typer.ml 52.8 KB
Newer Older
1
2
3
open Location
open Ast
open Ident
4

5
6
7
8
9
10
let (=) (x:int) y = x = y
let (<=) (x:int) y = x <= y
let (<) (x:int) y = x < y
let (>=) (x:int) y = x >= y
let (>) (x:int) y = x > y

11
let warning loc msg =
12
  let v = Location.get_viewport () in
13
  let ppf = if Html.is_html v then Html.ppf v else Format.err_formatter in
14
15
16
  Format.fprintf ppf "Warning %a:@\n" Location.print_loc (loc,`Full);
  Location.html_hilight (loc,`Full);
  Format.fprintf ppf "%s@." msg
17

18
19
20
21
22
23
24
25
exception NonExhaustive of Types.descr
exception Constraint of Types.descr * Types.descr
exception ShouldHave of Types.descr * string
exception ShouldHave2 of Types.descr * string * Types.descr
exception WrongLabel of Types.descr * label
exception UnboundId of id * bool
exception UnboundExtId of Types.CompUnit.t * id
exception Error of string
26
27
exception Warning of string * Types.t

28
29
30
31
let raise_loc loc exn = raise (Location (loc,`Full,exn))
let raise_loc_str loc ofs exn = raise (Location (loc,`Char ofs,exn))
let error loc msg = raise_loc loc (Error msg)

32
33
type item =
  | Type of Types.t
34
  | Val of Types.t
35

36
37
38
39
40
type ext =
  | ECDuce of Types.CompUnit.t   (* CDuce unit *)
  | EOCaml of string             (* OCaml module *)
  | ESchema of string            (* XML Schema *)

41
42
module UEnv = Map.Make(U)

43
type t = {
44
  ids : item Env.t;
45
  ns: Ns.table;
46
  cu: ext UEnv.t;
47
  keep_ns: bool
48
}
49

50
51
52
53
54
let load_schema = ref (fun _ _ -> assert false)
let from_comp_unit = ref (fun _ -> assert false)
let has_comp_unit = ref (fun _ -> assert false)
let has_ocaml_unit = ref (fun _ -> false)
let has_static_external = ref (fun _ -> assert false)
55

56
let schemas = Hashtbl.create 13
57

58
let type_schema env x uri =
59
60
  if not (Hashtbl.mem schemas uri) then
    Hashtbl.add schemas uri (!load_schema x uri);
61
  { env with cu = UEnv.add x (ESchema uri) env.cu }
62

63
(* TODO: filter out builtin defs ? *)
64
65
66
67
let serialize_item s = function
  | Type t -> Serialize.Put.bits 1 s 0; Types.serialize s t
  | Val t -> Serialize.Put.bits 1 s 1; Types.serialize s t

68
let serialize s env =
69
  Serialize.Put.env Id.serialize serialize_item Env.iter s env.ids;
70
71
72
  Ns.serialize_table s env.ns;

  let schs =
73
74
75
    UEnv.fold 
      (fun name cu accu -> 
	 match cu with ESchema uri -> (name,uri)::accu | _ -> accu) 
76
      env.cu [] in
77
78
  Serialize.Put.list 
    (Serialize.Put.pair U.serialize Serialize.Put.string) s schs
79

80
81
82
83
84
let deserialize_item s = match Serialize.Get.bits 1 s with
  | 0 -> Type (Types.deserialize s)
  | 1 -> Val (Types.deserialize s)
  | _ -> assert false

85
let deserialize s =
86
87
  let ids = 
    Serialize.Get.env Id.deserialize deserialize_item Env.add Env.empty s in
88
  let ns = Ns.deserialize_table s in
89
90
91
92
  let schs = 
    Serialize.Get.list 
      (Serialize.Get.pair U.deserialize Serialize.Get.string) s in
  let env = 
93
    { ids = ids; ns = ns; cu = UEnv.empty; keep_ns = false } in
94
  List.fold_left (fun env (name,uri) -> type_schema env name uri) env schs
95
96


97
98
let empty_env = {
  ids = Env.empty;
99
  ns = Ns.def_table;
100
  cu = UEnv.empty;
101
  keep_ns = false
102
103
}

104
let enter_cu x cu env =
105
  { env with cu = UEnv.add x (ECDuce cu) env.cu }
106

107
let find_cu loc x env =
108
  try UEnv.find x env.cu
109
110
111
112
  with Not_found ->
    if !has_comp_unit x then (ECDuce (Types.CompUnit.mk x))
    else if !has_ocaml_unit x then (EOCaml (U.get_str x))
    else error loc ("Cannot find external unit " ^ (U.to_string x))
113
114


115
let find_schema x env =
116
117
118
119
120
121
  try 
    (match UEnv.find x env.cu with
      | ESchema s -> s 
      | _ -> raise Not_found)
  with Not_found -> 
    raise (Error (Printf.sprintf "%s: no such schema" (U.to_string x)))
122

123
124
125
126
127
128
129
130
let enter_type id t env =
  { env with ids = Env.add id (Type t) env.ids }
let enter_types l env =
  { env with ids = 
      List.fold_left (fun accu (id,t) -> Env.add id (Type t) accu) env.ids l }
let find_type id env =
  match Env.find id env.ids with
    | Type t -> t
131
    | Val _ -> raise Not_found
132

133

134
let enter_value id t env = 
135
  { env with ids = Env.add id (Val t) env.ids }
136
137
let enter_values l env =
  { env with ids = 
138
      List.fold_left (fun accu (id,t) -> Env.add id (Val t) accu) env.ids l }
139
140
141
let enter_values_dummy l env =
  { env with ids = 
      List.fold_left (fun accu id -> Env.add id (Val Types.empty) accu) env.ids l }
142
143
let find_value id env =
  match Env.find id env.ids with
144
    | Val t -> t
145
    | _ -> raise Not_found
146
147
148
let find_value_global loc cu id env =
  try find_value id (!from_comp_unit cu)
  with Not_found -> raise_loc loc (UnboundExtId (cu,id))
149
	
