--- parser3/src/main/compile.y 2001/03/14 08:50:03 1.98 +++ parser3/src/main/compile.y 2021/11/09 20:55:21 1.293 @@ -1,34 +1,35 @@ -/* - Parser - Copyright (c) 2001 ArtLebedev Group (http://www.artlebedev.com) +%{ +/** @file + Parser: compiler(lexical parser and grammar). + + Copyright (c) 2001-2020 Art. Lebedev Studio (http://www.artlebedev.com) Author: Alexander Petrosyan (http://design.ru/paf) - $Id: compile.y,v 1.98 2001/03/14 08:50:03 paf Exp $ + */ -/* - TODO.parser4: - cache compiled code from request to request. to do that... - 1: make method definitions, @CLASS, @BASE, @USE instructions, +volatile const char * IDENT_COMPILE_Y = "$Id: compile.y,v 1.293 2021/11/09 20:55:21 moko Exp $"; + +/** + @todo parser4: + - cache compiled code from request to request. to do that... + -#: make method definitions, @CLASS, @BASE, @USE instructions, which would be executed afterwards, and actions now performed at compile time would be delayed to run time. - 2: make cache expiration on time and on disk-change of class source - 3: in apache use subpools for compiled class storage - 4: in iis make up specialized Pool object for that + -#: make cache expiration on time and on disk-change of class source + -#: in apache use subpools for compiled class storage + -#: in iis make up specialized Pool object for that */ -%{ -#define YYSTYPE Array/**/ * -#define YYPARSE_PARAM pc -#define YYLEX_PARAM pc +#define YYSTYPE ArrayOperation* #define YYDEBUG 1 -#define YYERROR_VERBOSE -#define yyerror(msg) real_yyerror((parse_control *)pc, msg) +#define YYERROR_VERBOSE 1 +#define yyerror(pc, msg) real_yyerror(pc, msg) #define YYPRINT(file, type, value) yyprint(file, type, value) +#define YYMALLOC pa_malloc +#define YYFREE pa_free -#include -#include -#include +// includes #include "compile_tools.h" #include "pa_value.h" @@ -36,38 +37,72 @@ #include "pa_vobject.h" #include "pa_vdouble.h" #include "pa_globals.h" +#include "pa_vmethod_frame.h" -#define SELF_ELEMENT_NAME "self" +// defines + +#define CLASS_NAME "CLASS" #define USE_CONTROL_METHOD_NAME "USE" +#define OPTIONS_CONTROL_METHOD_NAME "OPTIONS" + +// forwards -int real_yyerror(parse_control *pc, char *s); -static void yyprint(FILE *file, int type, YYSTYPE value); -int yylex(YYSTYPE *lvalp, void *pc); +static int real_yyerror(Parse_control* pc, const char* s); +static void yyprint(FILE* file, int type, YYSTYPE value); +static int yylex(YYSTYPE* lvalp, void* pc); +static const VBool vfalse(false); +static const VBool vtrue(true); +static const VString vempty; // local convinient inplace typecast & var -#define PC ((parse_control *)pc) -#define POOL *PC->pool -#undef NEW -#define NEW new(POOL) +#undef PC +#define PC (*pc) +#undef POOL +#define POOL (*PC.pool) +#ifndef DOXYGEN + +#define CLASS_ADD if(PC.class_add()){ \ + PC.error=pa_strcat(PC.cclass->type(), " - class is already defined"); \ + YYERROR; \ +} + +#define YYERROR3(header, value, footer) { \ + PC.error=pa_strcat(header, value, footer); \ + YYERROR; \ +} + +#define YYERROR1(value) { \ + PC.error=value; \ + YYERROR; \ +} + %} -%pure_parser +%pure-parser +%lex-param {Parse_control* pc} +%parse-param {Parse_control* pc} %token EON %token STRING %token BOGUS %token BAD_STRING_COMPARISON_OPERATOR +%token BAD_HEX_LITERAL +%token BAD_METHOD_DECL_START +%token BAD_METHOD_PARAMETER_NAME_CHARACTER %token LAND "&&" %token LOR "||" -%token LXOR "##" +%token LXOR "!||" +%token NXOR "!|" %token NLE "<=" %token NGE ">=" %token NEQ "==" %token NNE "!=" +%token NSL "<<" +%token NSR ">>" %token SLT "lt" %token SGT "gt" @@ -79,38 +114,40 @@ int yylex(YYSTYPE *lvalp, void *pc); %token DEF "def" %token IN "in" %token FEXISTS "-f" +%token DEXISTS "-d" %token IS "is" +%token LITERAL_TRUE "true" +%token LITERAL_FALSE "false" + /* logical */ -%left "is" -%left "lt" "gt" "le" "ge" -%left "eq" "ne" -%left '<' '>' "<=" ">=" "##" -%left "==" "!=" +%left "!||" %left "||" %left "&&" -%left "def" "in" "-f" -%left '!' +%left '<' '>' "<=" ">=" "lt" "gt" "le" "ge" +%left "==" "!=" "eq" "ne" +%left "is" "def" "in" "-f" "-d" /* bitwise */ -%left '#' -%left '&' '|' -%left '~' +%left "!|" +%left '|' +%left '&' +%left "<<" ">>" /* numerical */ -%left '-' '+' -%left '*' '/' '%' -%left NEG /* negation: unary - */ +%left '+' '-' +%left '*' '/' '\\' '%' +%left NUNARY /* unary - + */ + +/* out-of-group */ +%left '~' /* bitwise */ +%left '!' /* logical */ %% all: one_big_piece { - Method& method=*NEW Method(POOL, - *main_method_name, - 0, 0, /*min, max numbered_params_count*/ - 0/*param_names*/, 0/*local_names*/, - $1/*parser_code*/, 0/*native_code*/); - PC->cclass->add_method(*main_method_name, method); + Method* method=new Method(Method::CT_ANY, 0, 0 /*min, max numbered_params_count*/, 0 /*param_names*/, 0 /*local_names*/, $1 /*parser_code*/, 0 /*native_code*/, PC.cclass->is_vars_local()); + PC.cclass->set_method(PC.alias_method(main_method_name), method); } | methods; @@ -120,110 +157,132 @@ one_big_piece: maybe_codes; method: control_method | code_method; control_method: '@' STRING '\n' - control_strings { - const String& command=*SLA2S($2); + maybe_control_strings { + const String& command=LA2S(*$2)->trim(String::TRIM_END); YYSTYPE strings_code=$4; - if(strings_code->size()<1*2) { - strcpy(PC->error, "@"); - strcat(PC->error, command.cstr()); - strcat(PC->error, " is empty"); - YYERROR; - } + if(strings_code->count()<1*OPERATIONS_PER_OPVALUE) + YYERROR3("@", command.cstr(), " is empty"); if(command==CLASS_NAME) { - if(PC->cclass!=&PC->request->root_class) { // already changed from default? - strcpy(PC->error, "class already have a name '"); - strncat(PC->error, PC->cclass->name().cstr(), 100); - strcat(PC->error, "'"); - YYERROR; - } - if(strings_code->size()==1*2) { + if(strings_code->count()==1*OPERATIONS_PER_OPVALUE) { + CLASS_ADD; // new class' name - const String *name=SLA2S(strings_code); + const String& name=LA2S(*strings_code)->trim(String::TRIM_END); // creating the class - PC->cclass=NEW VClass(POOL); - PC->cclass->set_name(*name); - // defaulting base. may change with @BASE - PC->cclass->set_base(PC->request->root_class); - // append to request's classes - PC->request->classes().put(*name, PC->cclass); + VStateless_class* cclass=new VClass(name.cstr(), PC.request.get_used_filespec(PC.file_no)); + PC.cclass_new=cclass; + PC.append=false; } else { - strcpy(PC->error, "@"CLASS_NAME" must contain sole name"); - YYERROR; + YYERROR1("@" CLASS_NAME " must contain only one line with class name (contains more then one)"); } - } else { - if(command==USE_CONTROL_METHOD_NAME) { - for(int i=0; isize(); i+=2) { - String file(*SLA2S(strings_code, i)); - file.APPEND_CONST(".p"); - PC->request->use_file(file.cstr()); - } - } else if(command==BASE_NAME) { - if(PC->cclass->base()!