(*  Title:      Pure/Isar/outer_syntax.ML

     ID:         $Id$

     Author:     Markus Wenzel, TU Muenchen

     License:    GPL (GNU GENERAL PUBLIC LICENSE)

 The global Isabelle/Isar outer syntax.

 *)

 signature BASIC_OUTER_SYNTAX =

 sig

   val main: unit -> unit

   val loop: unit -> unit

   val sync_main: unit -> unit

   val sync_loop: unit -> unit

   val help: unit -> unit

 end;

 signature OUTER_SYNTAX =

 sig

   include BASIC_OUTER_SYNTAX

   structure Keyword:

     sig

       val control: string

       val diag: string

       val thy_begin: string

       val thy_switch: string

       val thy_end: string

       val thy_heading: string

       val thy_decl: string

       val thy_goal: string

       val qed: string

       val qed_block: string

       val qed_global: string

       val prf_goal: string

       val prf_block: string

       val prf_chain: string

       val prf_decl: string

       val prf_asm: string

       val prf_asm_goal: string

       val prf_script: string

       val kinds: string list

     end

   type token

   type parser

   val command: string -> string -> string ->

     (token list -> (Toplevel.transition -> Toplevel.transition) * token list) -> parser

   val markup_command: string -> string -> string ->

     (token list -> (Toplevel.transition -> Toplevel.transition) * token list) -> parser

   val markup_env_command: string -> string -> string ->

     (token list -> (Toplevel.transition -> Toplevel.transition) * token list) -> parser

   val verbatim_command: string -> string -> string ->

     (token list -> (Toplevel.transition -> Toplevel.transition) * token list) -> parser

   val improper_command: string -> string -> string ->

     (token list -> (Toplevel.transition -> Toplevel.transition) * token list) -> parser

   val dest_keywords: unit -> string list

   val dest_parsers: unit -> (string * string * string * bool) list

   val print_outer_syntax: unit -> unit

   val print_help: Toplevel.transition -> Toplevel.transition

   val add_keywords: string list -> unit

   val add_parsers: parser list -> unit

   val theory_header: token list -> (string * string list * (string * bool) list) * token list

   val deps_thy: string -> bool -> Path.T -> string list * Path.T list

   val load_thy: string -> bool -> bool -> Path.T -> unit

   val isar: bool -> bool -> Toplevel.isar

 end;

 structure OuterSyntax: OUTER_SYNTAX =

 struct

 structure T = OuterLex;

 structure P = OuterParse;

 (** outer syntax **)

 (* command keyword classification *)

 structure Keyword =

 struct

   val control = "control";

   val diag = "diag";

   val thy_begin = "theory-begin";

   val thy_switch = "theory-switch";

   val thy_end = "theory-end";

   val thy_heading = "theory-heading";

   val thy_decl = "theory-decl";

   val thy_goal = "theory-goal";

   val qed = "qed";

   val qed_block = "qed-block";

   val qed_global = "qed-global";

   val prf_goal = "proof-goal";

   val prf_block = "proof-block";

   val prf_chain = "proof-chain";

   val prf_decl = "proof-decl";

   val prf_asm = "proof-asm";

   val prf_asm_goal = "proof-asm-goal";

   val prf_script = "proof-script";

   val kinds = [control, diag, thy_begin, thy_switch, thy_end, thy_heading, thy_decl, thy_goal,

     qed, qed_block, qed_global, prf_goal, prf_block, prf_chain, prf_decl, prf_asm, prf_asm_goal,

     prf_script];

 end;

 (* parsers *)

 type token = T.token;

 type parser_fn = token list -> (Toplevel.transition -> Toplevel.transition) * token list;

 datatype markup_kind = Markup | MarkupEnv | Verbatim;

 datatype parser =

   Parser of string * (string * string * markup_kind option) * bool * parser_fn;

 fun parser int_only markup name comment kind parse =

   Parser (name, (comment, kind, markup), int_only, parse);

