.C to pass param to @unhandled_exception -VHash& exception2vhash(Pool& pool, const Exception& e) { - VHash& result=*new(pool) VHash(pool); - Hash& hash=result.hash(0); - if(const char *type=e.type()) - hash.put(*exception_type_part_name, new(pool) VString(*new(pool) String(pool, type))); - if(const String *asource=e.problem_source()) { - String& source=*new(pool) String(pool); - source.append(*asource, String::UL_TAINTED, true/*forced*/); - result.set_name(source); - - hash.put(*exception_source_part_name, new(pool) VString(source)); -#ifndef NO_STRING_ORIGIN - const Origin& origin=source.origin(); - hash.put(*new(pool) String(pool, "file"), - new(pool) VString(*new(pool) String(pool, origin.file))); - hash.put(*new(pool) String(pool, "lineno"), - new(pool) VInt(pool, 1+origin.line)); -#endif - } - if(const char *ecomment=e.comment()) { - int comment_size=strlen(ecomment); - char *pcomment=(char *)pool.malloc(comment_size); - memcpy(pcomment, ecomment, comment_size); - hash.put(*exception_comment_part_name, - new(pool) VString(*new(pool) String(pool, pcomment, comment_size, true/*tainted*/))); - } - hash.put(*exception_handled_part_name, - new(pool) VBool(pool, false)); +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; - return result; -} + Try_catch_result result; -static void _try_operator(Request& r, const String& method_name, MethodParams *params) { - Pool& pool=r.pool(); + try{ + // process try and catch code + result=try_catch(r, process_try_body_code, &body_code, &catch_code); + } catch(...){ + // process finally code but ignore the result + if(finally_code){ + Temp_skip temp(r); + Value &finally_result=r.process(*finally_code); + } + rethrow; + } - Value& body_code=params->as_junction(0, "body_code must be code"); - Value& catch_code=params->as_junction(1, "catch_code must be code"); + // process finally code + if(finally_code){ + Temp_skip temp(r); + Value& finally_result=r.process(*finally_code); - Value *result; + // no exception in try/catch or finally, writing processed body_code or catch_code + r.write(result.processed_code); - // taking snapshot of request processing status - //int ssexception_trace=r.exception_trace.top_index(); - int sstack=r.stack.top_index(); - Value *sself=r.self, *sroot=r.root, *srcontext=r.rcontext; - WContext *swcontext=r.wcontext; - try { - result=&r.process(body_code); - } catch(const Exception& e) { - // restoring request processing status - //r.exception_trace.top_index(ssexception_trace); - r.stack.top_index(sstack); - r.self=sself; r.root=sroot, r.rcontext=srcontext; r.wcontext=swcontext; - - - VHash& vhash=exception2vhash(pool, e); + // write out processed finally code + r.write(finally_result); + } else { + // no exception in try/catch, writing processed body_code or catch_code + r.write(result.processed_code); + } + +} - Junction *junction=catch_code.get_junction(); - Value *saved_exception_var_value=junction->root->get_element(*exception_var_name); - junction->root->put_element(*exception_var_name, &vhash); - result=&r.process(catch_code); - bool handled=false; - if(Value *value=static_cast(vhash.hash(0).get(*exception_handled_part_name))) - handled=value->as_bool(); - junction->root->put_element(*exception_var_name, saved_exception_var_value); - - if(!handled) - throw(e); // rethrow - } - // write it out - r.write_pass_lang(*result); -} - -static void _throw_operator(Request& r, const String& method_name, MethodParams *params) { - Pool& pool=r.pool(); - - if(params->size()==1) { - Value& param0=params->get(0); - if(Hash *hash=param0.get_hash(&method_name)) { - const char *type=0; - if(Value *value=static_cast(hash->get(*exception_type_part_name))) +static void _throw_operator(Request&, MethodParams& params) { + 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)) type=value->as_string().cstr(); - const String *source=0; - if(Value *value=static_cast(hash->get(*exception_source_part_name))) + const String* source=0; + if(Value* value=hash->get(exception_source_part_name)) source=&value->as_string(); - const char *comment=0; - if(Value *value= - static_cast(hash->get(*exception_comment_part_name))) + const