150
151
152
153
154
155
let value_name_ok id env =
  try match Env.find id env.ids with
    | Val t -> true
    | _ -> false
  with Not_found -> true

156
157
158
159
let iter_values env f =
  Env.iter (fun x ->
	      function Val t -> f x t;
		| _ -> ()) env.ids
160

161

162
let register_types cu env =
163
164
165
  Env.iter (fun x t -> match t with
	      | Type t -> Types.Print.register_global cu (Ident.value x) t
	      | _ -> ()) env.ids
166

167

168
(* Namespaces *)
169

170
let set_ns_table_for_printer env = 
171
  Ns.InternalPrinter.set_table env.ns
172

173
let get_ns_table tenv = tenv.ns
174

175
let type_ns env p ns =
176
  { env with ns = Ns.add_prefix p ns env.ns }
177

178
179
180
let type_keep_ns env k =
  { env with keep_ns = k }

181
182
183
184
185
let protect_error_ns loc f x =
  try f x
  with Ns.UnknownPrefix ns ->
    raise_loc_generic loc 
    ("Undefined namespace prefix " ^ (U.to_string ns))
186

187
188
189
let qname env loc t = 
  protect_error_ns loc (Ns.map_tag env.ns) t
    
190
191
192
193
194
195
196
197
198
199
let ident env loc t =
  let q = protect_error_ns loc (Ns.map_attr env.ns) t in
  Ident.ident q

let has_value id env =
  try match Env.find (Ident.ident (Ns.map_attr env.ns id)) env.ids with
    | Val t -> true
    | _ -> false
  with Not_found | Ns.UnknownPrefix _ -> false

200
let parse_atom env loc t =
201
  Atoms.V.of_qname (qname env loc t)
202
203
 
let parse_ns env loc ns =
204
  protect_error_ns loc (Ns.map_prefix env.ns) ns
205

206
let parse_label env loc t =
207
  let (ns,l) = protect_error_ns loc (Ns.map_attr env.ns) t in
208
  LabelPool.mk (ns,l)
209

210
211
212
213
214
215
216
217
218
219
220
221
222
let parse_record env loc f r =
  let r = List.map (fun (l,x) -> (parse_label env loc l, f x)) r in
  LabelMap.from_list (fun _ _ -> raise_loc_generic loc "Duplicated record field") r

let rec const env loc = function
  | LocatedExpr (loc,e) -> const env loc e
  | Pair (x,y) -> Types.Pair (const env loc x, const env loc y)
  | Xml (x,y) -> Types.Xml (const env loc x, const env loc y)
  | RecordLitt x -> Types.Record (parse_record env loc (const env loc) x)
  | String (i,j,s,c) -> Types.String (i,j,s,const env loc c)
  | Atom t -> Types.Atom (parse_atom env loc t)
  | Integer i -> Types.Integer i
  | Char c -> Types.Char c
223
  | Const c -> c
224
225
226
227
  | _ -> raise_loc_generic loc "This should be a scalar or structured constant"

(* I. Transform the abstract syntax of types and patterns into
      the internal form *)
228

229

230
231
232
233
let get_schema_names env = 
  UEnv.fold 
    (fun n cu acc -> match cu with ESchema _ -> n :: acc | _ -> acc) env.cu []
    
234
235
let find_schema_component uri name =
  Env.find (Ident.ident name) (Hashtbl.find schemas uri)
236

237
238
239
let get_schema_validator uri name =
  snd (find_schema_component uri name)

240
let find_schema_descr uri (name : Ns.qname) =
241
  try fst (find_schema_component uri name)
242
243
244
  with Not_found ->    
    raise (Error (Printf.sprintf "No component named '%s' found in schema '%s'"
		    (Ns.QName.to_string name) uri))
245
246


247
248
249
250
251
252
let find_type_global loc cu id env =
  match find_cu loc cu env with
    | ECDuce cu -> find_type id (!from_comp_unit cu)
    | EOCaml _ -> error loc "OCaml units don't export types" (* TODO *)
    | ESchema s -> find_schema_descr s (Ident.value id)
	
253

254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
module IType = struct
  type node = {
    mutable desc: desc;
    mutable smallhash: int;  (* Local hash *)
    mutable rechash: int;    (* Global (recursive) hash *)
    mutable sid: int;        (* Sequential id used to compute rechash *)
    mutable t: Types.t option;
    mutable tnode: Types.Node.t option;
    mutable p: Patterns.descr option;
    mutable pnode: Patterns.node option;
    mutable fv: fv option
  } 
  and desc =
    | ILink of node
    | IType of Types.descr * int
    | IOr of node * node
    | IAnd of node * node
    | IDiff of node * node
    | ITimes of node * node
    | IXml of node * node
    | IArrow of node * node
    | IOptional of node
    | IRecord of bool * (node * node option) label_map
    | ICapture of id
    | IConstant of id * Types.const
279
    | IConcat of node * node
280
    | IMerge of node * node
281
282
283
284
285
286
287
288

  let rec node_temp = { 
    desc = ILink node_temp;
    smallhash = 0; rechash = 0; sid = 0;
    t = None; tnode = None; p = None; pnode = None;
    fv = None
  }
			
