typer.ml 55 KB
Newer Older
1
(* TODO:
2
 - rewrite type-checking of operators to propagate constraint
3
 - check whether it is worth using recursive hash-consing internally
4
5
*)

6
7
8
open Location
open Ast
open Ident
9

10
11
12
13
14
15
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

16
17
let debug_schema = false

18
let warning loc msg =
19
  let v = Location.get_viewport () in
20
  let ppf = if Html.is_html v then Html.ppf v else Format.err_formatter in
21
22
23
  Format.fprintf ppf "Warning %a:@\n" Location.print_loc (loc,`Full);
  Location.html_hilight (loc,`Full);
  Format.fprintf ppf "%s@." msg
24

25
26
27
28
29
30
31
32
33
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

34
35
36

exception Warning of string * Types.t

37
38
39
40
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)

41
42
type item =
  | Type of Types.t
43
  | Val of Types.t
44

45
46
module UEnv = Map.Make(U)

47
type t = {
48
  ids : item Env.t;
49
  ns: Ns.table;
50
  cu: Types.CompUnit.t UEnv.t;
51
  schemas: string UEnv.t
52
}
53

54
55
56
57
58
let hash _ = failwith "Typer.hash"
let compare _ _ = failwith "Typer.compare"
let dump ppf _ = failwith "Typer.dump"
let equal _ _ = failwith "Typer.equal"
let check _ = failwith "Typer.check"
59

60
61
62
63
64
65
66
67

let load_schema_fwd = ref (fun x uri -> assert false)

let enter_schema x uri env =
  !load_schema_fwd x uri;
  { env with schemas = UEnv.add x uri env.schemas }


68
(* TODO: filter out builtin defs ? *)
69
70
71
72
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

73
let serialize s env =
74
  Serialize.Put.env Id.serialize serialize_item Env.iter s env.ids;
75
76
77
78
79
  Ns.serialize_table s env.ns;

  let schs =
    UEnv.fold (fun name uri accu -> (name,uri)::accu) env.schemas [] in
  Serialize.Put.list (Serialize.Put.pair U.serialize Serialize.Put.string) s schs
80

81
82
83
84
85
let deserialize_item s = match Serialize.Get.bits 1 s with
  | 0 -> Type (Types.deserialize s)
  | 1 -> Val (Types.deserialize s)
  | _ -> assert false

86
let deserialize s =
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
93
94
  let schs = 
    Serialize.Get.list 
      (Serialize.Get.pair U.deserialize Serialize.Get.string) s in
  let env = 
    { ids = ids; ns = ns; cu = UEnv.empty; schemas = UEnv.empty } in
  List.fold_left (fun env (name,uri) -> enter_schema name uri env) env schs
95
96


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

104
105
let from_comp_unit = ref (fun cu -> assert false)

106
let enter_cu x cu env =
107
  { env with cu = UEnv.add x cu env.cu }
108

109
110
111
let find_cu x env =
  try UEnv.find x env.cu
  with Not_found -> Types.CompUnit.mk x
112
113


114
115
116
117
let find_schema x env =
  try UEnv.find x env.schemas
  with Not_found -> raise (Error (Printf.sprintf "%s: no such schema" (U.get_str x)))

118
119
120
121
122
123
124
125
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
126
    | Val _ -> raise Not_found
127

128
let find_type_global loc cu id env =
129
  let cu = find_cu cu env in
130
131
132
  let env = !from_comp_unit cu in
  find_type id env

133
let enter_value id t env = 
134
  { env with ids = Env.add id (Val t) env.ids }
135
136
let enter_values l env =
  { env with ids = 
137
      List.fold_left (fun accu (id,t) -> Env.add id (Val t) accu) env.ids l }
138
139
140
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 }
141
142
let find_value id env =
  match Env.find id env.ids with
143
    | Val t -> t
144
    | _ -> raise Not_found
145
146
147
let find_value_global cu id env =
  let env = !from_comp_unit cu in
  find_value id env
148
	
149
150
151
152
153
154
let value_name_ok id env =
  try match Env.find id env.ids with
    | Val t -> true
    | _ -> false
  with Not_found -> true

