yecc.erl

OTP是开放电信平台的简称

atom_option(verbose) -> {verbose, true};
atom_option(Key) -> Key.

is_filename(T) ->
    try filename:flatten(T) of
        Filename -> Filename
    catch error: _ -> no
    end.    

shorten_filename(Name0) ->
    {ok,Cwd} = file:get_cwd(),
    case lists:prefix(Cwd, Name0) of
        false -> Name0;
        true ->
            case lists:nthtail(length(Cwd), Name0) of
                "/"++N -> N;
                N -> N
            end
    end.

start(Infilex, Options) ->
    Infile = assure_extension(Infilex, ".yrl"),
    {value, {_, Outfilex0}} = keysearch(parserfile, 1, Options),
    {value, {_, Includefilex}} = keysearch(includefile, 1, Options),
    Outfilex = case Outfilex0 of
                   [] -> filename:rootname(Infilex, ".yrl");
                   _ -> Outfilex0
               end,
    Includefile = case Includefilex of
                      [] -> [];
                      _ -> assure_extension(Includefilex,".hrl")
                  end,
    Outfile = assure_extension(Outfilex, ".erl"),
    Module = list_to_atom(filename:basename(Outfile, ".erl")),
    #yecc{infile = Infile, 
          outfile = Outfile,
          includefile = Includefile,
          module = Module,
          options = Options,
          verbose = member(verbose, Options),
          file_attrs = member(file_attributes, Options)}.

assure_extension(File, Ext) ->
    concat([strip_extension(File, Ext), Ext]).

%% Assumes File is a filename.
strip_extension(File, Ext) ->
    case filename:extension(File) of
        Ext -> filename:rootname(File);
        _Other -> File
    end.

infile(Parent, Infilex, Options) ->
    St0 = start(Infilex, Options),
    St = case file:open(St0#yecc.infile, [read, read_ahead]) of
             {ok, Inport} ->
                 try 
                     outfile(St0#yecc{inport = Inport})
                 after
                     ok = file:close(Inport)
                 end;
             {error, Reason} ->
                 add_error(St0#yecc.infile, none, {file_error, Reason}, St0)
         end,
    case St#yecc.errors of
        [] -> ok;
        _ -> _ = file:delete(St#yecc.outfile)
    end,
    Parent ! {self(), yecc_ret(St)}.

outfile(St0) ->
    case file:open(St0#yecc.outfile, [write, delayed_write]) of
        {ok, Outport} ->
            try 
                generate(St0#yecc{outport = Outport})
            catch 
                throw: St1  ->
                    St1;
                exit: Reason ->
                    add_error({internal_error, Reason}, St0)
            after
               ok = file:close(Outport)
            end;
        {error, Reason} ->
            add_error(St0#yecc.outfile, none, {file_error, Reason}, St0)
    end.

os_process_size() ->
    case os:type() of
        {unix, sunos} ->
            Size = os:cmd("ps -o vsz -p " ++ os:getpid() ++ " | tail -1"),
            list_to_integer(lib:nonl(Size));
        _ ->
            0
    end.            

timeit(Name, Fun, St0) ->
    Time = runtime,
    %% Time = wall_clock,
    {Before, _} = statistics(Time),
    St = Fun(St0), 
    {After, _} = statistics(Time),
    Mem0 = erts_debug:flat_size(St)*erlang:system_info(wordsize),
    Mem = lists:flatten(io_lib:format("~.1f kB", [Mem0/1024])),
    Sz = lists:flatten(io_lib:format("~.1f MB", [os_process_size()/1024])),
    io:fwrite(" ~-30w: ~10.2f s ~12s ~10s\n", 
              [Name, (After-Before)/1000, Mem, Sz]),
    St.

-define(PASS(P), {P, fun P/1}).

generate(St0) ->
    Passes = [?PASS(parse_grammar), ?PASS(check_grammar),
              ?PASS(states_and_goto_table), ?PASS(parse_actions),
              ?PASS(action_conflicts), ?PASS(write_file)],
    F = case member(time, St0#yecc.options) of
            true -> 
                io:fwrite("Generating parser from grammar in ~s\n", 
                          [format_filename(St0#yecc.infile)]),
                fun timeit/3;
            false ->
                fun(_Name, Fn, St) -> Fn(St) end
        end,
    Fun = fun({Name, Fun}, St) ->
                  St2 = F(Name, Fun, St),
                  if 
                      St2#yecc.errors =:= [] -> St2;
                      true -> throw(St2)
                  end
          end,
    foldl(Fun, St0, Passes).