289
(* Recursive hash-consing *)
290

291
292
293
294
295
296
  let hash_field f = function
    | (p, Some e) -> 1 + 17 * f p + 257 * f e
    | (p, None) -> 2 + 17 * f p

  let rec hash f n = match n.desc with
    | ILink n -> hash f n
297
    | IType (t,h) -> 1 + 17 * h
298
299
300
301
302
303
304
305
306
    | IOr (p1,p2) -> 2 + 17 * f p1 + 257 * f p2
    | IAnd (p1,p2) -> 3 + 17 * f p1 + 257 * f p2
    | IDiff (p1,p2) -> 4 + 17 * f p1 + 257 * f p2
    | ITimes (p1,p2) -> 5 + 17 * f p1 + 257 * f p2
    | IXml (p1,p2) -> 6 + 17 * f p1 + 257 * f p2
    | IArrow (p1,p2) -> 7 + 17 * f p1 + 257 * f p2
    | IOptional p -> 8 + 17 * f p
    | IRecord (o,r)->9+(if o then 17 else 0)+
	257*(LabelMap.hash (hash_field f) r)
307
308
    | ICapture x -> 10 + 17 * (Id.hash x)
    | IConstant (x,c) -> 11 + 17 * (Id.hash x) + 257*(Types.Const.hash c)
309
    | IConcat _ | IMerge _ -> assert false
310

311
312
313
314
315
  let hash0 = hash (fun n -> 1)
  let hash1 = hash hash0
  let hash2 = hash hash1
  let hash3 = hash hash2

316
317
  let smallhash n =
    if n.smallhash !=0 then n.smallhash
318
319
320
321
    else (
      let h = hash2 n in 
      n.smallhash <- h; h
    )
322
323

  let rec repr = function
324
    | { desc = ILink n } as m -> let z = repr n in m.desc <- ILink z; z
325
326
327
328
    | n -> n

  let back = ref []

329
330
331
332
  let rec prot_repr = function
    | { desc = ILink n } -> repr n
    | n -> n

333
334
335
336
337
338
339
340
341
  let link x y = match x,y with
    | { t = None } as x, y 
    | y, ({ t = None } as x) -> back := (x,x.desc) :: !back; x.desc <- ILink y
    | _ -> assert false

  exception Unify

  let rec unify x y =
    if x == y then ()
342
343
344
345
346
    else let x = prot_repr x and y = prot_repr y in if x == y then ()
    else if (smallhash x != smallhash y) then raise Unify 
    else if (x.t != None) && (y.t != None) then raise Unify
      (* x and y have been internalized; if they were equivalent,
	 they would be equal *)
347
    else match x.desc,y.desc with
348
      | IType (tx,_), IType (ty,_) when Types.equal tx ty -> link x y
349
350
351
352
353
      | IOr (x1,x2), IOr (y1,y2)
      | IAnd (x1,x2), IAnd (y1,y2)
      | IDiff (x1,x2), IDiff (y1,y2)
      | ITimes (x1,x2), ITimes (y1,y2)
      | IXml (x1,x2), IXml (y1,y2)
354
355
      | IArrow (x1,x2), IArrow (y1,y2) -> link x y; unify x1 y1; unify x2 y2
      | IOptional x1, IOptional y1 -> link x y; unify x1 y1
356
357
358
359
360
361
362
363
364
365
366
      | IRecord (xo,xr), IRecord (yo,yr) when xo == yo ->
	  link x y; LabelMap.may_collide unify_field Unify xr yr
      | ICapture xv, ICapture yv when Id.equal xv yv -> ()
      | IConstant (xv,xc), IConstant (yv,yc) when
	  Id.equal xv yv && Types.Const.equal xc yc -> ()
      | _ -> raise Unify
  and unify_field f1 f2 = match f1,f2 with
    | (p1, Some e1), (p2, Some e2) -> unify p1 p2; unify e1 e2
    | (p1, None), (p2, None) -> unify p1 p2
    | _ -> raise Unify

367

368
369
  let may_unify x y =
    try unify x y; back := []; true
370
    with Unify ->
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
      List.iter (fun (x,xd) -> x.desc <- xd) !back; back := []; false

  module SmallHash = Hashtbl.Make(
    struct 
      type t = node
      let equal = may_unify
      let hash = smallhash
    end
  )

  let iter_field f = function
    | (x, Some y) -> f x; f y
    | (x, None) -> f x
  let iter f = function
    | IOr (x,y) | IAnd (x,y) | IDiff (x,y)
    | ITimes (x,y) | IXml (x,y) | IArrow (x,y) -> f x; f y
    | IOptional x -> f x
    | IRecord (_,r) -> LabelMap.iter (iter_field f) r
    | _ -> ()

  let minimize ((mem,add) as h) =
    let rec aux n =
      let n = repr n in
394
395
396
397
      if mem n then () else (
	let n = repr n in add n (); 
	if n.t == None then iter aux n.desc
      )
398
399
400
401
402
403
404
    in aux

  let to_clear = ref []
  let sid = ref 0
  let rec rechash n =
    let n = repr n in
    if (n.sid != 0) then 17 * n.sid
405
    else (incr sid; n.sid <- !sid; to_clear := n :: !to_clear; hash rechash n)
406
407

  let clear () =
408
409
    sid := 0; List.iter (fun x -> x.sid <- 0) !to_clear;
    to_clear := []
410
411
412
413
414
415
416
417
418
419
420
421
422
423

  let rechash n =
    let n = repr n in
    if (n.rechash != 0) then n.rechash 
    else (let h = rechash n in clear (); n.rechash <- h; h)

  module RecHash = Hashtbl.Make(
    struct
      type t = node
      let equal = may_unify
      let hash = smallhash
    end
  )

424
425
426

(** Two-phases recursive hash-consing **)
(*
427
428
429
  let gtable = RecHash.create 17577

  let internalize n =
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
    let local = SmallHash.create 17 in
    minimize (SmallHash.mem local, SmallHash.add local) n; 
    minimize (RecHash.mem gtable, RecHash.add gtable) n;
    ()
*)