155
156
157
158
let iter_values env f =
  Env.iter (fun x ->
	      function Val t -> f x t;
		| _ -> ()) env.ids
159

160

161
162
163
164
165
166
167
168
169
let register_types cu env =
  let prefix = U.concat (Types.CompUnit.value cu) (U.mk ":") in
  Env.iter (fun x ->
	      function 
		| Type t ->
		    let n = U.concat prefix (Id.value x) in
		    Types.Print.register_global n t
		| _ -> ()) env.ids

170

171
(* Namespaces *)
172

173
let set_ns_table_for_printer env = 
174
  Ns.InternalPrinter.set_table env.ns
175

176
let get_ns_table tenv = tenv.ns
177

178
let enter_ns p ns env =
179
  { env with ns = Ns.add_prefix p ns env.ns }
180

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
let parse_atom env loc t =
191
  Atoms.V.of_qname (qname env loc t)
192
193
 
let parse_ns env loc ns =
194
  protect_error_ns loc (Ns.map_prefix env.ns) ns
195

196
let parse_label env loc t =
197
  let (ns,l) = protect_error_ns loc (Ns.map_attr env.ns) t in
198
  LabelPool.mk (ns,l)
199

200
201
202
203
204
205
206
207
208
209
210
211
212
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
213
  | Const c -> c
214
215
216
217
  | _ -> 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 *)
218

219

220
(* Schema *)
221

222
223
224
let is_registered_schema env s = UEnv.mem s env.schemas

(* uri -> schema binding *)
225
let schemas = State.ref "Typer.schemas" (Hashtbl.create 3)
226
227
228

let schema_types = State.ref "Typer.schema_types" (Hashtbl.create 51)
let schema_elements = State.ref "Typer.schema_elements" (Hashtbl.create 51)
229
let schema_attributes = State.ref "Typer.schema_attributes" (Hashtbl.create 51)
230
231
232
233
let schema_attribute_groups =
  State.ref "Typer.schema_attribute_groups" (Hashtbl.create 51)
let schema_model_groups =
  State.ref "Typer.schema_model_groups" (Hashtbl.create 51)
234

235

236
237
238
239
240
(*
let get_schema uri =
  try Hashtbl.find !schemas uri
  with Not_found -> assert false
*)
241

242
let find_schema_descr_uri kind uri (name : Ns.qname) =
243
  try
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
    let elt () = Hashtbl.find !schema_elements (uri, name) in
    let typ () = Hashtbl.find !schema_types (uri, name) in
    let att () = Hashtbl.find !schema_attributes (uri, name) in
    let att_group () = Hashtbl.find !schema_attribute_groups (uri, name) in
    let mod_group () = Hashtbl.find !schema_model_groups (uri, name) in
    let rec do_try n = function
      | [] -> raise Not_found
      | f :: rem -> (try f () with Not_found -> do_try n rem)
    in
    match kind with
      | Some `Element -> do_try "element" [ elt ]
      | Some `Type -> do_try "type" [ typ ]
      | Some `Attribute -> do_try "atttribute" [ att ]
      | Some `Attribute_group -> do_try "attribute group" [ att_group ]
      | Some `Model_group -> do_try "model group" [ mod_group ]
      | None ->
          (* policy for unqualified schema component resolution. This order should
           * be consistent with Schema_component.get_component *)
          do_try "component" [ elt; typ; att; att_group; mod_group ]
    with Not_found ->    
264
      raise (Error (Printf.sprintf "No %s named '%s' found in schema '%s'"
265
		      (Schema_common.string_of_component_kind kind) (Ns.QName.to_string name) uri))
266
267
268
269
270

let find_schema_descr env kind schema name =
  let uri = find_schema schema env in
  find_schema_descr_uri kind uri name

271

272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
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

  let rec node_temp = { 
    desc = ILink node_temp;
    smallhash = 0; rechash = 0; sid = 0;
    t = None; tnode = None; p = None; pnode = None;
    fv = None
  }
			