 (* parse command *)

 local

 fun command_body cmd (name, _) =

   (case cmd name of

     Some (int_only, parse) => P.!!! (Scan.prompt (name ^ "# ") (parse >> pair int_only))

   | None => sys_error ("no parser for outer syntax command " ^ quote name));

 fun terminator false = Scan.succeed ()

   | terminator true = P.group "terminator" (Scan.option P.sync -- P.$$$ ";" >> K ());

 in

 fun command term cmd =

   P.$$$ ";" >> K None ||

   P.sync >> K None ||

   (P.position P.command :-- command_body cmd) --| terminator term

     >> (fn ((name, pos), (int_only, f)) =>

       Some (Toplevel.empty |> Toplevel.name name |> Toplevel.position pos |>

         Toplevel.interactive int_only |> f));

 end;

 (** global syntax state **)

 local

 val global_lexicons = ref (Scan.empty_lexicon, Scan.empty_lexicon);

 val global_parsers =

   ref (Symtab.empty: (((string * string) * (bool * parser_fn)) * markup_kind option) Symtab.table);

 val global_markups = ref ([]: (string * markup_kind) list);

 fun change_lexicons f =

   let val lexs = f (! global_lexicons) in

     (case (op inter_string) (pairself Scan.dest_lexicon lexs) of

       [] => global_lexicons := lexs

     | bads => error ("Clash of outer syntax commands and keywords: " ^ commas_quote bads))

   end;

 fun get_markup (ms, (name, (_, Some m))) = (name, m) :: ms

   | get_markup (ms, _) = ms;

 fun make_markups () = global_markups := Symtab.foldl get_markup ([], ! global_parsers);

 fun change_parsers f = (global_parsers := f (! global_parsers); make_markups ());

 in

 (* get current syntax *)

 (*Note: the syntax for files is statically determined at the very

   beginning; for interactive processing it may change dynamically.*)

 fun get_lexicons () = ! global_lexicons;

 fun get_parsers () = ! global_parsers;

 fun get_parser () = apsome (#2 o #1) o curry Symtab.lookup (! global_parsers);

 fun lookup_markup name = assoc (! global_markups, name);

 fun is_markup kind name = (case lookup_markup name of Some k => k = kind | None => false);

 (* augment syntax *)

 fun add_keywords keywords = change_lexicons (apfst (fn lex =>

   (Scan.extend_lexicon lex (map Symbol.explode keywords))));

 fun add_parser (tab, Parser (name, (comment, kind, markup), int_only, parse)) =

  (if is_none (Symtab.lookup (tab, name)) then ()

   else warning ("Redefined outer syntax command " ^ quote name);

   Symtab.update ((name, (((comment, kind), (int_only, parse)), markup)), tab));

 fun add_parsers parsers =

   (change_parsers (fn tab => foldl add_parser (tab, parsers));

     change_lexicons (apsnd (fn lex => Scan.extend_lexicon lex

       (map (fn Parser (name, _, _, _) => Symbol.explode name) parsers))));

 end;

 (* print syntax *)

 fun dest_keywords () = Scan.dest_lexicon (#1 (get_lexicons ()));

 fun dest_parsers () =

   map (fn (name, (((cmt, kind), (int_only, _)), _)) => (name, cmt, kind, int_only))

     (Symtab.dest (get_parsers ()));

 fun help_outer_syntax () =

   let

     fun pretty_cmd (name, comment, _, _) =

       Pretty.block [Pretty.str (name ^ ":"), Pretty.brk 2, Pretty.str comment];

     val (int_cmds, cmds) = partition #4 (dest_parsers ());

   in

     [Pretty.strs ("syntax keywords:" :: map quote (dest_keywords ())),

       Pretty.big_list "proper commands:" (map pretty_cmd cmds),

       Pretty.big_list "improper commands (interactive-only):" (map pretty_cmd int_cmds)]

   end;

 val print_outer_syntax = Pretty.writeln o Pretty.chunks o help_outer_syntax;

 val print_help =

   Toplevel.keep (fn state =>

     let val opt_thy = try Toplevel.theory_of state in

       help_outer_syntax () @

       Method.help_methods opt_thy @

       Attrib.help_attributes opt_thy

       |> Pretty.chunks |> Pretty.writeln

     end);

 (** read theory **)

 (* special keywords *)

 val headerN = "header";

 val theoryN = "theory";

 val theory_keyword = P.$$$ theoryN;

 val header_keyword = P.$$$ headerN;

 val semicolon = P.$$$ ";";

 (* sources *)

 local

 val no_terminator =

   Scan.unless semicolon (Scan.one (T.not_sync andf T.not_eof));

 val recover = Scan.prompt "recover# " (Scan.repeat no_terminator);

 in

 fun source term do_recover cmd src =

   src

   |> Source.source T.stopper

     (Scan.bulk (fn xs => P.!!! (command term (cmd ())) xs))