(** Single-phase hash-consing **)
  let gtable = SmallHash.create 17

  let internalize n =
    minimize (SmallHash.mem gtable, SmallHash.add gtable) n



(*  let internalize n = () *)
445
446
447
448
449
450
451
452
453

(* Compute free variables *)

  let fv n =
    let fv = ref IdSet.empty in
    let rec aux n =
      let n = repr n in
      if (n.sid = 0) then (
	n.sid <- 1;
454
	to_clear := n :: !to_clear; 
455
456
457
458
459
460
	match n.fv, n.desc with
	  | Some x, _ -> fv := IdSet.cup !fv x
	  | None, (ICapture x | IConstant (x,_)) -> fv := IdSet.add x !fv
	  | None, d -> iter aux d
      )
    in
461
    assert(!to_clear == []);
462
463
464
465
    match n.fv with
      | Some x -> x
      | None -> aux n; clear (); n.fv <- Some !fv; !fv

466
467
468
(* optimized version to check closedness *)

  let no_fv = Some IdSet.empty
469
470
471
  exception FoundFv of id
  let peek_fv n =
    let err x = raise (FoundFv x) in
472
473
474
475
476
477
    let rec aux n =
      let n = repr n in
      if (n.sid = 0) then (
	n.sid <- 1;
	to_clear := n :: !to_clear; 
	match n.fv, n.desc with
478
479
	  | Some x, _ when IdSet.is_empty x -> ()
	  | Some x, _ -> err (IdSet.choose x)
480
481
482
483
	  | None, (ICapture x | IConstant (x,_)) -> err x;
	  | None, d -> iter aux d
      )
    in
484
    assert(!to_clear == []);
485
486
    try
      match n.fv with
487
488
	| Some x when IdSet.is_empty x -> ()
	| Some x -> err (IdSet.choose x)
489
490
491
492
493
	| None -> aux n; 
	    List.iter (fun n -> n.sid <- 0; n.fv <- no_fv) !to_clear;
	    to_clear := []
    with exn -> clear (); raise exn

494
495
496
497
498
499
500
501
502
503
504
  let check_no_fv loc n =
    try peek_fv n 
    with FoundFv x ->
      raise_loc_generic loc 
	("Capture variable not allowed: " ^ (Ident.to_string x))

  let has_no_fv n =
    try peek_fv n; true
    with FoundFv _ -> false


505
(* From the intermediate representation to the internal one *)
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521


  let rec typ n =
    let n = repr n in
    match n.t with
      | Some t -> t
      | None -> let t = compute_typ n.desc in n.t <- Some t; t
  and compute_typ = function
    | IType (t,_) -> t
    | IOr (s1,s2) -> Types.cup (typ s1) (typ s2)
    | IAnd (s1,s2) ->  Types.cap (typ s1) (typ s2)
    | IDiff (s1,s2) -> Types.diff (typ s1) (typ s2)
    | ITimes (s1,s2) -> Types.times (typ_node s1) (typ_node s2)
    | IXml (s1,s2) -> Types.xml (typ_node s1) (typ_node s2)
    | IArrow (s1,s2) -> Types.arrow (typ_node s1) (typ_node s2)
    | IOptional s -> Types.Record.or_absent (typ s)
522
    | IRecord (o,r) ->  Types.record_fields (o, LabelMap.map compute_typ_field r)
523
524
    | ILink _ -> assert false
    | ICapture _ | IConstant (_,_) -> assert false
525
    | IConcat _ | IMerge _ -> assert false
526
527
528
529
530
531
  and compute_typ_field = function
    | (s, None) -> typ_node s
    | (s, Some _) -> 
	raise (Patterns.Error "Or-else clauses are not allowed in types")

  and typ_node n =
532
    let n = repr n in
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
    match n.tnode with
      | Some t -> t
      | None ->
	  let x = Types.make () in
	  n.tnode <- Some x;
	  Types.define x (typ n);
	  x
      
  let rec pat n =
    let n = repr n in
    if IdSet.is_empty (fv n)
    then Patterns.constr (typ n)
    else match n.p with
      | Some p -> p
      | None -> let p = compute_pat n.desc in n.p <- Some p; p

  and compute_pat = function
    | IOr (s1,s2) -> Patterns.cup (pat s1) (pat s2)
    | IAnd (s1,s2) -> Patterns.cap (pat s1) (pat s2)
    | IDiff (s1,s2) when IdSet.is_empty (fv s2) ->
	let s2 = Types.neg (typ s2) in
	Patterns.cap (pat s1) (Patterns.constr s2)
    | IDiff _ ->
	raise (Patterns.Error "Differences are not allowed in patterns")
    | ITimes (s1,s2) -> Patterns.times (pat_node s1) (pat_node s2)
    | IXml (s1,s2) -> Patterns.xml (pat_node s1) (pat_node s2)
    | IOptional _ -> 
	raise (Patterns.Error "Optional fields are not allowed in record patterns")
    | IRecord (o,r) ->
	let pats = ref [] in
	let aux l = function
	  | (s,None) ->
	      if IdSet.is_empty (fv s) then typ_node s
	      else
		( pats := Patterns.record l (pat_node s) :: !pats;
		  Types.any_node )
	  | (s,Some e) ->
	      if IdSet.is_empty (fv s) then
		raise (Patterns.Error "Or-else clauses are not allowed in types")
	      else
		( pats := Patterns.cup 
		    (Patterns.record l (pat_node s))
		    (pat e) :: !pats;
		  Types.Record.any_or_absent_node )
	in
578
	let constr = Types.record_fields (o,LabelMap.mapi aux r) in
579
580
581
582
583
584
	List.fold_left Patterns.cap (Patterns.constr constr) !pats
	  (* TODO: can avoid constr when o=true, and all fields have fv *)
    | ICapture x -> Patterns.capture x
    | IConstant (x,c) -> Patterns.constant x c
    | IArrow _ ->
	raise (Patterns.Error "Arrows are not allowed in patterns")
585
    | IType _ | ILink _ | IConcat _ | IMerge _ -> assert false
586
587
      
  and pat_node n =
588
    let n = repr n in
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
    match n.pnode with
      | Some p -> p
      | None ->
	  let x = Patterns.make (fv n) in
	  try
	    n.pnode <- Some x;
	    Patterns.define x (pat n);
	    x
	  with exn -> n.pnode <- None; raise exn