305
(* Recursive hash-consing *)
306

307
308
309
310
311
312
  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
313
    | IType (t,h) -> 1 + 17 * h
314
315
316
317
318
319
320
321
322
    | 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)
323
324
325
    | ICapture x -> 10 + 17 * (Id.hash x)
    | IConstant (x,c) -> 11 + 17 * (Id.hash x) + 257*(Types.Const.hash c)

326
327
328
329
330
  let hash0 = hash (fun n -> 1)
  let hash1 = hash hash0
  let hash2 = hash hash1
  let hash3 = hash hash2

331
332
  let smallhash n =
    if n.smallhash !=0 then n.smallhash
333
334
335
336
    else (
      let h = hash2 n in 
      n.smallhash <- h; h
    )
337
338

  let rec repr = function
339
    | { desc = ILink n } as m -> let z = repr n in m.desc <- ILink z; z
340
341
342
343
    | n -> n

  let back = ref []

344
345
346
347
  let rec prot_repr = function
    | { desc = ILink n } -> repr n
    | n -> n

348
349
350
351
352
353
354
355
356
  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 ()
357
358
359
360
361
    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 *)
362
    else match x.desc,y.desc with
363
      | IType (tx,_), IType (ty,_) when Types.equal tx ty -> link x y
364
365
366
367
368
      | 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)
369
370
      | 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
371
372
373
374
375
376
377
378
379
380
381
      | 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

382

383
384
  let may_unify x y =
    try unify x y; back := []; true
385
    with Unify ->
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
      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
409
410
411
412
      if mem n then () else (
	let n = repr n in add n (); 
	if n.t == None then iter aux n.desc
      )
413
414
415
416
417
418
419
    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
420
    else (incr sid; n.sid <- !sid; to_clear := n :: !to_clear; hash rechash n)
421
422

  let clear () =
423
424
    sid := 0; List.iter (fun x -> x.sid <- 0) !to_clear;
    to_clear := []
425
426
427
428
429
430
431
432
433
434
435
436
437
438

  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
  )

439
440
441

(** Two-phases recursive hash-consing **)
(*
442
443
444
  let gtable = RecHash.create 17577

  let internalize n =
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
    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 = () *)
460
461
462
463
464
465
466
467
468

(* 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;
469
	to_clear := n :: !to_clear; 
470
471
472
473
474
475
	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
476
    assert(!to_clear == []);
477
478
479
480
    match n.fv with
      | Some x -> x
      | None -> aux n; clear (); n.fv <- Some !fv; !fv

481
482
483
(* optimized version to check closedness *)

  let no_fv = Some IdSet.empty
484
485
486
  exception FoundFv of id
  let peek_fv n =
    let err x = raise (FoundFv x) in
487
488
489
490
491
492
493
494
495
496
497
    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
	  | Some x, _ -> (match IdSet.pick x with Some x -> err x | None -> ())
	  | None, (ICapture x | IConstant (x,_)) -> err x;
	  | None, d -> iter aux d
      )
    in
498
    assert(!to_clear == []);
499
500
501
502
503
504
505
506
    try
      match n.fv with
	| Some x -> (match IdSet.pick x with Some x -> err x | None -> ())
	| None -> aux n; 
	    List.iter (fun n -> n.sid <- 0; n.fv <- no_fv) !to_clear;
	    to_clear := []
    with exn -> clear (); raise exn

507
508
509
510
511
512
513
514
515
516
517
  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


518
(* From the intermediate representation to the internal one *)
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543


  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)
    | IRecord (o,r) ->  Types.record' (o, LabelMap.map compute_typ_field r)
    | ILink _ -> assert false
    | ICapture _ | IConstant (_,_) -> assert false
  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 =
544
    let n = repr n in
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
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
    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
	let constr = Types.record' (o,LabelMap.mapi aux r) in
	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")
    | IType _ | ILink _ -> assert false
      
  and pat_node n =
600
    let n = repr n in
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
    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 }