=&PC->request->root_class) { // already changed from default? - strcpy(PC->error, "class already have a base '"); - strncat(PC->error, PC->cclass->base()->name().cstr(), 100); - strcat(PC->error, "'"); - YYERROR; - } - if(strings_code->size()==1*2) { - // TODO: преодолеть self и циклические base - const String& base_name=*SLA2S(strings_code); - VClass *base=static_cast( - PC->request->classes().get(base_name)); - if(!base) { - strcpy(PC->error, base_name.cstr()); - strcat(PC->error, ": undefined class in @"BASE_NAME); - YYERROR; - } - PC->cclass->set_base(*base); + } else if(command==USE_CONTROL_METHOD_NAME) { + CLASS_ADD; + for(size_t i=0; icount(); i+=OPERATIONS_PER_OPVALUE){ + PC.request.use_file(LA2S(*strings_code, i)->trim(String::TRIM_END), PC.request.get_used_filespec(PC.file_no), strings_code->get(i+1).origin); + } + } else if(command==BASE_NAME) { + if(PC.append) + YYERROR3("can't set base while appending methods to class '", PC.cclass->type(), "'"); + CLASS_ADD; + if(PC.cclass->base_class()) // already changed from default? + YYERROR3("class already have a base '", PC.cclass->base_class()->type(), "'"); + if(strings_code->count()==1*OPERATIONS_PER_OPVALUE) { + const String& base_name=LA2S(*strings_code)->trim(String::TRIM_END); + if(VStateless_class *base_class=PC.request.get_class(base_name)) { + // @CLASS == @BASE sanity check + if(PC.cclass==base_class) + YYERROR1("@" CLASS_NAME " equals @" BASE_NAME); + PC.cclass->get_class()->set_base(base_class); } else { - strcpy(PC->error, "@"BASE_NAME" must contain sole name"); - YYERROR; + YYERROR3("'", base_name.cstr(), "': undefined class in @" BASE_NAME); } } else { - strcpy(PC->error, "'"); - strncat(PC->error, command.cstr(), MAX_STRING/2); - strcat(PC->error, "' invalid special name. valid names are " - "'"CLASS_NAME"', '"USE_CONTROL_METHOD_NAME"' and '"BASE_NAME"'"); - YYERROR; + YYERROR1("@" BASE_NAME " must contain sole name"); + } + } else if(command==OPTIONS_CONTROL_METHOD_NAME) { + for(size_t i=0; icount(); i+=OPERATIONS_PER_OPVALUE) { + const String& option=LA2S(*strings_code, i)->trim(String::TRIM_END); + if(option==Symbols::LOCALS_SYMBOL){ + PC.set_all_vars_local(); + } else if(option==Symbols::PARTIAL_SYMBOL){ + if(PC.cclass_new){ + if(VStateless_class* existed=PC.get_existed_class(PC.cclass_new)){ + if(!PC.reuse_existed_class(existed)) + YYERROR3("can't append methods to '", PC.cclass_new->type(), "' - the class wasn't marked as partial"); + } else { + // marks the new class as partial. we will be able to add methods here later. + PC.cclass_new->set_partial(); + } + } else { + YYERROR1("'partial' option should be used straight after @" CLASS_NAME); + } + } else if(option==Symbols::STATIC_SYMBOL){ + PC.set_methods_call_type(Method::CT_STATIC); + } else if(option==Symbols::DYNAMIC_SYMBOL){ + PC.set_methods_call_type(Method::CT_DYNAMIC); + } else { + YYERROR3("'", option.cstr(), "' invalid option. valid options are 'partial', 'locals', 'static' and 'dynamic'"); + } } + } else { + YYERROR3("'", command.cstr(), "' invalid special name. valid names are '" CLASS_NAME "', '" USE_CONTROL_METHOD_NAME "', '" BASE_NAME "' and '" OPTIONS_CONTROL_METHOD_NAME "'."); } }; -control_strings: control_string | control_strings control_string { $$=$1; P($$, $2) }; +maybe_control_strings: empty | control_strings; +control_strings: control_string | control_strings control_string { $$=$1; P(*$$, *$2); }; control_string: maybe_string '\n'; maybe_string: empty | STRING; -code_method: '@' STRING bracketed_maybe_strings maybe_bracketed_strings maybe_comment '\n' - maybe_codes { - const String *name=SLA2S($2); +code_method: '@' STRING bracketed_maybe_strings maybe_bracketed_strings maybe_comment '\n' { + CLASS_ADD; + PC.explicit_result=false; YYSTYPE params_names_code=$3; - Array *params_names=0; - if(int size=params_names_code->size()) { - params_names=NEW Array(POOL); - for(int i=0; icount()) { + params_names=new ArrayString; + for(int i=0; isize()) { - locals_names=NEW Array(POOL); - for(int i=0; iis_vars_local(); + + if(int size=locals_names_code->count()) { + locals_names=new ArrayString; + for(int i=0; icclass->add_method(*name, method); + 0/*to be filled later in next {} */, 0, all_vars_local); + + *reinterpret_cast(&$$)=method; +} maybe_codes { + Method* method=reinterpret_cast($7); + // fill in the code + method->parser_code=$8; + + // register in class + const String& name=*LA2S(*$2); + PC.cclass->set_method(PC.alias_method(name), method); }; maybe_bracketed_strings: empty | bracketed_maybe_strings; -bracketed_maybe_strings: '[' maybe_strings ']' {$$=$2}; +bracketed_maybe_strings: '[' maybe_strings ']' {$$=$2;}; maybe_strings: empty | strings; -strings: STRING | strings ';' STRING { $$=$1; P($$, $3) }; +strings: STRING | strings ';' STRING { $$=$1; P(*$$, *$3); }; maybe_comment: empty | STRING; @@ -231,295 +290,520 @@ maybe_comment: empty | STRING; maybe_codes: empty | codes; -codes: code | codes code { $$=$1; P($$, $2) }; +codes: code | codes code { $$=$1; P(*$$, *$2); }; code: write_string | action; -action: get | put | with | call; +action: get | put | call; /* get */ get: get_value { - $$=$1; /* stack: resulting value */ - O($$, OP_WRITE); /* value=pop; wcontext.write(value) */ + $$=N(); + YYSTYPE code=$1; + size_t count=code->count(); + +#ifdef OPTIMIZE_BYTECODE_GET_ELEMENT + if(count!=3 || !change_first(*code, OP::OP_VALUE__GET_ELEMENT, /*=>*/OP::OP_VALUE__GET_ELEMENT__WRITE) ) +#endif + +#ifdef OPTIMIZE_BYTECODE_GET_SELF_ELEMENT + if(count!=3 || !change_first(*code, OP::OP_WITH_SELF__VALUE__GET_ELEMENT, /*=>*/OP::OP_WITH_SELF__VALUE__GET_ELEMENT__WRITE) ) +#endif + +#ifdef OPTIMIZE_BYTECODE_GET_OBJECT_ELEMENT + if(count!=5 || !change_first(*code, OP::OP_GET_OBJECT_ELEMENT, /*=>*/OP::OP_GET_OBJECT_ELEMENT__WRITE) ) +#endif + +#ifdef OPTIMIZE_BYTECODE_GET_OBJECT_VAR_ELEMENT + if(count!=5 || !change_first(*code, OP::OP_GET_OBJECT_VAR_ELEMENT, /*=>*/OP::OP_GET_OBJECT_VAR_ELEMENT__WRITE) ) +#endif + +#ifdef OPTIMIZE_BYTECODE_GET_ELEMENT__SPECIAL + if(!change(*code, count-1/* last */, OP::OP_GET_ELEMENT__SPECIAL, /*=>*/OP::OP_GET_ELEMENT__SPECIAL__WRITE) ) +#endif + + { + change_or_append(*code, count-1 /* last */, OP::OP_GET_ELEMENT, /*=>*/OP::OP_GET_ELEMENT__WRITE, /*or */OP::OP_WRITE_VALUE ); /* value=pop; wcontext.write(value) */ + } + + P(*$$, *code); }; -get_value: '$' get_name_value { $$=$2 } +get_value: '$' get_name_value { $$=$2; }; get_name_value: name_without_curly_rdive EON | name_in_curly_rdive; -name_in_curly_rdive: '{' name_without_curly_rdive '}' { $$=$2 }; +name_in_curly_rdive: '{' name_without_curly_rdive '}' { $$=$2; }; name_without_curly_rdive: name_without_curly_rdive_read -| name_without_curly_rdive_root | name_without_curly_rdive_class; name_without_curly_rdive_read: name_without_curly_rdive_code { - $$=N(POOL); - Array *diving_code=$1; - const String *first_name=SLA2S(diving_code); - if(first_name && *first_name==SELF_ELEMENT_NAME) { - O($$, OP_WITH_SELF); /* stack: starting context */ - P($$, diving_code, - /* skip over... */ - diving_code->size()>2?3/*OP_+string+get_element*/:2/*OP_+string*/); + $$=N(); + YYSTYPE diving_code=$1; + size_t count=diving_code->count(); + + if(maybe_make_self(*$$, *diving_code, count)) { + // $self. + } else + +#ifdef OPTIMIZE_BYTECODE_GET_OBJECT_ELEMENT + if(maybe_make_get_object_element(*$$, *diving_code, count)){ + // optimization for $object.field + ^object.method[ + } else +#endif + +#ifdef OPTIMIZE_BYTECODE_GET_OBJECT_VAR_ELEMENT + if(maybe_make_get_object_var_element(*$$, *diving_code, count)){ + // optimization for $object.