char* comment=0; + if(Value* value=hash->get(exception_comment_part_name)) comment=value->as_string().cstr(); throw Exception(type, - source?source:&method_name, - comment); + source?source:0, + "%s", comment?comment:""); } else - throw Exception("parser.runtime", - &method_name, - "one-param version has hash param"); + throw Exception(PARSER_RUNTIME, + 0, + "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->as_string(2, "comment must be string").cstr(); - throw Exception(type, &source, comment); - } + const char* type=params.as_string(0, "type must be string").cstr(); + 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:""); + } + } + +static void _sleep_operator(Request& r, MethodParams& params) { + double seconds=params.as_double(0, "seconds must be double", r); + if(seconds>0) + pa_sleep((int)trunc(seconds), (int)trunc((seconds-trunc(seconds))*1000000)); + } + +#if defined(WIN32) && defined(_DEBUG) +# define PA_BPT +static void _bpt(Request&, MethodParams&) { + _asm int 3; } - +#endif + // constructor -MOP::MOP(Pool& apool) : Methoded(apool), - main_sql_name(apool, MAIN_SQL_NAME), - main_sql_drivers_name(apool, MAIN_SQL_DRIVERS_NAME) -{ - set_name(*NEW String(pool(), OP_CLASS_NAME)); +VClassMAIN::VClassMAIN(): VClass(MAIN_CLASS_NAME) { + +#ifdef PA_BPT + // ^bpt[] + add_native_method("bpt", Method::CT_ANY, _bpt, 0, 0); +#endif // ^if(condition){code-when-true} // ^if(condition){code-when-true}{code-when-false} - add_native_method("if", Method::CT_ANY, _if, 2, 3); + // ^if(condition){code-when-true} (another condition){code-when-true} ... {code-when-false} + add_native_method("if", Method::CT_ANY, _if, 2, 10000, Method::CO_WITHOUT_FRAME); + + // ^untaint[as-is|uri|sql|js|html|html-typo|regex|parser-code]{code} + add_native_method("untaint", Method::CT_ANY, _untaint, 1, 2, Method::CO_WITHOUT_FRAME); - // ^untaint[as-is|uri|sql|js|html|html-typo]{code} - add_native_method("untaint", Method::CT_ANY, _untaint, 1, 2); + // ^taint[as-is|uri|sql|js|html|html-typo|regex|parser-code]{code} + add_native_method("taint", Method::CT_ANY, _taint, 1, 2, Method::CO_WITHOUT_FRAME); - // ^taint[as-is|uri|sql|js|html|html-typo]{code} - add_native_method("taint", Method::CT_ANY, _taint, 1, 2); + // ^apply-taint[untaint lang][string] + add_native_method("apply-taint", Method::CT_ANY, _apply_taint, 1, 2, Method::CO_WITHOUT_FRAME); - // ^process[code] - add_native_method("process", Method::CT_ANY, _process, 1, 1); + // ^process{code} + // ^process[context]{code}[options hash] + 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, 2); + add_native_method("while", Method::CT_ANY, _while, 2, 3, Method::CO_WITHOUT_FRAME); + + // ^use[file[;options hash]] + add_native_method("use", Method::CT_ANY, _use, 1, 2); - // ^use[file] - add_native_method("use", Method::CT_ANY, _use, 1, 1); + // ^break[] + // ^break(condition) + add_native_method("break", Method::CT_ANY, _break, 0, 1, Method::CO_WITHOUT_FRAME); + + // ^continue[] + // ^continue(condition) + add_native_method("continue", Method::CT_ANY, _continue, 0, 1, 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); // ^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); + // ^cache[file_spec](time){code}[{catch code}] time=0 no cache // ^cache[file_spec] delete cache - // ^cache[file_spec](time){code} time=0 no cache - add_native_method("cache", Method::CT_ANY, _cache, 1, 3); + // ^cache[] get current expiration time + add_native_method("cache", Method::CT_ANY, _cache, 0, 4); // 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); + add_native_method("sleep", Method::CT_ANY, _sleep_operator, 1, 1); } // constructor & configurator -Methoded *MOP_create(Pool& pool) { - return new(pool) MOP(pool); -} - -void MOP::configure_user(Request& r) { - Pool& pool=r.pool(); - - // $MAIN:SQL.drivers - if(Value *sql=r.main_class->get_element(main_sql_name)) - if(Value *element=sql->get_element(main_sql_drivers_name)) - if(Table *protocol2library=element->get_table()) - r.classes_conf.put(name(), protocol2library); +VStateless_class& VClassMAIN_create() { + return *new VClassMAIN; }