  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 =
626
627
    if p1.desc == iempty.desc then p2 
    else if p2.desc == iempty.desc then p1 
628
629
630
    else mk (IOr (p1,p2))

  let iand p1 p2 =
631
    if (p1.desc == iempty.desc) || (p2.desc == iempty.desc) then iempty 
632
633
634
635
636
    else mk (IAnd (p1,p2))

  type regexp =
    | PElem of node
    | PGuard of node
637
638
    | PSeq of regexp list
    | PAlt of regexp list
639
640
641
    | PStar of regexp
    | PWeakStar of regexp

642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
  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 []

  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
666
    | PElem p::PElem q::l -> merge_alt (PElem (ior p q) :: l)
667
668
    | r::l -> r::(merge_alt l)
    | [] -> []
669
670
671
672
673
674
675
676
677

(* 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

678
679
680
681
682
683
684
685
686
  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
687

688
689
  let rec remove_regexp r q = 
    match r with
690
691
692
693
    | PElem p ->
	mk (ITimes (p, q))
    | PGuard p ->
	iand p q
694
695
696
697
    | 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
698
699
700
    | PStar r ->
	let x = mk_delayed () in
	let res = ior x q in
701
	x.desc <- ILink (remove_regexp_nullable r res iempty);
702
703
704
705
	res
    | PWeakStar r ->
	let x = mk_delayed () in
	let res = ior q x in
706
	x.desc <- ILink (remove_regexp_nullable r res iempty);
707
	res
708
709
710
711
712

  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

713
  and remove_regexp2 r q_nonempty q_empty =
714
715
    (* Assume r is nullable *)
    if q_nonempty == q_empty then remove_regexp r q_nonempty
716
    else match r with
717
      | PSeq [] ->
718
719
          q_empty
      | PElem p ->
720
	  assert false
721
722
      | PGuard p ->
	  iand p q_empty
723
724
725
726
727
728
729
730
      | 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
731
732
      | PStar r ->
 	  let x = mk_delayed () in
733
          x.desc <- ILink (remove_regexp_nullable r (ior x q_nonempty) iempty);
734
735
736
          ior x q_empty
      | PWeakStar r ->
 	  let x = mk_delayed () in
737
          x.desc <- ILink (remove_regexp_nullable r (ior q_nonempty x) iempty);
738
739
740
741
742
743
744
745
746
          ior q_empty x


  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 
747
      (fun p x -> seq p (PGuard (mk (IConstant (x,cst_nil))))) p vars
748
749
750

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

751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
  let all_delayed = ref []

  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
    
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
  let rec derecurs env p = match p.descr with
    | PatVar v -> derecurs_var env p.loc v
    | SchemaVar (kind, schema_name, component_name) ->

	let name = qname env.penv_tenv  p.loc component_name in
	itype (find_schema_descr env.penv_tenv kind schema_name name)

    | Recurs (p,b) -> derecurs (derecurs_def env b) p
    | 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))
    | Constant (x,c) -> mk (IConstant (x,const env.penv_tenv p.loc c))
    | 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
	  
  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 -> 
813
	PSeq [], IdSet.empty
814
815
816
817
818
819
820
    | 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
821
	seq p1 p2, IdSet.cup v1 v2
822
823
824
    | 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
825
	alt (termin b (IdSet.diff v2 v1) p1) (termin b (IdSet.diff v1 v2) p2),
826
827
828
829
830
831
832
833
834
835
836
837
838
	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
    | SeqCapture (x,p) -> 
	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 
839
	 then seq (PGuard (mk (ICapture x))) p 
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
	 else termin (after || b) (IdSet.singleton x) p), 
	(if after then v else IdSet.add x v)
	  
	  
  and derecurs_var env loc v =
    match Ns.split_qname v with
      | "", v ->
	  let v = ident v in
	  (try Env.find v env.penv_derec
	   with Not_found -> 
	     try itype (find_type v env.penv_tenv)
	     with Not_found -> mk (ICapture v))
      | cu, v -> 
	  try 
	    let cu = U.mk cu in
	    itype (find_type_global loc cu (ident v) env.penv_tenv)
	  with Not_found ->
	    raise_loc_generic loc 
	      ("Unbound external type " ^ cu ^ ":" ^ (U.to_string v))
	      
  and derecurs_def env b =
861
    let b = List.map (fun (v,p) -> (v,p,delayed p.loc)) b in
862
863
864
865
866
867
    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;
    env