$var + } else +#endif + +#ifdef OPTIMIZE_BYTECODE_GET_ELEMENT + if(count>=4 && (*diving_code)[0].code==OP::OP_VALUE && (*diving_code)[3].code==OP::OP_GET_ELEMENT ){ + // optimization + O(*$$, + (PC.in_call_value && count==4) + ? OP::OP_VALUE__GET_ELEMENT_OR_OPERATOR // ^object[ : OP_VALUE+origin+string+OP_GET_ELEMENT => OP_VALUE__GET_ELEMENT_OR_OPERATOR+origin+string + : OP::OP_VALUE__GET_ELEMENT // $object : OP_VALUE+origin+string+OP_GET_ELEMENT => OP_VALUE__GET_ELEMENT+origin+string + ); + P(*$$, *diving_code, 1/*offset*/, 2/*limit*/); // copy origin+value + if(count>4) + P(*$$, *diving_code, 4); // copy tail } else { - O($$, OP_WITH_READ); /* stack: starting context */ - P($$, diving_code); + O(*$$, OP::OP_WITH_READ); /* stack: starting context */ + P(*$$, *diving_code); + } +#else + { + O(*$$, OP::OP_WITH_READ); /* stack: starting context */ + + // ^if OP_ELEMENT => ^if OP_ELEMENT_OR_OPERATOR + // optimized OP_VALUE+origin+string+OP_GET_ELEMENT. => OP_VALUE+origin+string+OP_GET_ELEMENT_OR_OPERATOR. + if(PC.in_call_value && count==4) + diving_code->put(count-1, OP::OP_GET_ELEMENT_OR_OPERATOR); + P(*$$, *diving_code); } +#endif /* diving code; stack: current context */ }; -name_without_curly_rdive_root: ':' name_without_curly_rdive_code { - $$=N(POOL); - O($$, OP_WITH_ROOT); /* stack: starting context */ - P($$, $2); /* diving code; stack: current context */ -}; -name_without_curly_rdive_class: class_prefix name_without_curly_rdive_code { $$=$1; P($$, $2) }; -name_without_curly_rdive_code: name_advance2 | name_path name_advance2 { $$=$1; P($$, $2) }; +name_without_curly_rdive_class: class_prefix name_without_curly_rdive_code { $$=$1; P(*$$, *$2); }; +name_without_curly_rdive_code: name_advance2 | name_path name_advance2 { $$=$1; P(*$$, *$2); }; /* put */ put: '$' name_expr_wdive construct { - $$=$2; /* stack: context,name */ - P($$, $3); /* stack: context,name,constructor_value */ + $$=N(); +#ifdef OPTIMIZE_BYTECODE_CONSTRUCT + if(maybe_optimize_construct(*$$, *$2, *$3)){ + // $a(expr), $.a(expr), $a[value], $.a[value], $self.a[value], $self.a(expr) + } else +#endif + { + P(*$$, *$2); /* stack: context,name */ + P(*$$, *$3); /* stack: context,name,constructor_value */ + } }; name_expr_wdive: - name_expr_wdive_write -| name_expr_wdive_root + name_expr_wdive_root +| name_expr_wdive_write | name_expr_wdive_class; -name_expr_wdive_write: name_expr_dive_code { - $$=N(POOL); - Array *diving_code=$1; - const String *first_name=SLA2S(diving_code); - if(first_name && *first_name==SELF_ELEMENT_NAME) { - O($$, OP_WITH_SELF); /* stack: starting context */ - P($$, diving_code, - /* skip over... */ - diving_code->size()>2?3/*OP_+string+get_element*/:2/*OP_+string*/); - } else { - O($$, OP_WITH_WRITE); /* stack: starting context */ - P($$, diving_code); +name_expr_wdive_root: name_expr_dive_code { + $$=N(); + YYSTYPE diving_code=$1; + size_t count=diving_code->count(); + + if(maybe_make_self(*$$, *diving_code, count)) { + // $self. + } else +#ifdef OPTIMIZE_BYTECODE_GET_ELEMENT + if(count>=4 && (*diving_code)[0].code==OP::OP_VALUE && (*diving_code)[3].code==OP::OP_GET_ELEMENT ){ + O(*$$, OP::OP_WITH_ROOT__VALUE__GET_ELEMENT); + P(*$$, *diving_code, 1/*offset*/, 2/*limit*/); // copy origin+value + if(count>4) + P(*$$, *diving_code, 4); // tail + } else +#endif + { + O(*$$, OP::OP_WITH_ROOT); /* stack: starting context */ + P(*$$, *diving_code); } /* diving code; stack: current context */ }; -name_expr_wdive_root: ':' name_expr_dive_code { - $$=N(POOL); - O($$, OP_WITH_ROOT); /* stack: starting context */ - P($$, $2); /* diving code; stack: context,name */ -}; -name_expr_wdive_class: class_prefix name_expr_dive_code { $$=$1; P($$, $2) }; - -construct: construct_by_code | construct_by_expr; -construct_by_code: '[' any_constructor_code_value ']' { - $$=$2; /* stack: context, name, value */ - O($$, OP_CONSTRUCT_VALUE); /* value=pop; name=pop; context=pop; construct(context,name,value) */ +name_expr_wdive_write: '.' name_expr_dive_code { + $$=N(); + O(*$$, OP::OP_WITH_WRITE); /* stack: starting context */ + P(*$$, *$2); /* diving code; stack: context,name */ +}; +name_expr_wdive_class: class_prefix name_expr_dive_code { $$=$1; P(*$$, *$2); }; + +construct: + construct_square +| construct_round +| construct_curly +; +construct_square: '[' { + // allow $result_or_other_variable[ letters here any time ] + *reinterpret_cast(&$$)=PC.explicit_result; PC.explicit_result=false; +} any_constructor_code_value { + PC.explicit_result=*reinterpret_cast(&$2); +} ']' { + // stack: context, name + $$=$3; // stack: context, name, value + O(*$$, OP::OP_CONSTRUCT_VALUE); /* value=pop; name=pop; context=pop; construct(context,name,value) */ } ; -construct_by_expr: '(' expr_value ')' { - $$=$2; /* stack: context, name, value */ - O($$, OP_CONSTRUCT_EXPR); /* value=pop; name=pop; context=pop; construct(context,name,value) */ +construct_round: '(' expr_value ')' { + $$=N(); + // stack: context, name + P(*$$, *$2); // stack: context, name, value + O(*$$, OP::OP_CONSTRUCT_EXPR); /* value=pop->as_expr_result; name=pop; context=pop; construct(context,name,value) */ } ; +construct_curly: '{' maybe_codes '}' { + // stack: context, name + $$=N(); + OA(*$$, OP::OP_CURLY_CODE__CONSTRUCT, $2); /* code=pop; name=pop; context=pop; construct(context,name,junction(code)) */ +}; + any_constructor_code_value: - empty_string_value /* optimized $var[] case */ + empty_value /* optimized $var[] case */ | STRING /* optimized $var[STRING] case */ | constructor_code_value /* $var[something complex] */ ; constructor_code_value: constructor_code { - $$=N(POOL); - O($$, OP_CREATE_EWPOOL); /* stack: empty write context */ - P($$, $1); /* some code that writes to that context */ - O($$, OP_REDUCE_EWPOOL); /* context=pop; stack: context.value() */ + $$=N(); + OA(*$$, OP::OP_OBJECT_POOL, $1); /* stack: empty write context */ + /* some code that writes to that context */ + /* context=pop; stack: context.value() */ }; constructor_code: codes__excluding_sole_str_literal; -codes__excluding_sole_str_literal: action | code codes { $$=$1; P($$, $2) }; +codes__excluding_sole_str_literal: action | code codes { $$=$1; P(*$$, *$2); }; /* call */ call: call_value { - $$=$1; /* stack: value */ - O($$, OP_WRITE); /* value=pop; wcontext.write(value) */ + size_t count=$1->count(); +#ifdef OPTIMIZE_BYTECODE_CUT_REM_OPERATOR + if(count) +#endif + { + $$=$1; /* stack: value */ + if(!change_first(*$$, OP::OP_CONSTRUCT_OBJECT, /*=>*/ OP::OP_CONSTRUCT_OBJECT__WRITE)) + change_or_append(*$$, count-2 /* second last */, OP::OP_CALL, /*=>*/ OP::OP_CALL__WRITE, /*or */ OP::OP_WRITE_VALUE); /* value=pop; wcontext.write(value) */ + } }; -call_value: '^' call_name store_params EON { /* ^field.$method{vasya} */ - $$=$2; /* with_xxx,diving code; stack: context,method_junction */ - O($$, OP_GET_METHOD_FRAME); /* stack: context,method_frame */ - P($$, $3); /* filling method_frame.store_params */ - O($$, OP_CALL); /* method_frame=pop; ncontext=pop; call(ncontext,method_frame) stack: value */ +call_value: '^' { + PC.in_call_value=true; +} +call_name { + PC.in_call_value=false; +} +store_params EON { /* ^field.