(* From AST to the intermediate representation *)

  type penv = {
    penv_tenv : t;
    penv_derec : node Env.t;
  }

  let penv tenv = { penv_tenv = tenv; penv_derec = Env.empty }

608
609
  let concats = ref []

610
611
612
613
614
615
  let mk d = { node_temp with desc = d }
  let mk_delayed () = { node_temp with desc = ILink node_temp }
  let itype t = mk (IType (t, Types.hash t))
  let iempty = itype Types.empty

  let ior p1 p2 =
616
617
    if p1.desc == iempty.desc then p2 
    else if p2.desc == iempty.desc then p1 
618
619
620
    else mk (IOr (p1,p2))

  let iand p1 p2 =
621
    if (p1.desc == iempty.desc) || (p2.desc == iempty.desc) then iempty 
622
623
    else mk (IAnd (p1,p2))

624
625
626
627
628
  let times x y = mk (ITimes (x,y))
  let xml x y = mk (IXml (x,y))
  let record o m = mk (IRecord (o,m))
  let optional x = mk (IOptional x)

629
630
631
  type regexp =
    | PElem of node
    | PGuard of node
632
633
    | PSeq of regexp list
    | PAlt of regexp list
634
635
636
    | PStar of regexp
    | PWeakStar of regexp

637
638
639
640
641
642
643
644
  let rec nullable = function
    | PElem _ -> false
    | PSeq rl -> List.for_all nullable rl
    | PAlt rl -> List.exists nullable rl
    | PStar _ | PWeakStar _ | PGuard _ -> true

  let eps = PSeq []
  let emp = PAlt []
645
646
  let star x = PStar x
  let elem x = PElem x
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662

  let seq r1 r2 =
    let r1 = match r1 with PSeq l -> l | x -> [ x ] in
    let r2 = match r2 with PSeq l -> l | x -> [ x ] in
    match r1 @ r2 with
      | [ x ] -> x
      | l -> PSeq l

  let alt r1 r2 =
    let r1 = match r1 with PAlt l -> l | x -> [ x ] in
    let r2 = match r2 with PAlt l -> l | x -> [ x ] in
    match r1 @ r2 with
      | [ x ] -> x
      | l -> PAlt l

  let rec merge_alt = function
663
    | PElem p::PElem q::l -> merge_alt (PElem (ior p q) :: l)
664
665
    | r::l -> r::(merge_alt l)
    | [] -> []
666
667
668
669
670
671
672
673
674

(* Works only for types, not patterns, because
   [ (x&Int|_) R' ] is possible *)
  let rec simplify_regexp = function
    | PSeq l -> PSeq (List.map simplify_regexp l)
    | PAlt l -> PAlt (merge_alt (List.map simplify_regexp l))
    | PStar r | PWeakStar r -> PStar (simplify_regexp r)
    | x -> x

675
676
677
678
679
680
681
682
683
  let rec print_regexp ppf = function
    | PElem _ -> Format.fprintf ppf "Elem"
    | PGuard _ -> Format.fprintf ppf "Guard"
    | PSeq l -> Format.fprintf ppf "Seq(%a)" print_regexp_list l
    | PAlt l -> Format.fprintf ppf "Alt(%a)" print_regexp_list l
    | PStar r -> Format.fprintf ppf "Star(%a)" print_regexp r
    | PWeakStar r -> Format.fprintf ppf "WStar(%a)" print_regexp r
  and print_regexp_list ppf l =
    List.iter (fun x -> Format.fprintf ppf "%a;" print_regexp x) l
684

685
686
  let rec remove_regexp r q = 
    match r with
687
688
689
690
    | PElem p ->
	mk (ITimes (p, q))
    | PGuard p ->
	iand p q
691
692
693
694
    | PSeq l ->
	List.fold_right (fun r a -> remove_regexp r a) l q
    | PAlt rl ->
	List.fold_left (fun a r -> ior a (remove_regexp r q)) iempty rl
695
696
697
    | PStar r ->
	let x = mk_delayed () in
	let res = ior x q in
698
	x.desc <- ILink (remove_regexp_nullable r res iempty);
699
700
701
702
	res
    | PWeakStar r ->
	let x = mk_delayed () in
	let res = ior q x in
703
	x.desc <- ILink (remove_regexp_nullable r res iempty);
704
	res
705
706
707
708
709

  and remove_regexp_nullable r q_nonempty q_empty =
    if nullable r then remove_regexp2 r q_nonempty q_empty
    else remove_regexp r q_nonempty

710
  and remove_regexp2 r q_nonempty q_empty =
711
712
    (* Assume r is nullable *)
    if q_nonempty == q_empty then remove_regexp r q_nonempty
713
    else match r with
714
      | PSeq [] ->
715
716
          q_empty
      | PElem p ->
717
	  assert false
718
719
      | PGuard p ->
	  iand p q_empty
720
721
722
723
724
725
726
727
      | PSeq (r::rl) ->
          remove_regexp2 r
            (remove_regexp (PSeq rl) q_nonempty)
            (remove_regexp2 (PSeq rl) q_nonempty q_empty)
      | PAlt rl ->
	  List.fold_left 
	    (fun a r -> ior a (remove_regexp_nullable r q_nonempty q_empty))
	    iempty rl
728
729
      | PStar r ->
 	  let x = mk_delayed () in
730
          x.desc <- ILink (remove_regexp_nullable r (ior x q_nonempty) iempty);
731
732
733
          ior x q_empty
      | PWeakStar r ->
 	  let x = mk_delayed () in
734
          x.desc <- ILink (remove_regexp_nullable r (ior q_nonempty x) iempty);
735
736
737
          ior q_empty x