868
869
870
871
872
  let derec penv p =
    let d = derecurs penv p in
    check_delayed ();
    internalize d;
    d
873
874


875
(* API *)
876
877
878
879
880
881
882
883
884
885
886
887
888
889

  module Ids = Set.Make(Id)
  let type_defs env b =
    ignore 
      (List.fold_left 
	 (fun seen (v,p) ->
	    if Ids.mem v seen then 
	      raise_loc_generic p.loc 
		("Multiple definitions for the type identifer " ^ 
		   (Ident.to_string v));
	    Ids.add v seen
	 ) Ids.empty b);
    
    let penv = derecurs_def (penv env) b in
890
891
892
893
894
895
    let aux t =
      let d = derec penv t in
      check_no_fv t.loc d;
      try typ d
      with Patterns.Error s -> raise_loc_generic t.loc s
    in
896
897
    let b = 
      List.map 
898
899
	(fun (v,p) ->
	   let t = aux p in
900
901
902
903
904
905
906
	   if (p.loc <> noloc) && (Types.is_empty t) then
	     warning p.loc 
	       ("This definition yields an empty type for " ^ (Ident.to_string v));
	   (v,t)) b in
    List.iter (fun (v,t) -> Types.Print.register_global (Id.value v) t) b;
    b

907

908
909
910
  let typ_descr d =
    internalize d;
    typ d
911

912
913
914
915
916
917
918
919
920
921
922
  let typ env t = 
    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

  let pat env t = 
    let d = derec (penv env) t in
    try pat_node d
    with Patterns.Error s -> raise_loc_generic t.loc s
end
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938

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



939

940
941
(* II. Build skeleton *)

942

943
type type_fun = Types.t -> bool -> Types.t
944

945
module Fv = IdSet
946

947
948
949
type branch = Branch of Typed.branch * branch list

let cur_branch : branch list ref = ref []
950

951
let exp loc fv e =
952
953
  fv,
  { Typed.exp_loc = loc;
954
    Typed.exp_typ = Types.empty;
955
    Typed.exp_descr = e;
956
  }
957

958
let ops = Hashtbl.create 13
959
960
let register_op op arity f = Hashtbl.add ops op (arity,f)
let typ_op op = snd (Hashtbl.find ops op)
961

962
963
964
965
966
let is_op env s = 
  if (Env.mem (ident s) env.ids) then None
  else 
    try let s = U.get_str s in Some (s, fst (Hashtbl.find ops s))
    with Not_found -> None
967