$method{vasya} */ +#ifdef OPTIMIZE_BYTECODE_CUT_REM_OPERATOR +#ifdef OPTIMIZE_BYTECODE_GET_ELEMENT + const String* operator_name=LA2S(*$3, 0, OP::OP_VALUE__GET_ELEMENT_OR_OPERATOR); +#else + const String* operator_name=LA2S(*$3, 1); +#endif + if(operator_name && SYMBOLS_EQ(*operator_name,REM_SYMBOL)){ + $$=N(); + } else +#endif + { + YYSTYPE params_code=$5; + if(params_code->count()==3) { // probably [] case. [OP::OP_VALUE+origin+Void] + if(Value* value=LA2V(*params_code)) // it is OP_VALUE+origin+value? + if(const String * string=value->get_string()) + if(string->is_empty()) // value is empty string? + params_code=0; // ^zzz[] case. don't append lone empty param. + } + /* stack: context, method_junction */ + + YYSTYPE var_code=$3; + if( + var_code->count()==8 + && ( (*var_code)[0].code==OP::OP_VALUE__GET_CLASS || (*var_code)[0].code==OP::OP_VALUE__GET_BASE_CLASS ) + && (*var_code)[3].code==OP::OP_PREPARE_TO_CONSTRUCT_OBJECT + && (*var_code)[4].code==OP::OP_VALUE +#ifdef FEATURE_GET_ELEMENT4CALL + && (*var_code)[7].code==OP::OP_GET_ELEMENT4CALL +#else + && (*var_code)[7].code==OP::OP_GET_ELEMENT +#endif + ){ + $$=N(); + O(*$$, OP::OP_CONSTRUCT_OBJECT); + P(*$$, *var_code, 1/*offset*/, 2/*limit*/); // class name + P(*$$, *var_code, 5/*offset*/, 2/*limit*/); // constructor name + OA(*$$, params_code); + } else + { + $$=var_code; /* with_xxx,diving code; stack: context,method_junction */ + OA(*$$, OP::OP_CALL, params_code); // method_frame=make frame(pop junction); ncontext=pop; call(ncontext,method_frame) stack: value + } + } }; -call_name: name_without_curly_rdive; +call_name: name_without_curly_rdive { +#ifdef FEATURE_GET_ELEMENT4CALL + size_t count=$1->count(); + if(count){ + $$=$1; +#ifdef OPTIMIZE_BYTECODE_GET_OBJECT_ELEMENT + !(count==5 && change_first(*$$, OP::OP_GET_OBJECT_ELEMENT, OP::OP_GET_OBJECT_ELEMENT4CALL)) && +#endif +#ifdef OPTIMIZE_BYTECODE_GET_OBJECT_VAR_ELEMENT + !(count==5 && change_first(*$$, OP::OP_GET_OBJECT_VAR_ELEMENT, OP::OP_GET_OBJECT_VAR_ELEMENT4CALL)) && +#endif + !change(*$$, count-1, OP::OP_GET_ELEMENT, OP::OP_GET_ELEMENT4CALL); + } +#endif +}; -store_params: store_param | store_params store_param { $$=$1; P($$, $2) }; +store_params: store_param | store_params store_param { $$=$1; P(*$$, *$2); }; store_param: store_square_param | store_round_param | store_curly_param ; -store_square_param: '[' store_code_param_parts ']' {$$=$2}; -store_round_param: '(' store_expr_param_parts ')' {$$=$2}; -store_curly_param: '{' store_curly_param_parts '}' {$$=$2}; +store_square_param: '[' { + // allow ^call[ letters here any time ] + *reinterpret_cast(&$$)=PC.explicit_result; PC.explicit_result=false; +} store_code_param_parts { + PC.explicit_result=*reinterpret_cast(&$2); +} ']' {$$=$3;}; +store_round_param: '(' store_expr_param_parts ')' {$$=$2;}; +store_curly_param: '{' store_curly_param_parts '}' {$$=$2;}; store_code_param_parts: store_code_param_part -| store_code_param_parts ';' store_code_param_part { $$=$1; P($$, $3) } +| store_code_param_parts ';' store_code_param_part { $$=$1; P(*$$, *$3); } ; store_expr_param_parts: store_expr_param_part -| store_expr_param_parts ';' store_expr_param_part { $$=$1; P($$, $3) } +| store_expr_param_parts ';' store_expr_param_part { $$=$1; P(*$$, *$3); } ; store_curly_param_parts: store_curly_param_part -| store_curly_param_parts ';' store_curly_param_part { $$=$1; P($$, $3) } +| store_curly_param_parts ';' store_curly_param_part { $$=$1; P(*$$, *$3); } ; -store_code_param_part: - empty /* optimized [] case */ -| STRING { /* optimized [STRING] case */ +store_code_param_part: code_param_value { $$=$1; - O($$, OP_STORE_PARAM); -} -| constructor_code_value { /* [something complex] */ - $$=$1; - O($$, OP_STORE_PARAM); -} -; -store_expr_param_part: write_expr_value { - $$=N(POOL); - PCA($$, $1); }; -store_curly_param_part: maybe_codes { - $$=N(POOL); - PCA($$, $1); +store_expr_param_part: expr_value { + YYSTYPE expr_code=$1; + if(expr_code->count()==3 + && (*expr_code)[0].code==OP::OP_VALUE) { // optimizing (double/bool/incidently 'string' too) case. [OP::OP_VALUE+origin+Double]. no evaluating + $$=expr_code; + } else { + YYSTYPE code=N(); + P(*code, *expr_code); + O(*code, OP::OP_WRITE_EXPR_RESULT); + $$=N(); + OA(*$$, OP::OP_EXPR_CODE__STORE_PARAM, code); + } }; -write_expr_value: expr_value { - $$=$1; - O($$, OP_WRITE); +store_curly_param_part: maybe_codes { + $$=N(); + OA(*$$, OP::OP_CURLY_CODE__STORE_PARAM, $1); }; +code_param_value: + empty_value /* optimized [;...] case */ +| STRING /* optimized [STRING] case */ +| constructor_code_value /* [something complex] */ +; /* name */ -name_expr_dive_code: name_expr_value | name_path name_expr_value { $$=$1; P($$, $2) }; +name_expr_dive_code: name_expr_value | name_path name_expr_value { $$=$1; P(*$$, *$2); }; -name_path: name_step | name_path name_step { $$=$1; P($$, $2) }; +name_path: name_step | name_path name_step { $$=$1; P(*$$, *$2); }; name_step: name_advance1 '.'; name_advance1: name_expr_value { + // we know that name_advance1 not called from ^xxx context + // so we'll not check for operator call possibility as we do in name_advance2 + /* stack: context */ $$=$1; /* stack: context,name */ - O($$, OP_GET_ELEMENT); /* name=pop; context=pop; stack: context.get_element(name) */ +#ifdef OPTIMIZE_BYTECODE_GET_ELEMENT__SPECIAL + O(*$$, is_special_element(*$$) ? OP::OP_GET_ELEMENT__SPECIAL : OP::OP_GET_ELEMENT); +#else + O(*$$, OP::OP_GET_ELEMENT); /* name=pop; context=pop; stack: context.get_element(name) */ +#endif }; name_advance2: name_expr_value { /* stack: context */ $$=$1; /* stack: context,name */ - O($$, OP_GET_ELEMENT); /* name=pop; context=pop; stack: context.get_element(name) */ +#ifdef OPTIMIZE_BYTECODE_GET_ELEMENT__SPECIAL + O(*$$, is_special_element(*$$) ? OP::OP_GET_ELEMENT__SPECIAL : OP::OP_GET_ELEMENT); +#else + O(*$$, OP::OP_GET_ELEMENT); /* name=pop; context=pop; stack: context.get_element(name) */ +#endif } | STRING BOGUS ; name_expr_value: STRING /* subname_is_const */ | name_expr_subvar_value /* $subname_is_var_value */ -| name_expr_with_subvar_value /* xxx$part_of_subname_is_var_value[$...] */ +| name_expr_with_subvar_value /* xxx$part_of_subname_is_var_value */ +| name_square_code_value /* [codes] */ ; name_expr_subvar_value: '$' subvar_ref_name_rdive { $$=$2; - O($$, OP_GET_ELEMENT); + O(*$$, OP::OP_GET_ELEMENT); }; name_expr_with_subvar_value: STRING subvar_get_writes { - $$=N(POOL); - O($$, OP_CREATE_EWPOOL); - P($$, $1); - O($$, OP_WRITE); - P($$, $2); - O($$, OP_REDUCE_EWPOOL); -}; -subvar_ref_name_rdive: subvar_ref_name_rdive_read | subvar_ref_name_rdive_root; -subvar_ref_name_rdive_read: STRING { - $$=N(POOL); - O($$, OP_WITH_READ); - P($$, $1); -}; -subvar_ref_name_rdive_root: ':' STRING { - $$=N(POOL); - O($$, OP_WITH_ROOT); - P($$, $2); + YYSTYPE code; + { + change_string_literal_to_write_string_literal(*(code=$1)); + P(*code, *$2); + } + $$=N(); + OA(*$$, OP::OP_STRING_POOL, code); +}; +name_square_code_value: '[' { + // allow $result_or_other_variable[ letters here any time ] + *reinterpret_cast(&$$)=PC.explicit_result; PC.explicit_result=false; +} codes { + PC.explicit_result=*reinterpret_cast(&$2); +} ']' { + $$=N(); +#ifdef OPTIMIZE_BYTECODE_GET_ELEMENT__SPECIAL + if(!maybe_append_simple_diving_code(*$$, *$3)) +#endif + { + OA(*$$, OP::OP_OBJECT_POOL, $3); /* stack: empty write context */ + /* some code that writes to that context */ + /* context=pop; stack: context.value() */ + } +}; +subvar_ref_name_rdive: STRING { + $$=N(); + O(*$$, OP::OP_WITH_READ); + P(*$$, *$1); }; -subvar_get_writes: subvar__get_write | subvar_get_writes subvar__get_write { $$=$1; P($$, $2) }; +subvar_get_writes: subvar__get_write | subvar_get_writes subvar__get_write { $$=$1; P(*$$, *$2); }; subvar__get_write: '$' subvar_ref_name_rdive { $$=$2; - O($$, OP_GET_ELEMENT__WRITE); + O(*$$, OP::OP_GET_ELEMENT__WRITE); }; -class_prefix: STRING ':' { +class_prefix: + class_static_prefix +| class_constructor_prefix +; +class_static_prefix: STRING ':' { $$=$1; // stack: class name string - O($$, OP_GET_CLASS); + OP::OPCODE code = OP::OP_VALUE__GET_CLASS; + if(*LA2S(*$$) == BASE_NAME) { // pseudo BASE class + if(VStateless_class* base=PC.cclass->base_class()) { + change_string_literal_value(*$$, *new String(base->type())); + } else { + YYERROR1("no base class declared"); + } + code = OP::OP_VALUE__GET_BASE_CLASS; + } else { + // can't use get_class because it will call @autouse[] if the class wasn't loaded + VStateless_class* base=PC.request.classes().get(*LA2S(*$$)); + if(base && PC.cclass->derived_from(*base)) + code = OP::OP_VALUE__GET_BASE_CLASS; + } + // optimized OP_VALUE+origin+string+OP_GET_CLASS => OP_VALUE__GET_CLASS+origin+string + change_first(*$$, OP::OP_VALUE, code); }; - - -/* with */ - -with: '$' name_without_curly_rdive '{' codes '}' { - $$=$2; - O($$, OP_CREATE_RWPOOL); - P($$, $4); - O($$, OP_REDUCE_RWPOOL); - O($$, OP_WRITE); +class_constructor_prefix: class_static_prefix ':' { + $$=$1; + if(!PC.in_call_value) + YYERROR1(":: not allowed here"); + O(*$$, OP::OP_PREPARE_TO_CONSTRUCT_OBJECT); }; + /* expr */ -expr_value: expr { - if(($$=$1)->size()==2) // only one string literal in there? - change_string_literal_to_double_literal($$); // make that string literal Double -}; +expr_value: expr; expr: - STRING + double_or_STRING +| true_value +| false_value | get_value | call_value | '"' string_inside_quotes_value '"' { $$ = $2; } +| '\'' string_inside_quotes_value '\'' { $$ = $2; } | '(' expr ')' { $$ = $2; } /* stack: operand // stack: @operand */ -| '-' expr %prec NEG { $$=$2; O($$, OP_NEG) } -| '~' expr { $$=$2; O($$, OP_INV) } -| '!' expr { $$=$2; O($$, OP_NOT) } -| "def" expr { $$=$2; O($$, OP_DEF) } -| "in" expr { $$=$2; O($$, OP_IN) } -| "-f" expr { $$=$2; O($$, OP_FEXISTS) } +| '-' expr %prec NUNARY { $$=$2; O(*$$, OP::OP_NEG); } +| '+' expr %prec NUNARY { $$=$2; } +| '~' expr { $$=$2; O(*$$, OP::OP_INV); } +| '!' expr { $$=$2; O(*$$, OP::OP_NOT); } +| "def" expr { $$=$2; O(*$$, OP::OP_DEF); } +| "in" expr { $$=$2; O(*$$, OP::OP_IN); } +| "-f" expr { $$=$2; O(*$$, OP::OP_FEXISTS); } +| "-d" expr { $$=$2; O(*$$, OP::OP_DEXISTS); } /* stack: a,b // stack: a@b */ -| expr '-' expr { $$=$1; P($$, $3); O($$, OP_SUB) } -| expr '+' expr { $$=$1; P($$, $3); O($$, OP_ADD) } -| expr '*' expr { $$=$1; P($$, $3); O($$, OP_MUL) } -| expr '/' expr { $$=$1; P($$, $3); O($$, OP_DIV) } -| expr '%' expr { $$=$1; P($$, $3); O($$, OP_MOD) } -| expr '&' expr { $$=$1; P($$, $3); O($$, OP_BIN_AND) } -| expr '|' expr { $$=$1; P($$, $3); O($$, OP_BIN_OR) } -| expr '#' expr { $$=$1; P($$, $3); O($$, OP_BIN_XOR) } -| expr "&&" expr { $$=$1; P($$, $3); O($$, OP_LOG_AND) } -| expr "||" expr { $$=$1; P($$, $3); O($$, OP_LOG_OR) } -| expr "##" expr { $$=$1; P($$, $3); O($$, OP_LOG_XOR) } -| expr '<' expr { $$=$1; P($$, $3); O($$, OP_NUM_LT) } -| expr '>' expr { $$=$1; P($$, $3); O($$, OP_NUM_GT) } -| expr "<=" expr { $$=$1; P($$, $3); O($$, OP_NUM_LE) } -| expr ">=" expr { $$=$1; P($$, $3); O($$, OP_NUM_GE) } -| expr "==" expr { $$=$1; P($$, $3); O($$, OP_NUM_EQ) } -| expr "!=" expr { $$=$1; P($$, $3); O($$, OP_NUM_NE) } -| expr "lt" expr { $$=$1; P($$, $3); O($$, OP_STR_LT) } -| expr "gt" expr { $$=$1; P($$, $3); O($$, OP_STR_GT) } -| expr "le" expr { $$=$1; P($$, $3); O($$, OP_STR_LE) } -| expr "ge" expr { $$=$1; P($$, $3); O($$, OP_STR_GE) } -| expr "eq" expr { $$=$1; P($$, $3); O($$, OP_STR_EQ) } -| expr "ne" expr { $$=$1; P($$, $3); O($$, OP_STR_NE) } -| expr "is" expr { $$=$1; P($$, $3); O($$, OP_IS) } +| expr '-' expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_SUB); } +| expr '+' expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_ADD); } +| expr '*' expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_MUL); } +| expr '/' expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_DIV); } +| expr '%' expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_MOD); } +| expr '\\' expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_INTDIV); } +| expr "<<" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_BIN_SL); } +| expr ">>" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_BIN_SR); } +| expr '&' expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_BIN_AND); } +| expr '|' expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_BIN_OR); } +| expr "!|" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_BIN_XOR); } +| expr "&&" expr { $$=$1; OA(*$$, OP::OP_NESTED_CODE, $3); O(*$$, OP::OP_LOG_AND); } +| expr "||" expr { $$=$1; OA(*$$, OP::OP_NESTED_CODE, $3); O(*$$, OP::OP_LOG_OR); } +| expr "!||" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_LOG_XOR); } +| expr '<' expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_NUM_LT); } +| expr '>' expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_NUM_GT); } +| expr "<=" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_NUM_LE); } +| expr ">=" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_NUM_GE); } +| expr "==" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_NUM_EQ); } +| expr "!=" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_NUM_NE); } +| expr "lt" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_STR_LT); } +| expr "gt" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_STR_GT); } +| expr "le" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_STR_LE); } +| expr "ge" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_STR_GE); } +| expr "eq" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_STR_EQ); } +| expr "ne" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_STR_NE); } +| expr "is" expr { $$=$1; P(*$$, *$3); O(*$$, OP::OP_IS); } ; +double_or_STRING: STRING { + // optimized OP_STRING => OP_VALUE for doubles + maybe_change_string_literal_to_double_literal(*($$=$1)); +}; + string_inside_quotes_value: maybe_codes { - $$=N(POOL); - O($$, OP_CREATE_SWPOOL); /* stack: empty write context */ - P($$, $1); /* some code that writes to that context */ - O($$, OP_REDUCE_SWPOOL); /* context=pop; stack: context.get_string() */ +#ifdef OPTIMIZE_BYTECODE_STRING_POOL + // it brakes ^if(" 09 "){...