738
739
740
741
742
743
744
745
746
747
748
749
750
  let pcdata = star (PElem (itype (Types.char Chars.any)))
  let mix regexp =
    let rec aux = function
      | PSeq [] -> eps
      | PElem re -> PElem re
      | PGuard re -> assert false
      | PSeq (r::rl) -> seq (aux r) (seq pcdata (aux (PSeq rl)))
      | PAlt rl -> PAlt (List.map aux rl)
      | PStar re -> star (seq pcdata (aux re))
      | PWeakStar re -> assert false
    in
    seq pcdata (seq (aux regexp) pcdata)

751
752
753
754
755
756
  let cst_nil = Types.Atom Sequence.nil_atom
  let capture_all vars p = 
    IdSet.fold (fun p x -> iand p (mk (ICapture x))) p vars
  let termin b vars p = 
    if b then p 
    else IdSet.fold 
757
      (fun p x -> seq p (PGuard (mk (IConstant (x,cst_nil))))) p vars
758
759
760

  let rexp r = remove_regexp r (itype Sequence.nil_type)

761
762
  let all_delayed = ref []

763
764
  let clean_on_err () = all_delayed := []

765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
  let delayed loc =
    let s = mk_delayed () in
    all_delayed := (loc,s) :: !all_delayed;
    s

  let check_one_delayed (loc,p) =
    let rec aux q = if p == q then raise Exit; aux2 q.desc
    and aux2 = function
      | IOr (q1,q2) | IAnd (q1,q2) | IDiff (q1,q2) -> aux q1; aux q2
      | ILink q -> aux q
      | _ -> ()
    in
    try aux2 p.desc
    with Exit -> error loc "Ill-formed recursion"
    
  let check_delayed () =
    let l = !all_delayed in
    all_delayed := []; 
    List.iter check_one_delayed l
784

785
    
786
  let rec derecurs env p = match p.descr with
787
    | PatVar (cu,v) -> derecurs_var env p.loc cu v
788
    | Recurs (p,b) -> derecurs (fst (derecurs_def env b)) p
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
    | Internal t -> itype t
    | NsT ns -> 
	itype (Types.atom (Atoms.any_in_ns (parse_ns env.penv_tenv p.loc ns)))
    | Or (p1,p2) -> mk (IOr (derecurs env p1, derecurs env p2))
    | And (p1,p2) -> mk (IAnd (derecurs env p1, derecurs env p2))
    | Diff (p1,p2) -> mk (IDiff (derecurs env p1, derecurs env p2))
    | Prod (p1,p2) -> mk (ITimes (derecurs env p1, derecurs env p2))
    | XmlT (p1,p2) -> mk (IXml (derecurs env p1, derecurs env p2))
    | Arrow (p1,p2) -> mk (IArrow (derecurs env p1, derecurs env p2))
    | Optional p -> mk (IOptional (derecurs env p))
    | Record (o,r) -> 
	let aux = function
	  | (p,Some e) -> (derecurs env p, Some (derecurs env e))
	  | (p,None) -> derecurs env p, None in
	mk (IRecord (o, parse_record env.penv_tenv p.loc aux r))
804
805
    | Constant (x,c) -> mk (IConstant (ident env.penv_tenv p.loc x,
				       const env.penv_tenv p.loc c))
806
807
808
809
    | Cst c -> itype (Types.constant (const env.penv_tenv p.loc c))
    | Regexp r ->
	let r,_ = derecurs_regexp IdSet.empty false IdSet.empty true env r in
	rexp r
810
811
812
813
    | Concat (p1,p2) -> 
	let n = mk (IConcat (derecurs env p1, derecurs env p2)) in
	concats := n :: !concats;
	n
814
815
816
817
    | Merge (p1,p2) -> 
	let n = mk (IMerge (derecurs env p1, derecurs env p2)) in
	concats := n :: !concats;
	n
818
819
820
821
822
823
824
825
826
827
828
829
	  
  and derecurs_regexp vars b rvars f env = function
      (* - vars: seq variables to be propagated top-down and added
	 to each captured element
	 - b: below a star ?
	 - rvars: seq variables that appear on the right of the regexp
	 - f: tail position
	 
	 returns the set of seq variable of the regexp minus rvars
	 (they have already been terminated if not below a star)
      *)
    | Epsilon -> 
830
	PSeq [], IdSet.empty
831
832
833
834
835
836
837
    | Elem p -> 
	PElem (capture_all vars (derecurs env p)), IdSet.empty
    | Guard p ->
	PGuard (derecurs env p), IdSet.empty
    | Seq (p1,p2) -> 
	let (p2,v2) = derecurs_regexp vars b rvars f env p2 in
	let (p1,v1) = derecurs_regexp vars b (IdSet.cup rvars v2) false env p1 in
838
	seq p1 p2, IdSet.cup v1 v2
839
840
841
    | Alt (p1,p2) -> 
	let (p1,v1) = derecurs_regexp vars b rvars f env p1
	and (p2,v2) = derecurs_regexp vars b rvars f env p2 in
842
	alt (termin b (IdSet.diff v2 v1) p1) (termin b (IdSet.diff v1 v2) p2),
843
844
845
846
847
848
849
	IdSet.cup v1 v2
    | Star p -> 
	let (p,v) = derecurs_regexp vars true rvars false env p in
	termin b v (PStar p), v
    | WeakStar p -> 
	let (p,v) = derecurs_regexp vars true rvars false env p in
	termin b v (PWeakStar p), v
850
851
    | SeqCapture (loc,x,p) -> 
	let x = ident env.penv_tenv loc x in
852
853
854
855
856
	let vars = if f then vars else IdSet.add x vars in
	let after = IdSet.mem rvars x in
	let rvars = IdSet.add x rvars in
	let (p,v) = derecurs_regexp vars b rvars false env p in
	(if f 
857
	 then seq (PGuard (mk (ICapture x))) p 
858
859
860
861
	 else termin (after || b) (IdSet.singleton x) p), 
	(if after then v else IdSet.add x v)
	  