968
969
let rec expr env loc = function
  | LocatedExpr (loc,e) -> expr env loc e
970
  | Forget (e,t) ->
971
      let (fv,e) = expr env loc e and t = typ env t in
972
      exp loc fv (Typed.Forget (e,t))
973
974
  | Check (e,t) ->
      let (fv,e) = expr env loc e and t = typ env t in
975
      exp loc fv (Typed.Check (ref Types.empty,e,t))
976
  | Var s -> var env loc s
977
  | Apply (e1,e2) -> 
978
979
980
981
982
983
984
985
      let (fv1,e1) = expr env loc e1 and (fv2,e2) = expr env loc e2 in
      let fv = Fv.cup fv1 fv2 in
      (match e1.Typed.exp_descr with
	 | Typed.Op (op,arity,args) when arity > 0 -> 
	     exp loc fv (Typed.Op (op,arity - 1,args @ [e2]))
	 | _ ->
	     exp loc fv (Typed.Apply (e1,e2)))
  | Abstraction a -> abstraction env loc a
986
  | (Integer _ | Char _ | Atom _ | Const _) as c -> 
987
      exp loc Fv.empty (Typed.Cst (const env loc c))
988
  | Pair (e1,e2) ->
989
      let (fv1,e1) = expr env loc e1 and (fv2,e2) = expr env loc e2 in
990
991
      exp loc (Fv.cup fv1 fv2) (Typed.Pair (e1,e2))
  | Xml (e1,e2) ->
992
      let (fv1,e1) = expr env loc e1 and (fv2,e2) = expr env loc e2 in
993
994
      exp loc (Fv.cup fv1 fv2) (Typed.Xml (e1,e2))
  | Dot (e,l) ->
995
996
      let (fv,e) = expr env loc e in
      exp loc fv (Typed.Dot (e,parse_label env loc l))
997
  | RemoveField (e,l) ->
998
999
      let (fv,e) = expr env loc e in
      exp loc fv (Typed.RemoveField (e,parse_label env loc l))
1000
1001
  | RecordLitt r -> 
      let fv = ref Fv.empty in
1002
      let r = parse_record env loc
1003
		(fun e -> 
1004
		   let (fv2,e) = expr env loc e 
1005
1006
1007
		   in fv := Fv.cup !fv fv2; e)
		r in
      exp loc !fv (Typed.RecordLitt r)
1008
  | String (i,j,s,e) ->
1009
      let (fv,e) = expr env loc e in
1010
      exp loc fv (Typed.String (i,j,s,e))
1011
  | Match (e,b) -> 
1012
1013
      let (fv1,e) = expr env loc e
      and (fv2,b) = branches env b in
1014
      exp loc (Fv.cup fv1 fv2) (Typed.Match (e, b))
1015
  | Map (e,b) ->
1016
1017
      let (fv1,e) = expr env loc e
      and (fv2,b) = branches env b in
1018
1019
      exp loc (Fv.cup fv1 fv2) (Typed.Map (e, b))
  | Transform (e,b) ->
1020
1021
      let (fv1,e) = expr env loc e
      and (fv2,b) = branches env b in
1022
      exp loc (Fv.cup fv1 fv2) (Typed.Transform (e, b))
1023
  | Xtrans (e,b) ->
1024
1025
      let (fv1,e) = expr env loc e
      and (fv2,b) = branches env b in
1026
      exp loc (Fv.cup fv1 fv2) (Typed.Xtrans (e, b))
1027
  | Validate (e,kind,schema,elt) ->
1028
      let (fv,e) = expr env loc e in
1029
      let uri = find_schema schema env in
1030
      exp loc fv (Typed.Validate (e, kind, uri, qname env loc elt))
1031
  | Try (e,b) ->
1032
1033
      let (fv1,e) = expr env loc e
      and (fv2,b) = branches env b in
1034
      exp loc (Fv.cup fv1 fv2) (Typed.Try (e, b))
1035
  | NamespaceIn (pr,ns,e) ->
1036
1037
      let env = enter_ns pr ns env in
      expr env loc e
1038
  | Ref (e,t) ->
1039
      let (fv,e) = expr env loc e and t = typ env t in
1040
      exp loc fv (Typed.Ref (e,t))
1041
  | External (s,args) ->
1042
      extern loc env s args
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
	
and extern loc env s args = 
  let args = List.map (typ env) args in
  try
    let (i,t) = Externals.resolve s args in
    exp loc Fv.empty (Typed.External (t,i))
  with exn -> raise_loc loc exn
    
and var env loc s =
  match is_op env s with
1053
    | Some (s,arity) -> 
1054
1055
	let need_ns = match s with "print_xml" | "print_xml_utf8" -> true
	  | _ -> false in
1056
1057
1058
	let e = Typed.Op (s, arity, []) in
	let e = if need_ns then Typed.NsTable (env.ns,e) else e in
	exp loc Fv.empty e
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
    | None ->
	match Ns.split_qname s with
	  | "", id -> 
	      let s = U.get_str id in
	      if String.contains s '.' then
		extern loc env s []
	      else
		let id = ident id in
		(try ignore (find_value id env)
		 with Not_found -> raise_loc loc (UnboundId (id, Env.mem id env.ids)));
1069
	  exp loc (Fv.singleton id) (Typed.Var id)
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
	  | cu, id -> 
	      let cu = find_cu (U.mk cu) env in
	      let id = ident id in
	      let t =
		try find_value_global cu id env
		with Not_found ->
		  raise_loc loc (UnboundExtId (cu,id) ) in
	      exp loc Fv.empty (Typed.ExtVar (cu, id, t))

