/** @file Parser: parser @b operators. Copyright (c) 2001-2003 ArtLebedev Group (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) */ static const char* IDENT_OP_C="$Date: 2003/03/18 11:54:55 $"; #include "classes.h" #include "pa_vmethod_frame.h" #include "pa_common.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" // limits #define MAX_LOOPS 10000 // defines #define CASE_DEFAULT_VALUE "DEFAULT" // class class VClassMAIN: public VClass { public: VClassMAIN(); override ~VClassMAIN() {} }; // defines for statics #define SWITCH_DATA_NAME "SWITCH-DATA" #define CACHE_DATA_NAME "CACHE-DATA" #define EXCEPTION_VAR_NAME "exception" #define EXCEPTION_TYPE_PART_NAME "type" #define EXCEPTION_SOURCE_PART_NAME "source" #define EXCEPTION_COMMENT_PART_NAME "comment" #define EXCEPTION_HANDLED_PART_NAME "handled" // statics //^switch ^case static StringPtr switch_data_name(new String(SWITCH_DATA_NAME)); //^cache static StringPtr cache_data_name(new String(CACHE_DATA_NAME)); static StringPtr exception_var_name(new String(EXCEPTION_VAR_NAME)); static StringPtr exception_type_part_name(new String(EXCEPTION_TYPE_PART_NAME)); static StringPtr exception_source_part_name(new String(EXCEPTION_SOURCE_PART_NAME)); static StringPtr exception_comment_part_name(new String(EXCEPTION_COMMENT_PART_NAME)); static StringPtr exception_handled_part_name(new String(EXCEPTION_HANDLED_PART_NAME)); // helpers class Untaint_lang_name2enum: public Hash { public: Untaint_lang_name2enum() { #define ULN(name, LANG) \ put(StringPtr(new String(name)), (value_type)String::UL_##LANG); ULN("as-is", AS_IS); ULN("optimized-as-is", AS_IS|String::UL_OPTIMIZE_BIT); ULN("file-spec", FILE_SPEC); ULN("http-header", HTTP_HEADER); ULN("mail-header", MAIL_HEADER); ULN("uri", URI); ULN("table", TABLE); ULN("sql", SQL); ULN("js", JS); ULN("xml", XML); ULN("optimized-xml", XML|String::UL_OPTIMIZE_BIT); ULN("html", HTML); ULN("optimized-html", HTML|String::UL_OPTIMIZE_BIT); #undef ULN } } untaint_lang_name2enum; // methods static void _if(Request& r, StringPtr /*method_name*/, MethodParams* params) { ValuePtr condition_code=params->as_junction(0, "condition must be expression"); bool condition=r.process_to_value(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->count()>2) r.write_pass_lang(r.process(params->as_junction(2, "'else' parameter must be code"))); } static void _untaint(Request& r, StringPtr method_name, MethodParams* params) { Pool& pool=r.pool(); String_UL lang; if(params->count()==1) lang=String::UL_AS_IS; // mark as simply 'tainted'. useful in html from sql else { StringPtr 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"); } { ValuePtr vbody=params->as_junction(params->count()-1, "body must 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 } } static void _taint(Request& r, StringPtr /*method_name*/, MethodParams* params) { Pool& pool=r.pool(); String_UL lang; if(params->count()==1) lang=String::UL_TAINTED; // mark as simply 'tainted'. useful in table:set else { StringPtr 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"); } { ValuePtr vbody=params->as_no_junction(params->count()-1, "body must not be code"); String result; result.append( *vbody->as_string(&pool), // process marking tainted with that lang lang, true); // force result language to specified r.write_pass_lang(result); } } static void _process(Request& r, StringPtr method_name, MethodParams* params) { Pool& pool=r.pool(); StringPtr source; // must be declared here to prevent premature destruction MethodPtr main_method; ValuePtr target_self=params->count()>1? params->as_no_junction(0, "target must not be code") :ValuePtr(&r.get_method_frame()->caller()->self()); { ValuePtr vjunction=params->as_junction(params->count()-1, "body must be code"); VStateless_class *target_class=target_self->get_last_derived_class(); if(!target_class) throw Exception("parser.runtime", method_name, "no target class"); // evaluate source to process source=r.process_to_string(vjunction); // 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(*target_class, main_method_name, MethodPtr(0)); // temporary