     (if do_recover then Some recover else None)

   |> Source.mapfilter I;

 end;

 fun token_source (src, pos) =

   src

   |> Symbol.source false

   |> T.source false (K (get_lexicons ())) pos;

 fun filter_proper src =

   src

   |> Source.filter T.is_proper;

 (* scan header *)

 fun scan_header get_lex scan (src, pos) =

   src

   |> Symbol.source false

   |> T.source false (fn () => (get_lex (), Scan.empty_lexicon)) pos

   |> filter_proper

   |> Source.source T.stopper (Scan.single scan) None

   |> (fst o the o Source.get_single);

 (* detect new/old header *)

 local

 val check_header_lexicon = Scan.make_lexicon [Symbol.explode headerN, Symbol.explode theoryN];

 val check_header = Scan.option (header_keyword || theory_keyword);

 in

 fun is_old_theory src = is_none (scan_header (K check_header_lexicon) check_header src);

 end;

 (* deps_thy --- inspect theory header *)

 local

 val header_lexicon =

   Scan.make_lexicon (map Symbol.explode ["(", ")", "+", ":", ";", "=", "files", headerN, theoryN]);

 val file_name =

   (P.$$$ "(" |-- P.!!! (P.name --| P.$$$ ")")) >> rpair false || P.name >> rpair true;

 in

 val theory_header =

   (P.name -- (P.$$$ "=" |-- P.enum1 "+" P.name) --

     Scan.optional (P.$$$ "files" |-- P.!!! (Scan.repeat1 file_name)) [] --|  P.$$$ ":")

   >> (fn ((A, Bs), files) => (A, Bs, files));

 val new_header =

   header_keyword |-- (P.!!! (P.text -- Scan.option semicolon -- theory_keyword |-- theory_header))

     || theory_keyword |-- P.!!! theory_header;

 val old_header =

   P.!!! (P.group "theory header"

     (P.name -- (P.$$$ "=" |-- P.name -- Scan.repeat (P.$$$ "+" |-- P.name))))

   >> (fn (A, (B, Bs)) => (A, B :: Bs, []: (string * bool) list));

 fun deps_thy name ml path =

   let

     val src = Source.of_string (File.read path);

     val pos = Path.position path;

     val (name', parents, files) =

       (*unfortunately, old-style headers dynamically depend on the current lexicon*)

       if is_old_theory (src, pos) then

         scan_header ThySyn.get_lexicon (Scan.error old_header) (src, pos)

       else scan_header (K header_lexicon) (Scan.error new_header) (src, pos);

     val ml_path = ThyLoad.ml_path name;

     val ml_file = if ml andalso is_some (ThyLoad.check_file ml_path) then [ml_path] else [];

   in

     if name <> name' then

       error ("Filename " ^ quote (Path.pack path) ^

         " does not agree with theory name " ^ quote name')

     else (parents, map (Path.unpack o #1) files @ ml_file)

   end;

 end;

 (* present theory source *)

 local

 val is_improper = not o (T.is_proper orf T.is_begin_ignore orf T.is_end_ignore);

 val improper = Scan.any is_improper;

 val improper_keep_indent = Scan.repeat

   (Scan.unless (Scan.one T.is_indent -- Scan.one T.is_proper) (Scan.one is_improper));

 val improper_end =

   (improper -- semicolon) |-- improper_keep_indent ||

   improper_keep_indent;

 val ignore =

   Scan.depend (fn d => Scan.one T.is_begin_ignore >> pair (d + 1)) ||

   Scan.depend (fn 0 => Scan.fail | d => Scan.one T.is_end_ignore >> pair (d - 1)) ||

   Scan.lift (Scan.one (OuterLex.not_eof andf (not o OuterLex.is_end_ignore)));

 val opt_newline = Scan.option (Scan.one T.is_newline);

 val ignore_stuff =

   opt_newline -- Scan.one T.is_begin_ignore --

     P.!!!! (Scan.pass 0 (Scan.repeat ignore) -- Scan.one T.is_end_ignore -- opt_newline);

 fun markup_kind k = Scan.one (T.is_kind T.Command andf is_markup k o T.val_of);

 val markup = markup_kind Markup >> T.val_of;

 val markup_env = markup_kind MarkupEnv >> T.val_of;

 val verbatim = markup_kind Verbatim;

 val present_token =

   ignore_stuff >> K None ||

   (improper |-- markup -- P.!!!! (improper |-- P.text --| improper_end) >> Present.markup_token ||

     improper |-- markup_env -- P.!!!! (improper |-- P.text --| improper_end) >> Present.markup_env_token ||