and abstraction env loc a =
  let iface = 
    List.map 
      (fun (t1,t2) -> (typ env t1, typ env t2)) a.fun_iface in
  let t = 
    List.fold_left 
      (fun accu (t1,t2) -> Types.cap accu (Types.arrow t1 t2)) 
      Types.any iface in
  let iface = 
    List.map 
      (fun (t1,t2) -> (Types.descr t1, Types.descr t2)) 
      iface in
  let env' = 
    match a.fun_name with 
      | None -> env
      | Some f -> enter_values_dummy [ f ] env
  in
  let (fv0,body) = branches env' a.fun_body in
  let fv = match a.fun_name with
    | None -> fv0
    | Some f -> Fv.remove f fv0 in
  let e = Typed.Abstraction 
	    { Typed.fun_name = a.fun_name;
	      Typed.fun_iface = iface;
	      Typed.fun_body = body;
	      Typed.fun_typ = t;
	      Typed.fun_fv = fv
	    } in
  exp loc fv e
    
and branches env b = 
  let fv = ref Fv.empty in
  let accept = ref Types.empty in
  let branch (p,e) = 
    let cur_br = !cur_branch in
    cur_branch := [];
    let p' = pat env p in
    let fvp = Patterns.fv p' in
    let env' = enter_values_dummy fvp env in
    let (fv2,e) = expr env' noloc e in
    let br_loc = merge_loc p.loc e.Typed.exp_loc in
    (match Fv.pick (Fv.diff fvp fv2) with
       | None -> ()
       | Some x ->
	   let x = U.to_string (Id.value x) in
	   warning br_loc 
	     ("The capture variable " ^ x ^ 
	      " is declared in the pattern but not used in the body of this branch. It might be a misspelled or undeclared type or name (if it isn't, use _ instead)."));
    let fv2 = Fv.diff fv2 fvp in
    fv := Fv.cup !fv fv2;
    accept := Types.cup !accept (Types.descr (Patterns.accept p'));
    let br = 
      { 
	Typed.br_loc = br_loc;
1133
	Typed.br_used = br_loc == noloc;
1134
	Typed.br_vars_empty = Patterns.fv p';
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
	Typed.br_pat = p';
	Typed.br_body = e } in
    cur_branch := Branch (br, !cur_branch) :: cur_br;
    br in
  let b = List.map branch b in
  (!fv, 
   { 
     Typed.br_typ = Types.empty; 
     Typed.br_branches = b; 
     Typed.br_accept = !accept;
     Typed.br_compiled = None;
   } 
  )
1148

1149
let expr env e = snd (expr env noloc e)
1150

1151
1152
let let_decl env p e =
  { Typed.let_pat = pat env p;
1153
    Typed.let_body = expr env e;
1154
1155
    Typed.let_compiled = None }

1156
1157
1158

(* Hide global "typing/parsing" environment *)

1159

1160
1161
(* III. Type-checks *)

1162
1163
open Typed

1164
1165
1166
1167
1168
1169
let localize loc f x =
  try f x
  with 
    | (Error _ | Constraint (_,_)) as exn -> raise (Location.Location (loc,`Full,exn))
    | Warning (s,t) -> warning loc s; t

1170
1171
let require loc t s = 
  if not (Types.subtype t s) then raise_loc loc (Constraint (t, s))
1172

1173
let verify loc t s = 
1174
1175
  require loc t s; t

1176
1177
1178
1179
let verify_noloc t s =
  if not (Types.subtype t s) then raise (Constraint (t, s));
  t

1180
1181
1182
1183
1184
let check_str loc ofs t s = 
  if not (Types.subtype t s) then raise_loc_str loc ofs (Constraint (t, s));
  t

let should_have loc constr s = 
1185
1186
  raise_loc loc (ShouldHave (constr,s))

1187
1188
1189
let should_have_str loc ofs constr s = 
  raise_loc_str loc ofs (ShouldHave (constr,s))

1190
let flatten arg constr precise =