zero @auto so it wouldn't be auto-called in Request::use_buf Temp_method temp_method_auto(*target_class, auto_method_name, MethodPtr(0)); // calculate pseudo file name of processed chars // would be something like "/some/file(4) process" char local_place[MAX_STRING]; #ifndef NO_STRING_ORIGIN const String_fragment::Origin& source_origin=source->origin(); const String_fragment::Origin& method_origin=method_name->origin(); size_t place_size; if(source_origin.file==method_origin.file) place_size=snprintf(local_place, MAX_STRING, "%s(%d) %s", // same file source_origin.file?source_origin.file:"unknown_file", 1+source_origin.line, method_name->cstr().get())+1; else // different files ^process{external__file_text__or__sql} place_size=snprintf(local_place, MAX_STRING, "%s", source_origin.file?source_origin.file:"unknown_file")+1; #else strncpy(local_place, method_name->cstr(), MAX_STRING-1); place[MAX_STRING-1]=0; size_t place_size=strlen(local_place)+1; #endif char *heap_place=pool.copy(local_place, place_size); // process source code, append processed methods to 'self' class // maybe-define new @main r.use_buf(*target_class, source->cstr(pool, String::UL_UNSPECIFIED, r.connection(StringPtr(0))), StringPtr(new String(heap_place, place_size, true /*tainted*/)), heap_place); // main_method main_method=target_class->get_method(main_method_name); } // after restoring current-request-lang // maybe-execute @main[] if(main_method) { // temporarily set method_frame's self to target_self Temp_method_frame_self tmfs(*r.get_method_frame(), target_self); // execute! r.execute(*main_method->parser_code); } } static void _rem(Request& r, StringPtr /*method_name*/, MethodParams* params) { params->as_junction(0, "body must be code"); } static void _while(Request& r, StringPtr method_name, MethodParams* params) { Pool& pool=r.pool(); ValuePtr vcondition=params->as_junction(0, "condition must be expression"); ValuePtr body=params->as_junction(1, "body must be code"); // 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_to_value(vcondition, /*0/*no name* /,*/ false/*don't intercept string*/)->as_bool(); if(!condition) // ...condition is true break; // write processed body r.write_pass_lang(r.process(body)); } } static void _use(Request& r, StringPtr method_name, MethodParams* params) { Pool& pool=r.pool(); ValuePtr vfile=params->as_no_junction(0, "file name must not be code"); r.use_file(*r.main_class, vfile->as_string(&pool)); } static void _for(Request& r, StringPtr method_name, MethodParams* params) { Pool& pool=r.pool(); StringPtr 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); ValuePtr body_code=params->as_junction(3, "body must be code"); ValuePtr delim_maybe_code=params->count()>4?(*params)[4]:ValuePtr(0); if(to-from>=MAX_LOOPS) // too long loop? throw Exception("parser.runtime", method_name, "endless loop detected"); bool need_delim=false; VIntPtr vint(new VInt(0)); r.get_method_frame()->caller()->put_element(var_name, vint, false); for(int i=from; i<=to; i++) { vint->set_int(i); StringOrValue sv_processed=r.process(body_code); StringPtr s_processed=sv_processed.get_string(); if(delim_maybe_code && s_processed && s_processed->size()) { // 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; } r.write_pass_lang(sv_processed); } } static void _eval(Request& r, StringPtr method_name, MethodParams* params) { ValuePtr expr=params->as_junction(0, "need expression"); // evaluate expresion ValuePtr value_result=r.process_to_value(expr, /*0/*no name YET* /,*/ true/*don't intercept string*/)->as_expr_result(); if(params->count()>1) { ValuePtr fmt=params->as_no_junction(1, "fmt must not be code"); Pool& pool=r.pool(); String string; string.APPEND_CONST( format(pool, value_result->as_double(), fmt->as_string(&pool)->cstr())); r.write_no_lang(string); } else r.write_no_lang(value_result); } static void _connect(Request& r, StringPtr method_name, MethodParams* params) { Pool& pool=r.pool(); #ifdef RESOURCES_DEBUG struct timeval mt[2]; #endif ValuePtr