	  
862
863
864
865
866
  and derecurs_var env loc cu v =
    let v = ident env.penv_tenv loc v in
    match cu with
      | None ->
	  (try Env.find v env.penv_derec 
867
868
869
	   with Not_found -> 
	     try itype (find_type v env.penv_tenv)
	     with Not_found -> mk (ICapture v))
870
871
872
873
874
875
      | Some cu ->
	  (try itype (find_type_global loc cu v env.penv_tenv)
	   with Not_found ->
	     raise_loc_generic loc 
	       ("Unbound external type " ^ (U.get_str cu) ^ "." ^ 
		  (Ident.to_string v)))
876
877
	      
  and derecurs_def env b =
878
879
880
881
882
883
884
885
886
887
888
889
890
    let seen = ref IdSet.empty in
    let b = 
      List.map 
	(fun (loc,v,p) -> 
	   let v = ident env.penv_tenv loc v in
	   if IdSet.mem !seen v then 
	     raise_loc_generic loc
	       ("Multiple definitions for the type identifer " ^ 
		  (Ident.to_string v));
	   seen := IdSet.add v !seen;
	   (v,p,delayed loc))
	b in

891
892
893
894
    let n = 
      List.fold_left (fun env (v,p,s) -> Env.add v s env) env.penv_derec b in
    let env = { env with penv_derec = n } in
    List.iter (fun (v,p,s) -> s.desc <- ILink (derecurs env p)) b;
895
896
897
898
899
900
901
    (env, b)

  module H = Hashtbl.Make(Types)

  let rec elim_concat n =
    match n.desc with
      | IConcat (a,b) ->
902
903
	  if (n.sid > 0) 
	  then 	raise (Patterns.Error "Ill-formed concatenation loop");
904
905
	  n.sid <- 1;
	  n.desc <- ILink (elim_conc a b)
906
907
908
909
910
      | IMerge (a,b) ->
	  if (n.sid > 0) 
	  then 	raise (Patterns.Error "Ill-formed concatenation loop");
	  n.sid <- 1;
	  n.desc <- ILink (elim_merge a b)
911
      | _ -> ()
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
  and elim_merge a b =
    let get_rec t =
      let t = Types.Record.get t in
      List.map (fun (l,o,_) ->
		  o, 
		  LabelMap.map 
		    (fun (opt,x) ->
		       let x = itype x in 
		       (if opt then mk (IOptional x) else x),
		       None)
		    l) t in
    let merge (o1,l1) (o2,l2) =
      mk (IRecord (o1||o2, LabelMap.merge (fun _ x -> x) l1 l2)) in
    (* Problem: repr can loop with ill-formed recursion.
       type t = s + t where s = s | s;; *)
    match (repr a).desc, (repr b).desc with
      | IType (t1,_), IType (t2,_) -> 
	  if not (Types.subtype t1 Types.Record.any) then
	    raise 
	      (Patterns.Error 
		 "Left argument of record concatenation is not a record type");
	  if not (Types.subtype t2 Types.Record.any) then
	    raise 
	      (Patterns.Error 
		 "Right argument of record concatenation is not a record type");
	  itype (Types.Record.merge t1 t2)
      | IOr (a1,a2), _ -> ior (elim_merge a1 b) (elim_merge a2 b)
      | _, IOr (b1,b2) -> ior (elim_merge a b1) (elim_merge a b2)
      | IRecord (o1,l1), IRecord (o2,l2) -> merge (o1,l1) (o2,l2)
      | IType (t1,_), IRecord (o2,l2) ->
	  if not (Types.subtype t1 Types.Record.any) then
	    raise 
	      (Patterns.Error 
		 "Left argument of record concatenation is not a record type");
	  List.fold_left (fun accu (o1,l1) -> 
			    ior accu (merge (o1,l1) (o2,l2)))
	    iempty (get_rec t1)
      | IRecord (o1,l1), IType (t2,_) ->
	  if not (Types.subtype t2 Types.Record.any) then
	    raise 
	      (Patterns.Error 
		 "Right argument of record concatenation is not a record type");
	  List.fold_left (fun accu (o2,l2) -> 
			    ior accu (merge (o1,l1) (o2,l2)))
	    iempty (get_rec t2)
      | _ -> raise (Patterns.Error "Cannot compute record concatenation")
958
959
960
961
962
963
964
965
  and elim_conc n q =
    let mem = ref [] in
    let rec aux n =
      try List.assq n !mem
      with Not_found ->
	let r = mk_delayed () in
	mem := (n,r) :: !mem;
	let rec aux2 n =
966
	  match n.desc with
967
968
969
970
971
	    | ILink n' -> aux2 n'
	    | IOr (a,b) -> ior (aux a) (aux b)
	    | ITimes (a,b) -> mk (ITimes (a, aux b))
	    | IConcat (a,b) -> elim_concat n; aux2 n
	    | IType (t,_) -> elim_concat_type t q
972
	    | _ -> raise (Patterns.Error "Cannot compute concatenation")
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
	in
	r.desc <- ILink (aux2 n);
	r
    in
    aux n
  and elim_concat_type t q =
    if not (Types.subtype t Sequence.any) then
      raise (Patterns.Error "Left argument of concatenation is not a sequence type");
    (* TODO: check t <= [ Any* ] *)
    let mem = H.create 17 in
    let rec aux t =
      try H.find mem t 
      with Not_found ->
	let n = mk_delayed () in
	H.add mem t n;
	let d = 
	  List.fold_left
	    (fun accu (t1,t2) -> ior accu (mk (ITimes (itype t1, aux t2))))
	    (if Types.Atom.has_atom t Sequence.nil_atom then q else iempty)
	    (Types.Product.get t) in
	n.desc <- d.desc;
	n
    in
    aux t
    