} + YYSTYPE code=$1; + $$=N(); + if(code->count()==3 && change_first(*code, OP::OP_STRING__WRITE, OP::OP_VALUE)){ + // optimized OP_STRING__WRITE+origin+value => OP_VALUE+origin+value without starting OP_STRING_POOL + P(*$$, *code); + } else { + OA(*$$, OP::OP_STRING_POOL, code); /* stack: empty write context */ + } +#else + $$=N(); + OA(*$$, OP::OP_STRING_POOL, $1); /* stack: empty write context */ +#endif + /* some code that writes to that context */ + /* context=pop; stack: context.get_string() */ }; /* basics */ write_string: STRING { - // optimized from OP_STRING+OP_WRITE to OP_STRING__WRITE - change_string_literal_to_write_string_literal($$=$1) + // optimized OP_STRING+OP_WRITE_VALUE => OP_STRING__WRITE + change_string_literal_to_write_string_literal(*($$=$1)); }; -empty_string_value: /* empty */ { $$=VL(NEW VString(POOL)) }; -empty: /* empty */ { $$=N(POOL) }; +empty_value: /* empty */ { $$=VL(/*we know that we will not change it*/const_cast(&vempty), 0, 0, 0); } +true_value: "true" { $$ = VL(/*we know that we will not change it*/const_cast(&vtrue), 0, 0, 0); } +false_value: "false" { $$ = VL(/*we know that we will not change it*/const_cast(&vfalse), 0, 0, 0); } + +empty: /* empty */ { $$=N(); }; %% +#endif /* 000$111(2222)00 @@ -537,84 +821,224 @@ empty: /* empty */ { $$=N(POOL) }; 4:[^({]=pop */ -int yylex(YYSTYPE *lvalp, void *pc) { - #define lexical_brackets_nestage PC->brackets_nestages[PC->sp] +inline void ungetc(Parse_control& pc, uint last_line_end_col) { + pc.source--; + if(pc.pos.col==0) { + --pc.pos.line; pc.pos.col=last_line_end_col; + } else + --pc.pos.col; + +} +static int yylex(YYSTYPE *lvalp, void *apc) { + register Parse_control& pc=*static_cast(apc); + + #define lexical_brackets_nestage pc.brackets_nestages[pc.ls_sp] #define RC {result=c; goto break2; } - register int c; - int result; + register int c; + int result; - if(PC->pending_state) { - result=PC->pending_state; - PC->pending_state=0; + if(pc.pending_state) { + result=pc.pending_state; + pc.pending_state=0; return result; } - const char *begin=PC->source; + const char *begin=pc.source; + Pos begin_pos=pc.pos; const char *end; - int begin_line=PC->line; int skip_analized=0; while(true) { - c=*(end=(PC->source++)); + c=*(end=(pc.source++)); +// fprintf(stderr, "\nchar: %c %02X; nestage: %d, sp=%d", c, c, lexical_brackets_nestage, pc.sp); - if(c=='\n') { - PC->line++; - PC->col=0; - } else - PC->col++; - - // todo: # in 0+1 column comment - - // escaping: ^^ ^$ ^; ^) ^} ^( ^{ ^" - if(c=='^') - switch(*PC->source) { - case '^': case '$': case ';': + if(c=='\n') + pc.pos_next_line(); + else + pc.pos_next_c(c); +// fprintf(stderr, "\nchar: %c file(%d:%d)", c, pc.pos.line, pc.pos.col); + + if(pc.pos.col==0+1 && c=='@') { + if(pc.ls==LS_DEF_SPECIAL_BODY) { + // @SPECIAL + // ... + // @string->APPEND(begin, end-begin, PC->file, begin_line); + pc.string.append_strdup_know_length(begin, end-begin); } // reset piece 'begin' position & line - begin=PC->source; // ^ - begin_line=PC->line; - // skip over ^ and _ - PC->source++; PC->col++; + begin=pc.source; // ->punctuation + begin_pos=pc.pos; + // skip over _ after ^ + pc.source++; pc.pos.col++; + // skip analysis = forced literal + continue; + + // converting ^#HH into char(hex(HH)) + case '#': + if(end!=begin) { + if(!pc.string_start) + pc.string_start=begin_pos; + // append piece till ^ + pc.string.append_strdup_know_length(begin, end-begin); + } + // #HH ? + if(pc.source[1] && isxdigit(pc.source[1]) && pc.source[2] && isxdigit(pc.source[2])) { + char c=(char)( + hex_value[(unsigned char)pc.source[1]]*0x10+ + hex_value[(unsigned char)pc.source[2]]); + if(c==0) { + result=BAD_HEX_LITERAL; + goto break2; // wrong hex value[no ^#00 chars allowed]: bail out + } + // append char(hex(HH)) + pc.string.append(c); + // skip over ^#HH + pc.source+=3; + pc.pos.col+=3; + // reset piece 'begin' position & line + begin=pc.source; // ->after ^#HH + begin_pos=pc.pos; + // skip analysis = forced literal + continue; + } + // just escaped char + // reset piece 'begin' position & line + begin=pc.source; + begin_pos=pc.pos; + // skip over _ after ^ + pc.source++; pc.pos.col++; // skip analysis = forced literal continue; } - switch(PC->ls) { + break; + } + } + // #comment start skipping + if(c=='#' && pc.pos.col==1) { + if(end!=begin) { + if(!pc.string_start) + pc.string_start=begin_pos; + // append piece till # + pc.string.append_strdup_know_length(begin, end-begin); + } + // fall into COMMENT lexical state [wait for \n] + push_LS(pc, LS_USER_COMMENT); + continue; + } + switch(pc.ls) { // USER'S = NOT OURS case LS_USER: + case LS_NAME_SQUARE_PART: // name.[here].xxx + if(pc.trim_bof) + switch(c) { + case '\n': case ' ': case '\t': + begin=pc.source; + begin_pos=pc.pos; + continue; // skip it + default: + pc.trim_bof=false; + } switch(c) { case '$': - push_LS(PC, LS_VAR_NAME_SIMPLE); + push_LS(pc, LS_VAR_NAME_SIMPLE_WITH_COLON); RC; case '^': - push_LS(PC, LS_METHOD_NAME); + push_LS(pc, LS_METHOD_NAME); RC; - case '@': - if(PC->col==0+1) { - push_LS(PC, LS_DEF_NAME); - RC; - } + case ']': + if(pc.ls==LS_NAME_SQUARE_PART) + if(--lexical_brackets_nestage==0) {// $name.[co<]?>de<]?> + pop_LS(pc); // $name.[co<]>de<]!> + RC; + } + break; + case '[': // $name.[co<[>de] + if(pc.ls==LS_NAME_SQUARE_PART) + lexical_brackets_nestage++; break; } + if(pc.explicit_result && c) + switch(c) { + default: + pc.string.append(c); + case '\n': case ' ': case '\t': + begin=pc.source; + begin_pos=pc.pos; + continue; + } + break; + + // #COMMENT + case LS_USER_COMMENT: + if(c=='\n') { + // skip comment + begin=pc.source; + begin_pos=pc.pos; + + pop_LS(pc); + continue; + } break; // STRING IN EXPRESSION - case LS_EXPRESSION_STRING: + case LS_EXPRESSION_STRING_QUOTED: + case LS_EXPRESSION_STRING_APOSTROFED: switch(c) { case '"': - pop_LS(PC); //"abc". - RC; + case '\'': + if( + (pc.ls == LS_EXPRESSION_STRING_QUOTED && c=='"') || + (pc.ls == LS_EXPRESSION_STRING_APOSTROFED && c=='\'') ) { + pop_LS(pc); //"abc". | 'abc'. + RC; + } + break; case '$': - push_LS(PC, LS_VAR_NAME_SIMPLE); + push_LS(pc, LS_VAR_NAME_SIMPLE_WITH_COLON); RC; case '^': - push_LS(PC, LS_METHOD_NAME); + push_LS(pc, LS_METHOD_NAME); RC; } break; @@ -623,23 +1047,26 @@ int yylex(YYSTYPE *lvalp, void *pc) { case LS_DEF_NAME: switch(c) { case '[': - PC->ls=LS_DEF_PARAMS; + pc.ls=LS_DEF_PARAMS; RC; case '\n': - PC->ls=LS_DEF_SPECIAL_BODY; + pc.ls=LS_DEF_SPECIAL_BODY; RC; } break; case LS_DEF_PARAMS: switch(c) { + case '$': // common error + result=BAD_METHOD_PARAMETER_NAME_CHARACTER; + goto break2; case ';': RC; case ']': - PC->ls=*PC->source=='['?LS_DEF_LOCALS:LS_DEF_COMMENT; + pc.ls=*pc.source=='['?LS_DEF_LOCALS:LS_DEF_COMMENT; RC; case '\n': // wrong. bailing out - pop_LS(PC); + pop_LS(pc); RC; } break; @@ -650,30 +1077,24 @@ int yylex(YYSTYPE *lvalp, void *pc) { case ';': RC; case ']': - PC->ls=LS_DEF_COMMENT; + pc.ls=LS_DEF_COMMENT; RC; case '\n': // wrong. bailing out - pop_LS(PC); + pop_LS(pc); RC; } break; case LS_DEF_COMMENT: if(c=='\n') { - pop_LS(PC); + pop_LS(pc); RC; } break; case LS_DEF_SPECIAL_BODY: - if(c=='\n') { - switch(*PC->source) { - case '@': case 0: // end of special_code - pop_LS(PC); - break; - } + if(c=='\n') RC; - } break; // (EXPRESSION) @@ -681,83 +1102,140 @@ int yylex(YYSTYPE *lvalp, void *pc) { case LS_METHOD_ROUND: switch(c) { case ')': - if(--lexical_brackets_nestage==0) - if(PC->ls==LS_METHOD_ROUND) // method round param ended - PC->ls=LS_METHOD_AFTER; // look for method end - else // PC->ls==LS_VAR_ROUND // variable constructor ended - pop_LS(PC); // return to normal life + if(--lexical_brackets_nestage==0) { + if(pc.ls==LS_METHOD_ROUND) // method round param ended + pc.ls=LS_METHOD_AFTER; // look for method end + else // pc.ls==LS_VAR_ROUND // variable constructor ended + pop_LS(pc); // return to normal life + } RC; + case '#': // comment start skipping + if(end!