parse_grammar(St) ->
    parse_grammar(St#yecc.inport, 1, St).

parse_grammar(Inport, Line, St) ->
    {NextLine, Grammar} = read_grammar(Inport, Line),
    parse_grammar(Grammar, Inport, NextLine, St).

parse_grammar(eof, _Inport, _NextLine, St) ->
    St;
parse_grammar({#symbol{name = 'Erlang'}, [#symbol{name = code}]}, _Inport, 
              NextLine, St) ->
    St#yecc{erlang_code = NextLine};
parse_grammar(Grammar, Inport, NextLine, St0) ->
    St = parse_grammar(Grammar, St0),
    parse_grammar(Inport, NextLine, St).

parse_grammar({error,ErrorLine,Error}, St) ->
    add_error(ErrorLine, Error, St);
parse_grammar({rule, Rule, Tokens}, St0) ->
    NmbrOfDaughters = case Rule of
                          [_, #symbol{name = '$empty'}]  -> 0;
                          _ -> length(Rule) - 1
                      end,
    St1 = check_action(Tokens, St0),
    {Tokens1, St2} = subst_pseudo_vars(Tokens,
                                       NmbrOfDaughters,
                                       St1),
    RuleDef = #rule{symbols = Rule, tokens = Tokens1},
    St2#yecc{rules_list = [RuleDef | St2#yecc.rules_list]};
parse_grammar({prec, Prec}, St) ->
    St#yecc{prec = Prec ++ St#yecc.prec};
parse_grammar({#symbol{line = Line, name = Name}, Symbols}, St) ->
    CF = fun(I) ->
                 case element(I, St) of
                     [] -> 
                         setelement(I, St, Symbols);
                     _ -> 
                         add_error(Line, {duplicate_declaration, Name}, St)
                 end
         end,
    OneSymbol = length(Symbols) =:= 1,
    case Name of
        'Nonterminals' -> CF(#yecc.nonterminals);
        'Terminals' -> CF(#yecc.terminals);
        'Rootsymbol' when OneSymbol -> CF(#yecc.rootsymbol);
        'Endsymbol' when OneSymbol ->  CF(#yecc.endsymbol);
        'Expect' when OneSymbol -> CF(#yecc.expect_shift_reduce);
        'States' when OneSymbol -> CF(#yecc.expect_n_states); % undocumented
        _ -> add_warning(Line, bad_declaration, St)
    end.

read_grammar(Inport, Line) ->
    case yeccscan:scan(Inport, '', Line) of
        {eof, NextLine} ->
            {NextLine, eof};
        {error, {ErrorLine, Mod, What}, NextLine} ->
            {NextLine, {error, ErrorLine, {error, Mod, What}}};
        {ok, Input, NextLine} ->
            {NextLine, case yeccparser:parse(Input) of
                           {error, {ErrorLine, Mod, Message}} ->
                               {error, ErrorLine, {error, Mod, Message}};
                           {ok, {rule, Rule, {erlang_code, Tokens}}} ->
                               {rule, Rule, Tokens};
                           {ok, {#symbol{name=P}, 
                                 [#symbol{name=I} | OpL]}=Ss} ->
                               A = precedence(P),
                               if
                                   A =/= unknown, 
                                   is_integer(I),
                                   OpL =/= [] ->
                                       Ps = [{Op, I , A} || Op <- OpL],
                                       {prec, Ps};
                                   true -> 
                                       Ss
                               end
                       end}
    end.

precedence('Left') -> left;
precedence('Right') -> right;
precedence('Unary') -> unary;
precedence('Nonassoc') -> nonassoc;
precedence(_) -> unknown.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

check_grammar(St0) ->
    Empty = #symbol{line = none, name = '$empty'},
    AllSymbols = St0#yecc.nonterminals ++ St0#yecc.terminals ++ [Empty],
    St1 = St0#yecc{all_symbols = AllSymbols},
    Cs = [fun check_nonterminals/1, fun check_terminals/1, 
          fun check_rootsymbol/1, fun check_endsymbol/1, 
          fun check_expect/1, fun check_states/1,
          fun check_precedences/1, fun check_rules/1],
    foldl(fun(F, St) -> F(St) end, St1, Cs).

check_nonterminals(St) ->
    case St#yecc.nonterminals of 
        [] ->
            add_error(nonterminals_missing, St);
        Nonterminals ->
            {Unique, Dups} = duplicates(names(Nonterminals)),
            St1 = add_warnings(Dups, duplicate_nonterminal, St),
            St2 = check_reserved(Unique, St1),
            St2#yecc{nonterminals = [?ACCEPT | Unique]}
    end.