   
    
  let elim_concats () =
    try
      List.iter elim_concat !concats;
      List.iter (fun n -> n.sid <- 0) !concats;
      concats := []
    with exn ->
      List.iter (fun n -> n.sid <- 0) !concats;
      concats := [];
      raise exn
1009

1010
1011
  let derec penv p =
    let d = derecurs penv p in
1012
    elim_concats ();
1013
1014
1015
    check_delayed ();
    internalize d;
    d
1016
1017


1018
(* API *)
1019
1020
1021

  module Ids = Set.Make(Id)
  let type_defs env b =
1022
1023
1024
1025
1026
1027
    let _,b' = derecurs_def (penv env) b in
    elim_concats ();
    check_delayed ();
    let aux loc d =
      internalize d;
      check_no_fv loc d;
1028
      try typ d
1029
      with Patterns.Error s -> raise_loc_generic loc s
1030
    in
1031
    let b = 
1032
1033
1034
      List.map2 
	(fun (loc,v,p) (v',_,d) ->
	   let t = aux loc d in
1035
1036
1037
1038
	   if (loc <> noloc) && (Types.is_empty t) then
	     warning loc 
	       ("This definition yields an empty type for " ^ (U.to_string v));
	   let v = ident env loc v in
1039
	   (v',t)) b b' in
1040
1041
    List.iter (fun (v,t) -> Types.Print.register_global 
		 (Types.CompUnit.get_current ()) (Id.value v) t) b;
1042
    enter_types b env
1043

1044
1045
1046
1047
  let type_defs env b =
    try type_defs env b
    with exn -> clean_on_err (); raise exn

1048

1049
1050
  let get_type d =
    check_delayed ();
1051
1052
    try internalize d; typ d
    with exn -> clean_on_err (); raise exn
1053

1054
  let typ env t = 
1055
1056
1057
1058
1059
1060
    try
      let d = derec (penv env) t in
      check_no_fv t.loc d;
      try typ_node d
      with Patterns.Error s -> raise_loc_generic t.loc s
    with exn -> clean_on_err (); raise exn
1061
1062

  let pat env t = 
1063
1064
1065
1066
1067
    try
      let d = derec (penv env) t in
      try pat_node d
      with Patterns.Error s -> raise_loc_generic t.loc s
    with exn -> clean_on_err (); raise exn
1068
1069
1070
1071
1072
1073

  let delayed () = delayed noloc
  let link a b = a.desc <- ILink b

  let get_ct c =
    match c.desc with
1074
      | ITimes ({ desc = IType (t,_) },content) -> (t,content)
1075
1076
1077
      | _ -> assert false


1078
end
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094

let typ = IType.typ
let pat = IType.pat
let type_defs = IType.type_defs

let dump_types ppf env =
  Env.iter (fun v -> 
	      function 
		  (Type _) -> Format.fprintf ppf " %a" Ident.print v
		| _ -> ()) env.ids

let dump_ns ppf env =
  Ns.dump_table ppf env.ns



1095

1096
1097
(* II. Build skeleton *)

1098

1099
type type_fun = Types.t -> bool -> Types.t
1100

1101
module Fv = IdSet
1102

1103
1104
1105
type branch = Branch of Typed.branch * branch list

let cur_branch : branch list ref = ref []
1106

1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
let exp' loc e = 
  { Typed.exp_loc = loc; Typed.exp_typ = Types.empty; Typed.exp_descr = e; }

let exp loc fv e = fv, exp' loc e

let exp_nil = exp' noloc (Typed.Cst Sequence.nil_cst)

let pat_true = 
  let n = Patterns.make Fv.empty in
  Patterns.define n (Patterns.constr Builtin_defs.true_type);
  n

let pat_false =   
  let n = Patterns.make Fv.empty in
  Patterns.define n (Patterns.constr Builtin_defs.false_type);
  n

1124

1125
let ops = Hashtbl.create 13
1126
1127
let register_op op arity f = Hashtbl.add ops op (arity,f)
let typ_op op = snd (Hashtbl.find ops op)
1128

1129
1130
1131
1132
1133
let fun_name env a =
  match a.fun_name with
    | None -> None
    | Some (loc,s) -> Some (ident env loc s)

1134
let is_op env s = 
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
  if (Env.mem s env.ids) then None
  else
    let (ns,s) = Id.value s in
    if Ns.equal ns Ns.empty then
      let s = U.get_str s in
      try 
	let o = Hashtbl.find ops s in
	Some (s, fst o)
      with Not_found -> None
    else None
1145

1146
1147
let rec expr env loc = function
  | LocatedExpr (loc,e) -> expr env loc e
1148
  | Forget (e,t) ->
1149
      let (fv,e) = expr env loc e and t = typ env t in
1150
      exp loc fv (Typed.Forget (e,t))
1151
1152
  | Check (e,t) ->
      let (fv,e) = expr env loc e and t = typ env t in
1153
      exp loc fv (Typed.Check (ref Types.empty,e,t))