url=params->as_no_junction(0, "url must not be code"); ValuePtr body_code=params->as_junction(1, "body must be code"); Table* protocol2driver_and_client=0; if(ValuePtr sql=r.main_class->get_element(StringPtr(new String(MAIN_SQL_NAME)), *r.main_class, false)) { if(ValuePtr element=sql->get_element(StringPtr(new String(MAIN_SQL_DRIVERS_NAME)), *sql, false)) { protocol2driver_and_client=element->get_table(); } } #ifdef RESOURCES_DEBUG //measure:before gettimeofday(&mt[0],NULL); #endif // connect SQL_ConnectionPtr connection=SQL_driver_manager.get_connection(pool, url->as_string(&pool), method_name, protocol2driver_and_client); #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; 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)); } catch(...) { // process problem connection->mark_to_rollback(); rethrow; } } #ifndef DOXYGEN class Switch_data: public PA_Object { public: Request& r; ValuePtr searching; ValuePtr found; ValuePtr _default; public: Switch_data(Request& ar, ValuePtr asearching): r(ar), searching(asearching) {} }; DECLARE_OBJECT_PTR(Switch_data); #endif static void _switch(Request& r, StringPtr /*method_name*/, MethodParams* params) { Switch_dataPtr data(new Switch_data(r, r.process_to_value(params->get(0)))); Temp_hash_value switch_data_setter(r.classes_conf, switch_data_name, data); ValuePtr 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, true/*intercept_string*/); if(ValuePtr selected_code=data->found? data->found: data->_default) { // setting code context, would execute in ^switch[...]{>>context<<} //selected_code->get_junction()->change_context(cases_code.get_junction()); r.write_pass_lang(r.process(selected_code)); } } static void _case(Request& r, StringPtr method_name, MethodParams* params) { Pool& pool=r.pool(); Switch_data* data=static_cast(r.classes_conf.get(switch_data_name).get()); if(!data) throw Exception("parser.runtime", method_name, "without switch"); int count=params->count(); ValuePtr code=params->as_junction(--count, "case result must be code"); // killing context for safety, would execute in ^switch[...]{>>context<<} // reason: context is stacked, and it would become invalid afterwards //code->get_junction()->change_context(0); for(int i=0; iget(i)); if(*value->as_string(&pool) == CASE_DEFAULT_VALUE) { data->_default=code; break; } bool matches; if(data->searching->is_string()) matches=*data->searching->as_string(&pool) == *value->as_string(&pool); else matches=data->searching->as_double() == value->as_double(); if(matches) { if(data->found) throw Exception("parser.runtime", method_name, "duplicate found"); data->found=code; break; } } } // cache-- // consts const int DATA_STRING_SERIALIZED_VERSION=0x0003; // helper types #ifndef DOXYGEN struct Data_string_serialized_prolog { int version; time_t expires; }; #endif void cache_delete(StringPtr file_spec) { file_delete(file_spec, false/*fail_on_read_problem*/); } #ifndef DOXYGEN class Cache_data: public PA_Object { public: time_t expires; }; DECLARE_OBJECT_PTR(Cache_data); struct Locked_process_and_cache_put_action_info { Request *r; Cache_data *data; ValuePtr body_code; StringPtr evaluated_body; }; #endif /* @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); Pool& pool=info.r->pool(); // body->process info.evaluated_body=info.r->process_to_string(info.body_code); // expiration time not spoiled by ^cache(0) or something? if(info.data->expires > time(0)) { // string -serialize> buffer char *data; size_t data_size; info.evaluated_body->serialize(pool, sizeof(Data_string_serialized_prolog), data, data_size); Data_string_serialized_prolog& prolog= *reinterpret_cast(data); prolog.version=DATA_STRING_SERIALIZED_VERSION; prolog.expires=info.data->expires; // buffer -write> file write(f, data, data_size); } else // expired! info.data->expires=0; // flag it so that could be easily checked by caller } StringPtr locked_process_and_cache_put(Request& r, ValuePtr body_code, Cache_data& data, StringPtr file_spec) { Locked_process_and_cache_put_action_info info; info.r=&r; info.data=&data; info.body_code=body_code; StringPtr