     (P.$$$ "--" >> K "cmt") -- P.!!!! (improper |-- P.text) >> Present.markup_token ||

     (improper -- verbatim) |-- P.!!!! (improper |-- P.text --| improper_end)

       >> Present.verbatim_token ||

     Scan.one T.not_eof >> Present.basic_token) >> Some;

 in

 (*note: lazy evaluation ahead*)

 fun present_toks text pos () =

   token_source (Source.of_list (Library.untabify text), pos)

   |> Source.source T.stopper (Scan.bulk (Scan.error present_token)) None

   |> Source.mapfilter I

   |> Source.exhaust;

 fun present_text text () =

   Source.exhaust (Symbol.source false (Source.of_list (Library.untabify text)));

 end;

 (* load_thy --- read text (including header) *)

 local

 fun try_ml_file name time =

   let

     val path = ThyLoad.ml_path name;

     val tr = Toplevel.imperative (fn () => ThyInfo.load_file time path);

     val tr_name = if time then "time_use" else "use";

   in

     if is_none (ThyLoad.check_file path) then ()

     else Toplevel.excursion_error [Toplevel.empty |> Toplevel.name tr_name |> tr]

   end;

 fun parse_thy src_pos =

   src_pos

   |> token_source

   |> filter_proper

   |> source false false (K (get_parser ()))

   |> Source.exhaust;

 fun run_thy name path =

   let

     val text = explode (File.read path);

     val src = Source.of_list text;

     val pos = Path.position path;

   in

     Present.init_theory name;

     Present.verbatim_source name (present_text text);

     if is_old_theory (src, pos) then (ThySyn.load_thy name text;

       Present.old_symbol_source name (present_text text))   (*note: text presented twice!*)

     else (Toplevel.excursion_error (parse_thy (src, pos));

       Present.token_source name (present_toks text pos))

   end;

 in

 fun load_thy name ml time path =

  (if time then

     timeit (fn () =>

      (writeln ("\n**** Starting theory " ^ quote name ^ " ****");

       setmp Library.timing true (run_thy name) path;

       writeln ("**** Finished theory " ^ quote name ^ " ****\n")))

   else run_thy name path;

   Context.context (ThyInfo.get_theory name);

   if ml then try_ml_file name time else ());

 end;

 (* interactive source of state transformers *)

 fun isar term no_pos =

   Source.tty

   |> Symbol.source true

   |> T.source true get_lexicons

     (if no_pos then Position.none else Position.line_name 1 "stdin")

   |> filter_proper

   |> source term true get_parser;

 (** the read-eval-print loop **)

 (* main loop *)

 fun gen_loop term no_pos =

  (Context.reset_context ();

   Toplevel.loop (isar term no_pos));

 fun gen_main term no_pos =

  (Toplevel.set_state Toplevel.toplevel;

   writeln (Session.welcome ());

   gen_loop term no_pos);

 fun main () = gen_main false false;

 fun loop () = gen_loop false false;

 fun sync_main () = gen_main true true;

 fun sync_loop () = gen_loop true true;

 (* help *)

 fun help () =

   writeln ("This is Isabelle's underlying ML system (" ^ ml_system ^ ");\n\

     \invoke 'loop();' to enter the Isar loop.");

 (*final declarations of this structure!*)

 val command = parser false None;

 val markup_command = parser false (Some Markup);

 val markup_env_command = parser false (Some MarkupEnv);

 val verbatim_command = parser false (Some Verbatim);

 val improper_command = parser true None;

 end;

 (*setup theory syntax dependent operations*)

 ThyLoad.deps_thy_fn := OuterSyntax.deps_thy;

 ThyLoad.load_thy_fn := OuterSyntax.load_thy;

 structure ThyLoad: THY_LOAD = ThyLoad;

 structure BasicOuterSyntax: BASIC_OUTER_SYNTAX = OuterSyntax;

 open BasicOuterSyntax;