check_terminals(St0) ->
    case St0#yecc.terminals of
        [] ->
            add_error(terminals_missing, St0);
        Terminals ->
            {Unique, Dups} = duplicates(names(Terminals)),
            St1 = add_warnings(Dups, duplicate_terminal, St0),
            Common = intersect(St1#yecc.nonterminals, Unique),
            St2 = add_errors(Common, symbol_terminal_and_nonterminal, St1),
            St3 = check_reserved(Unique, St2),
            St3#yecc{terminals = ['$empty' | Unique]}
    end.

check_reserved(Names, St) ->
    add_errors(intersect(Names, ['$empty', '$end', '$undefined']),
               reserved, St).

check_rootsymbol(St) ->
    case St#yecc.rootsymbol of
        [] ->
            add_error(rootsymbol_missing, St);
        [#symbol{line = Line, name = SymName}] ->
            case kind_of_symbol(St, SymName) of
                nonterminal ->
                    St#yecc{rootsymbol = SymName};
                _ ->
                    add_error(Line, {bad_rootsymbol, SymName}, St)
            end
    end.

check_endsymbol(St) ->
    case St#yecc.endsymbol of
        [] ->
            St#yecc{endsymbol = '$end'};
        [#symbol{line = Line, name = SymName}] ->
            case kind_of_symbol(St, SymName) of
                nonterminal ->
                    add_error(Line, {endsymbol_is_nonterminal, SymName}, St);
                terminal ->
                    add_error(Line, {endsymbol_is_terminal, SymName}, St);
                _ ->
                    St#yecc{endsymbol = SymName}
            end
    end.

check_expect(St0) ->
    case St0#yecc.expect_shift_reduce of
        [] ->
            St0#yecc{expect_shift_reduce = 0};
        [#symbol{name = Expect}] when is_integer(Expect) ->
            St0#yecc{expect_shift_reduce = Expect};
        [#symbol{line = Line, name = Name}] ->
            St1 = add_error(Line, {bad_expect, Name}, St0),
            St1#yecc{expect_shift_reduce = 0}
    end.

check_states(St) ->
    case St#yecc.expect_n_states of
        [] ->
            St;
        [#symbol{name = NStates}] when is_integer(NStates) ->
            St#yecc{expect_n_states = NStates};
        [#symbol{line = Line, name = Name}] ->
            add_error(Line, {bad_states, Name}, St)
    end.

check_precedences(St0) ->
    {St1, _} = 
        foldr(fun({#symbol{line = Line, name = Op},_I,_A}, {St,Ps}) ->
                      case member(Op, Ps) of
                          true ->
                              {add_error(Line, {duplicate_precedence,Op}, St),
                               Ps};
                          false ->
                              {St, [Op | Ps]}
                      end
              end, {St0,[]}, St0#yecc.prec),
    foldl(fun({#symbol{line = Line, name = Op},I,A}, St) ->
                  case kind_of_symbol(St, Op) of
                      endsymbol ->
                          add_error(Line,{precedence_op_is_endsymbol,Op}, St);
                      unknown ->
                          add_error(Line, {precedence_op_is_unknown, Op}, St);
                      _ -> 
                          St#yecc{prec = [{Op,I,A} | St#yecc.prec]}
                  end
          end, St1#yecc{prec = []}, St1#yecc.prec).

check_rule(Rule0, {St0,Rules}) ->
    Symbols = Rule0#rule.symbols,
    #symbol{line = HeadLine, name = Head} = hd(Symbols),
    case member(Head, St0#yecc.nonterminals) of
        false -> 
            {add_error(HeadLine, {undefined_nonterminal, Head}, St0), Rules};
        true ->
            St = check_rhs(tl(Symbols), St0),
            Rule = Rule0#rule{line = HeadLine, symbols = names(Symbols)},
            {St, [Rule | Rules]}
    end.

check_rules(St0) ->
    {St,Rules0} = foldl(fun check_rule/2, {St0,[]}, St0#yecc.rules_list),
    case St#yecc.rules_list of
        [] ->
            add_error(no_grammar_rules, St);
        _ ->
            Rule = #rule{line = none, 
                         symbols = [?ACCEPT, St#yecc.rootsymbol],
                         tokens = []},
            Rules1 = [Rule | Rules0],
            Rules = map(fun({R,I}) -> R#rule{n = I} end,  count(0, Rules1)),