=begin) { + if(!pc.string_start) + pc.string_start=begin_pos; + // append piece till # + pc.string.append_strdup_know_length(begin, end-begin); + } + // fall into COMMENT lexical state [wait for \n] + push_LS(pc, LS_EXPRESSION_COMMENT); + lexical_brackets_nestage=1; + continue; case '$': - push_LS(PC, LS_EXPRESSION_VAR_NAME); + push_LS(pc, LS_EXPRESSION_VAR_NAME_WITH_COLON); RC; case '^': - push_LS(PC, LS_METHOD_NAME); + push_LS(pc, LS_METHOD_NAME); RC; case '(': lexical_brackets_nestage++; RC; case '-': - if(*PC->source=='f') { // -f + switch(*pc.source) { + case 'f': // -f skip_analized=1; result=FEXISTS; - } else + goto break2; + case 'd': // -d + skip_analized=1; + result=DEXISTS; + goto break2; + default: // minus result=c; + goto break2; + } goto break2; - case '+': case '*': case '/': case '%': + case '+': case '*': case '/': case '%': case '\\': case '~': case ';': RC; - case '&': case '|': case '#': - if(*PC->source==c) { // && || - result=c=='#'?LXOR:c=='&'?LAND:LOR; + case '&': case '|': + if(*pc.source==c) { // && || + result=c=='&'?LAND:LOR; skip_analized=1; } else result=c; goto break2; - case '<': case '>': case '=': case '!': - if(*PC->source=='=') { // <= >= == != + case '!': + switch(pc.source[0]) { + case '|': // !| !|| skip_analized=1; - switch(c) { - case '<': result=NLE; break; - case '>': result=NGE; break; - case '=': result=NEQ; break; - case '!': result=NNE; break; - } - } else - result=c; + if(pc.source[1]=='|') { + skip_analized++; + result=LXOR; + } else + result=NXOR; + goto break2; + case '=': // != + skip_analized=1; + result=NNE; + goto break2; + } + RC; + + case '<': // <<, <=, < + switch(*pc.source) { + case '<': // <[<] + skip_analized=1; result=NSL; break; + case '=': // <[=] + skip_analized=1; result=NLE; break; + default: // <[] + result=c; break; + } + goto break2; + case '>': // >>, >=, > + switch(*pc.source) { + case '>': // >[>] + skip_analized=1; result=NSR; break; + case '=': // >[=] + skip_analized=1; result=NGE; break; + default: // >[] + result=c; break; + } + goto break2; + case '=': // == + switch(*pc.source) { + case '=': // =[=] + skip_analized=1; result=NEQ; break; + default: // =[] + result=c; break; // not used now + } goto break2; + case '"': - push_LS(PC, LS_EXPRESSION_STRING); + push_LS(pc, LS_EXPRESSION_STRING_QUOTED); + RC; + case '\'': + push_LS(pc, LS_EXPRESSION_STRING_APOSTROFED); RC; case 'l': case 'g': case 'e': case 'n': if(end==begin) // right after whitespace - switch(*PC->source) { -// case '?': // ok [and bad cases, yacc would bark at them] - case 't': // lt gt [et nt] - result=c=='l'?SLT:c=='g'?SGT:BAD_STRING_COMPARISON_OPERATOR; - skip_analized=1; - goto break2; - case 'e': // le ge ne [ee] - result=c=='l'?SLE:c=='g'?SGE:c=='n'?SNE:BAD_STRING_COMPARISON_OPERATOR; - skip_analized=1; - goto break2; - case 'q': // eq [lq gq nq] - result=c=='e'?SEQ:BAD_STRING_COMPARISON_OPERATOR; - skip_analized=1; - goto break2; + if(isspace(pc.source[1])) { + switch(*pc.source) { + // case '?': // ok [and bad cases, yacc would bark at them] + case 't': // lt gt [et nt] + result=c=='l'?SLT:c=='g'?SGT:BAD_STRING_COMPARISON_OPERATOR; + skip_analized=1; + goto break2; + case 'e': // le ge ne [ee] + result=c=='l'?SLE:c=='g'?SGE:c=='n'?SNE:BAD_STRING_COMPARISON_OPERATOR; + skip_analized=1; + goto break2; + case 'q': // eq [lq gq nq] + result=c=='e'?SEQ:BAD_STRING_COMPARISON_OPERATOR; + skip_analized=1; + goto break2; + } } break; case 'i': if(end==begin) // right after whitespace - switch(PC->source[0]) { - case 'n': - { // in + if(isspace(pc.source[1])) { + switch(pc.source[0]) { + case 'n': // in skip_analized=1; result=IN; goto break2; - } - case 's': - { // is + case 's': // is skip_analized=1; result=IS; goto break2; @@ -766,9 +1244,29 @@ int yylex(YYSTYPE *lvalp, void *pc) { break; case 'd': if(end==begin) // right after whitespace - if(PC->source[0]=='e' && PC->source[1]=='f') { // def - skip_analized=2; - result=DEF; + if(pc.source[0]=='e' && pc.source[1]=='f') { // def + switch(pc.source[2]){ + case ' ': case '\t': case '\n': case '"': case '\'': case '^': case '$': // non-quoted string without whitespace after 'def' is not allowed + skip_analized=2; + result=DEF; + goto break2; + } + // error: incorrect char after 'def' + } + break; + case 't': + if(end==begin) // right after whitespace + if(pc.source[0]=='r' && pc.source[1]=='u' && pc.source[2]=='e') { // true + skip_analized=3; + result=LITERAL_TRUE; + goto break2; + } + break; + case 'f': + if(end==begin) // right after whitespace + if(pc.source[0]=='a' && pc.source[1]=='l' && pc.source[2]=='s' && pc.source[3]=='e') { // false + skip_analized=4; + result=LITERAL_FALSE; goto break2; } break; @@ -780,21 +1278,58 @@ int yylex(YYSTYPE *lvalp, void *pc) { // that's a leading|traling space or after-operator-space // ignoring it // reset piece 'begin' position & line - begin=PC->source; // after whitespace char - begin_line=PC->line; + begin=pc.source; // after whitespace char + begin_pos=pc.pos; + continue; + } + break; + case LS_EXPRESSION_COMMENT: + if(c=='(') + lexical_brackets_nestage++; + + switch(*pc.source) { + case '\n': case ')': + if(*pc.source==')') + if(--lexical_brackets_nestage!=0) + continue; + + // skip comment + begin=pc.source; + begin_pos=pc.pos; + + pop_LS(pc); continue; } break; // VARIABLE GET/PUT/WITH - case LS_VAR_NAME_SIMPLE: - case LS_EXPRESSION_VAR_NAME: - if(PC->ls==LS_EXPRESSION_VAR_NAME) { - // name in expr ends also before binary operators + case LS_VAR_NAME_SIMPLE_WITH_COLON: + case LS_VAR_NAME_SIMPLE_WITHOUT_COLON: + case LS_EXPRESSION_VAR_NAME_WITH_COLON: + case LS_EXPRESSION_VAR_NAME_WITHOUT_COLON: + if( + pc.ls==LS_EXPRESSION_VAR_NAME_WITH_COLON || + pc.ls==LS_EXPRESSION_VAR_NAME_WITHOUT_COLON) { + // name in expr ends also before switch(c) { + // expression minus case '-': - pop_LS(PC); - PC->source--; if(--PC->col<0) { PC->line--; PC->col=-1; } + // expression integer division + case '\\': + pop_LS(pc); + pc.ungetc(); + result=EON; + goto break2; + } + } + if( + pc.ls==LS_VAR_NAME_SIMPLE_WITHOUT_COLON || + pc.ls==LS_EXPRESSION_VAR_NAME_WITHOUT_COLON) { + // name already has ':', stop before next + switch(c) { + case ':': + pop_LS(pc); + pc.ungetc(); result=EON; goto break2; } @@ -803,43 +1338,71 @@ int yylex(YYSTYPE *lvalp, void *pc) { case 0: case ' ': case '\t': case '\n': case ';': - case ']': case '}': case ')': case '"': - case '<': case '>': // these stand for HTML brackets and expression binary ops - case '+': case '*': case '/': case '%': + case ']': case '}': case ')': + case '"': case '\'': + case '<': case '>': // these stand for HTML brackets AND expression binary ops + case '+': case '*': case '/': case '\\': case '%': case '&': case '|': case '=': case '!': - pop_LS(PC); - PC->source--; if(--PC->col<0) { PC->line--; PC->col=-1; } + // common delimiters + case ',': case '?': case '#': + // mysql column separators + case '`': + // before call + case '^': + pop_LS(pc); + pc.ungetc(); result=EON; goto break2; case '[': - PC->ls=LS_VAR_SQUARE; + // $name.