result=file_write_action_under_lock( file_spec, "cache_put", locked_process_and_cache_put_action, &info, false/*as_text*/, false/*do_append*/, false/*block*/, false/*fail on lock problem*/) ? info.evaluated_body: StringPtr(0); time_t now=time(0); if(data.expires<=now) cache_delete(file_spec); return result; } StringPtr cache_get(Pool& pool, Charset& charset, StringPtr file_spec, time_t now) { File_read_result file=file_read(pool, charset, file_spec, false/*as_text*/, 0, //no params false/*fail_on_read_problem*/); if(file.success && file.size/* ignore reads which are empty due to non-unary open+lockEX conflict with lockSH */) { Data_string_serialized_prolog& prolog= *reinterpret_cast(file.data); StringPtr result(new String); if( file.size>=sizeof(Data_string_serialized_prolog) && prolog.version==DATA_STRING_SERIALIZED_VERSION && prolog.expires > now && result->deserialize( sizeof(Data_string_serialized_prolog), file.data, file.size, file_spec->cstr(pool))) return result; } return StringPtr(0); } static time_t as_expires(Request& r, StringPtr method_name, MethodParams* params, int index, time_t now) { time_t result; if(Value* vdate=(*params)[index]->as(VDATE_TYPE, false)) result=static_cast(vdate)->get_time(); else result=now+(time_t)params->as_double(index, "lifespan must be date or number", r); return result; } static StringPtr 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, StringPtr method_name, MethodParams* params) { Pool& pool=r.pool(); time_t now=time(0); // ^cache[filename] ^cache(seconds) ^cache[expires date] if(params->count()==1) { if((*params)[0]->is_string()) { // filename? cache_delete(as_file_spec(r, params, 0)); return; } // secods|expires date Cache_data* data=static_cast(r.classes_conf.get(cache_data_name).get()); if(!data) throw Exception("parser.runtime", method_name, "expire-time reducing instruction without cache"); time_t expires=as_expires(r, method_name, params, 0, now); if(expires < data->expires) data->expires=expires; return; } // file_spec, expires, body code StringPtr file_spec=r.absolute(params->as_string(0, "filespec must be string")); Cache_dataPtr data(new Cache_data); Temp_hash_value cache_data_setter(r.classes_conf, cache_data_name, data); data->expires=as_expires(r, method_name, params, 1, now); ValuePtr body_code=params->as_junction(2, "body must be code"); if(data->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++) { if(StringPtr cached_body=cache_get(pool, r.charsets.source(), file_spec, now)) { // have cached copy? // write it out r.write_assign_lang(*cached_body); // happy with it return; } // non-blocked lock; process; cache it if(StringPtr processed_body= locked_process_and_cache_put(r, body_code, *data, 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"); } else { // instructed not to cache; forget cached copy cache_delete(file_spec); // process StringPtr processed_body=r.process_to_string(body_code); // write it out r.write_assign_lang(*processed_body); // happy with it return; } // never reached } // also used in pa_request.C to pass param to @unhandled_exception VHashPtr exception2vhash(Pool& pool, const Exception& e) { VHashPtr result(new VHash); HashStringValue& hash=result->hash(Exception::undefined_source); if(const char* type=e.type(true)) hash.put(exception_type_part_name, ValuePtr(new VString(StringPtr(new String(type))))); if(StringPtr asource=e.problem_source()) { StringPtr source(new String); source->append(*asource, String::UL_TAINTED, true/*forced*/); hash.put(exception_source_part_name, ValuePtr(new VString(source))); #ifndef NO_STRING_ORIGIN const String_fragment::Origin& origin=source->origin(); hash.put(StringPtr(new String("file", 0, true)), ValuePtr(new VString(StringPtr(new String(origin.file))))); hash.put(StringPtr(new String("lineno")), ValuePtr(new VInt(1+origin.line))); #endif } if(const char* ecomment=e.comment(true)) { char *pcomment=pool.copy(ecomment); hash.put(exception_comment_part_name, ValuePtr(new VString(StringPtr(new String(pcomment, 0, true/*tainted*/))))); } hash.put(exception_handled_part_name, ValuePtr(new VBool(false))); return result; } static void _try_operator(Request& r, StringPtr method_name, MethodParams* params) { Pool& pool=r.pool(); ValuePtr body_code=params->as_junction(0, "body_code must be code"); ValuePtr catch_code=params->as_junction(1, "catch_code must be code"); StringOrValue result; // taking snapshot of try-context Request_context_saver try_context(r); try { result=r.process(body_code); } catch(const Exception& e) { Request_context_saver throw_context(r); // taking snapshot of throw-context [stack trace contains error] try_context.restore(); // restoring try-context to perform catch-code VHashPtr vhash=exception2vhash(pool, e); JunctionPtr junction=catch_code->get_junction(); Value* method_frame=junction->method_frame; ValuePtr saved_exception_var_value=method_frame->get_element(exception_var_name, *method_frame, false); junction->method_frame->put_element(exception_var_name, vhash, false); result=r.process(catch_code); bool handled=false; if(ValuePtr value= vhash->hash(Exception::undefined_source).get(exception_handled_part_name)) handled=value->as_bool(); junction->method_frame->put_element(exception_var_name, saved_exception_var_value, false); if(!handled) { throw_context.restore(); // restoring throw-context [exception were not handled] throw(e); // rethrow } } // write out result r.write_pass_lang(result); } static void _throw_operator(Request& r, StringPtr method_name, MethodParams* params) { Pool& pool=r.pool(); if(params->count()==1) { ValuePtr param0=(*params)[0]; if(HashStringValue *hash=param0->get_hash(method_name)) { const char* type=0; if(ValuePtr value=hash->get(exception_type_part_name)) type=value->as_string(&pool)->cstr(pool); StringPtr source; if(ValuePtr value=hash->get(exception_source_part_name)) source=value->as_string(&pool); CharPtr comment; if(ValuePtr value=hash->get(exception_comment_part_name)) comment=value->as_string(&pool)->cstr(); throw Exception(type, source?source:method_name, "%s", comment?comment.get():""); } else throw Exception("parser.runtime", method_name, "one-param version has hash param"); } else { const char* type=params->as_string(0, "type must be string")->cstr(pool); StringPtr source=params->as_string(1, "source must be string"); CharPtr comment=params->count()>2? params->as_string(2, "comment must be string")->cstr() :CharPtr(0); throw Exception(type, source, "%s", comment?comment.get():""); } } #if defined(WIN32) && defined(_DEBUG) # define PA_BPT static void _bpt(Request& r, StringPtr method_name, MethodParams* params) { _asm int 3; } #endif // constructor VClassMAIN::VClassMAIN(): VClass() { set_name(StringPtr(new String(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); // ^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]{code} add_native_method("taint", Method::CT_ANY, _taint, 1, 2); // ^process[code] add_native_method("process", Method::CT_ANY, _process, 1, 2); // ^rem{code} add_native_method("rem", Method::CT_ANY, _rem, 1, 10000); // ^while(condition){code} add_native_method("while", Method::CT_ANY, _while, 2, 2); // ^use[file] add_native_method("use", Method::CT_ANY, _use, 1, 1); // ^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); // ^connect[protocol://user:pass@host[:port]/database]{code with ^sql-s} add_native_method("connect", Method::CT_ANY, _connect, 2, 2); // ^cache[file_spec] delete cache // ^cache[file_spec](time){code} time=0 no cache add_native_method("cache", Method::CT_ANY, _cache, 1, 3); // switch // ^switch[value]{cases} add_native_method("switch", Method::CT_ANY, _switch, 2, 2); // ^case[value]{code} add_native_method("case", Method::CT_ANY, _case, 2, 10000); // try-catch // ^try{code}{catch code} add_native_method("try", Method::CT_ANY, _try_operator, 2, 2); // ^throw[$exception hash] // ^throw[type;source;comment] add_native_method("throw", Method::CT_ANY, _throw_operator, 1, 3); } // constructor & configurator VStateless_classPtr VClassMAIN_create() { return VStateless_classPtr(new VClassMAIN); }