<[>code] + if(pc.pos.col>1/*not first column*/ && ( + end[-1]=='$'/*was start of get*/ || + end[-1]==':'/*was class name delim */ || + end[-1]=='.'/*was name delim */ + )) { + push_LS(pc, LS_NAME_SQUARE_PART); + lexical_brackets_nestage=1; + RC; + } + pc.ls=LS_VAR_SQUARE; lexical_brackets_nestage=1; RC; case '{': if(begin==end) { // ${name}, no need of EON, switching LS - PC->ls=LS_VAR_NAME_CURLY; + pc.ls=LS_VAR_NAME_CURLY; } else { - PC->ls=LS_VAR_CURLY; + pc.ls=LS_VAR_CURLY; lexical_brackets_nestage=1; } RC; case '(': - PC->ls=LS_VAR_ROUND; + pc.ls=LS_VAR_ROUND; lexical_brackets_nestage=1; RC; case '.': // name part delim case '$': // name part subvar - case ':': // ':name' or 'class:name' + case ':': // class<:>name + // go to _WITHOUT_COLON state variant... + if(pc.ls==LS_VAR_NAME_SIMPLE_WITH_COLON) + pc.ls=LS_VAR_NAME_SIMPLE_WITHOUT_COLON; + else if(pc.ls==LS_EXPRESSION_VAR_NAME_WITH_COLON) + pc.ls=LS_EXPRESSION_VAR_NAME_WITHOUT_COLON; + // ...stop before next ':' RC; } break; case LS_VAR_NAME_CURLY: switch(c) { + case '[': + // ${name.<[>code]} + push_LS(pc, LS_NAME_SQUARE_PART); + lexical_brackets_nestage=1; + RC; case '}': // ${name} finished, restoring LS - pop_LS(PC); + pop_LS(pc); RC; case '.': // name part delim case '$': // name part subvar @@ -851,14 +1414,14 @@ int yylex(YYSTYPE *lvalp, void *pc) { case LS_VAR_SQUARE: switch(c) { case '$': - push_LS(PC, LS_VAR_NAME_SIMPLE); + push_LS(pc, LS_VAR_NAME_SIMPLE_WITH_COLON); RC; case '^': - push_LS(PC, LS_METHOD_NAME); + push_LS(pc, LS_METHOD_NAME); RC; case ']': if(--lexical_brackets_nestage==0) { - pop_LS(PC); + pop_LS(pc); RC; } break; @@ -873,14 +1436,14 @@ int yylex(YYSTYPE *lvalp, void *pc) { case LS_VAR_CURLY: switch(c) { case '$': - push_LS(PC, LS_VAR_NAME_SIMPLE); + push_LS(pc, LS_VAR_NAME_SIMPLE_WITH_COLON); RC; case '^': - push_LS(PC, LS_METHOD_NAME); + push_LS(pc, LS_METHOD_NAME); RC; case '}': if(--lexical_brackets_nestage==0) { - pop_LS(PC); + pop_LS(pc); RC; } break; @@ -894,20 +1457,33 @@ int yylex(YYSTYPE *lvalp, void *pc) { case LS_METHOD_NAME: switch(c) { case '[': - PC->ls=LS_METHOD_SQUARE; + // ^name.<[>code].xxx + if(pc.pos.col>1/*not first column*/ && ( + end[-1]=='^'/*was start of call*/ || // never, ^[ is literal... + end[-1]==':'/*was class name delim */ || + end[-1]=='.'/*was name delim */ + )) { + push_LS(pc, LS_NAME_SQUARE_PART); + lexical_brackets_nestage=1; + RC; + } + pc.ls=LS_METHOD_SQUARE; lexical_brackets_nestage=1; RC; case '{': - PC->ls=LS_METHOD_CURLY; + pc.ls=LS_METHOD_CURLY; lexical_brackets_nestage=1; RC; case '(': - PC->ls=LS_METHOD_ROUND; + pc.ls=LS_METHOD_ROUND; lexical_brackets_nestage=1; RC; case '.': // name part delim case '$': // name part subvar case ':': // ':name' or 'class:name' + case '^': // ^abc^xxx wrong. bailing out + case ']': case '}': case ')': // ^abc]}) wrong. bailing out + case ' ': // ^if ( wrong. bailing out RC; } break; @@ -915,16 +1491,16 @@ int yylex(YYSTYPE *lvalp, void *pc) { case LS_METHOD_SQUARE: switch(c) { case '$': - push_LS(PC, LS_VAR_NAME_SIMPLE); + push_LS(pc, LS_VAR_NAME_SIMPLE_WITH_COLON); RC; case '^': - push_LS(PC, LS_METHOD_NAME); + push_LS(pc, LS_METHOD_NAME); RC; case ';': // param delim RC; case ']': if(--lexical_brackets_nestage==0) { - PC->ls=LS_METHOD_AFTER; + pc.ls=LS_METHOD_AFTER; RC; } break; @@ -937,16 +1513,16 @@ int yylex(YYSTYPE *lvalp, void *pc) { case LS_METHOD_CURLY: switch(c) { case '$': - push_LS(PC, LS_VAR_NAME_SIMPLE); + push_LS(pc, LS_VAR_NAME_SIMPLE_WITH_COLON); RC; case '^': - push_LS(PC, LS_METHOD_NAME); + push_LS(pc, LS_METHOD_NAME); RC; case ';': // param delim RC; case '}': if(--lexical_brackets_nestage==0) { - PC->ls=LS_METHOD_AFTER; + pc.ls=LS_METHOD_AFTER; RC; } break; @@ -954,26 +1530,35 @@ int yylex(YYSTYPE *lvalp, void *pc) { lexical_brackets_nestage++; break; } + if(pc.explicit_result && c) + switch(c) { + default: + pc.string.append(c); + case '\n': case ' ': case '\t': + begin=pc.source; + begin_pos=pc.pos; + continue; + } break; case LS_METHOD_AFTER: if(c=='[') {/* ][ }[ )[ */ - PC->ls=LS_METHOD_SQUARE; + pc.ls=LS_METHOD_SQUARE; lexical_brackets_nestage=1; RC; - } + } if(c=='{') {/* ]{ }{ ){ */ - PC->ls=LS_METHOD_CURLY; + pc.ls=LS_METHOD_CURLY; lexical_brackets_nestage=1; RC; - } + } if(c=='(') {/* ]( }( )( */ - PC->ls=LS_METHOD_ROUND; + pc.ls=LS_METHOD_ROUND; lexical_brackets_nestage=1; RC; - } - pop_LS(PC); - PC->source--; if(--PC->col<0) { PC->line--; PC->col=-1; } + } + pop_LS(pc); + pc.ungetc(); result=EON; goto break2; } @@ -985,44 +1570,42 @@ int yylex(YYSTYPE *lvalp, void *pc) { break2: if(end!=begin) { // there is last piece? - if((c=='@' || c==0) && end[-1]=='\n') { // we are before LS_DEF_NAME or EOF? - // strip last \n - end--; - } - if(end!=begin) { // last piece still alive? + if(c=='@' || c==0) // we are before LS_DEF_NAME or EOF? + while(end!=begin && end[-1]=='\n') // trim all empty lines before LS_DEF_NAME and EOF + end--; + if(end!=begin && pc.ls!=LS_USER_COMMENT) { // last piece still alive and not comment? + if(!pc.string_start) + pc.string_start=begin_pos; // append it - PC->string->APPEND(begin, end-begin, PC->file, begin_line/*, start_col*/); + pc.string.append_strdup_know_length(begin, end-begin); } } - if(PC->string->size()) { // something accumulated? - // create STRING value: array of OP_VALUE+vstring - *lvalp=VL(NEW VString(*PC->string)); + if(!pc.string.is_empty()) { // something accumulated? + // create STRING value: array of OP_VALUE+origin+vstring +#ifdef SYMBOLS_CACHING + Value *lookup=symbols->get(pc.string); +#else + Value *lookup=0; +#endif + *lvalp=VL(lookup ? lookup : new VString(*new String(pc.string, String::L_CLEAN)), pc.file_no, pc.string_start.line, pc.string_start.col); // new pieces storage - PC->string=NEW String(POOL); + pc.string.clear(); + pc.string_start.clear(); // make current result be pending for next call, return STRING for now - PC->pending_state=result; result=STRING; + pc.pending_state=result; result=STRING; } if(skip_analized) { - PC->source+=skip_analized; PC->col+=skip_analized; + pc.source+=skip_analized; pc.pos.col+=skip_analized; } return result; } -int real_yyerror(parse_control *pc, char *s) /* Called by yyparse on error */ - { - //fprintf(stderr, "[%s]\n", s); - - strncpy(pc->error, s, MAX_STRING); // TODO: перепроверить с треклятым последним байтом - return 1; - } - -static void - yyprint( - FILE *file, - int type, - YYSTYPE value) - { - if(type==STRING) - fprintf(file, " \"%s\"", SLA2S(value)->cstr()); - } +static int real_yyerror(Parse_control *pc, const char *s) { // Called by yyparse on error + PC.error=pa_strdup(s); + return 1; +} +static void yyprint(FILE *file, int type, YYSTYPE value) { + if(type==STRING) + fprintf(file, " \"%s\"", LA2S(*value)->cstr()); +}