types.ml 48.7 KB
Newer Older
1
open Ident
2
open Encodings
3

4
5
6
7
8
(* TODO:
   - I store hash in types to avoid computing it several times.
     Does not seem to help a lot.
*)

9
10
11
12
13
14
15
16
17
18
(*
To be sure not to use generic comparison ...
*)
let (=) : int -> int -> bool = (==)
let (<) : int -> int -> bool = (<)
let (<=) : int -> int -> bool = (<=)
let (<>) : int -> int -> bool = (<>)
let compare = 1


19
20
21
module CompUnit = struct   
  include Pool.Make(Utf8)
  module Tbl = Inttbl
22
      
23
  let pervasives = mk (U.mk "Pervasives") 
24
      
25
  let close_serialize_ref = ref (fun () -> assert false)
26
      
27
  let depend = Inttbl.create ()
28
      
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
  let serialize t cu =
    if cu != pervasives then Inttbl.add depend cu ();
    serialize t cu

  let close_serialize () =
    !close_serialize_ref ();
    let deps = Inttbl.fold depend (fun cu () l -> cu :: l) [] in
    Inttbl.clear depend;
    deps


  let stack = ref []
  let current = ref dummy_min
		
  let enter i = stack := !current :: !stack; current := i
		  
  let leave () =
    match !stack with
      | hd::tl -> current := hd; stack := tl
      | _ -> assert false


  let () = enter pervasives
end	  


55
type const = 
56
  | Integer of Intervals.V.t
57
  | Atom of Atoms.V.t 
58
  | Char of Chars.V.t
59
60
61
62
  | Pair of const * const
  | Xml of const * const
  | Record of const label_map
  | String of U.uindex * U.uindex * U.t * const
63

64
65
66
module Const = struct
  type t = const

67
68
69
70
71
72
73
74
75
76
77
  let rec check = function
    | Integer i -> Intervals.V.check i
    | Atom i -> Atoms.V.check i
    | Char i -> Chars.V.check i
    | Pair (x,y) | Xml (x,y) -> check x; check y
    | Record l -> LabelMap.iter check l
    | String (i,j,s,q) -> U.check s; check q

  let dump ppf _ =
    Format.fprintf ppf "<Types.Const.t>"

78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
  let rec serialize s = function
    | Integer x ->
	Serialize.Put.bits 3 s 0;
	Intervals.V.serialize s x
    | Atom x ->
	Serialize.Put.bits 3 s 1;
	Atoms.V.serialize s x
    | Char x ->
	Serialize.Put.bits 3 s 2;
	Chars.V.serialize s x
    | Pair (x,y) ->
	Serialize.Put.bits 3 s 3;
	serialize s x;
	serialize s y
    | Xml (x,y) ->
	Serialize.Put.bits 3 s 4;
	serialize s x;
	serialize s y
    | Record r ->
	Serialize.Put.bits 3 s 5;
	LabelMap.serialize serialize s r
    | String (i,j,st,q) ->
	Serialize.Put.bits 3 s 6;
	U.serialize_sub s st i j;
	serialize s q

  let rec deserialize s =
    match Serialize.Get.bits 3 s with
      | 0 ->
	  Integer (Intervals.V.deserialize s)
      | 1 ->
	  Atom (Atoms.V.deserialize s)
110
111
      | 2 ->
	  Char (Chars.V.deserialize s)
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
      | 3 ->
	  let x = deserialize s in
	  let y = deserialize s in
	  Pair (x,y)
      | 4 ->
	  let x = deserialize s in
	  let y = deserialize s in
	  Xml (x,y)
      | 5 ->
	  Record (LabelMap.deserialize deserialize s)
      | 6 ->
	  let st = U.deserialize s in
	  let q = deserialize s in
	  String (U.start_index st, U.end_index st, st, q)
      | _ ->
	  assert false

  let rec compare c1 c2 = match (c1,c2) with
130
    | Integer x, Integer y -> Intervals.V.compare x y
131
132
    | Integer _, _ -> -1
    | _, Integer _ -> 1
133
    | Atom x, Atom y -> Atoms.V.compare x y
134
135
    | Atom _, _ -> -1
    | _, Atom _ -> 1
136
    | Char x, Char y -> Chars.V.compare x y
137
138
139
    | Char _, _ -> -1
    | _, Char _ -> 1
    | Pair (x1,x2), Pair (y1,y2) ->
140
141
	let c = compare x1 y1 in
	if c <> 0 then c else compare x2 y2
142
143
144
    | Pair (_,_), _ -> -1
    | _, Pair (_,_) -> 1
    | Xml (x1,x2), Xml (y1,y2) ->
145
146
	let c = compare x1 y1 in
	if c <> 0 then c else compare x2 y2
147
148
149
    | Xml (_,_), _ -> -1
    | _, Xml (_,_) -> 1
    | Record x, Record y ->
150
	LabelMap.compare compare x y
151
152
153
154
155
156
157
    | Record _, _ -> -1
    | _, Record _ -> 1
    | String (i1,j1,s1,r1), String (i2,j2,s2,r2) ->
	let c = Pervasives.compare i1 i2 in if c <> 0 then c 
	else let c = Pervasives.compare j1 j2 in if c <> 0 then c
	else let c = U.compare s1 s2 in if c <> 0 then c (* Should compare
							    only the substring *)
158
159
160
161
162
163
164
165
166
167
	else compare r1 r2

  let rec hash = function
    | Integer x -> 1 + 17 * (Intervals.V.hash x)
    | Atom x -> 2 + 17 * (Atoms.V.hash x)
    | Char x -> 3 + 17 * (Chars.V.hash x)
    | Pair (x,y) -> 4 + 17 * (hash x) + 257 * (hash y)
    | Xml (x,y) -> 5 + 17 * (hash x) + 257 * (hash y)
    | Record x -> 6 + 17 * (LabelMap.hash hash x)
    | String (i,j,s,r) -> 7 + 17 * (U.hash s) + 257 * hash r
168
      (* Note: improve hash for String *)
169

170
171
  let equal c1 c2 = compare c1 c2 = 0
end
172

173
174
type pair_kind = [ `Normal | `XML ]

175

176
177
178
179
180
181
182
183
184
module rec Descr : 
sig
(*
  Want to write:
    type s = { ... }
    include Custom.T with type t = s
  but a  bug in OCaml 3.07+beta 2 makes it impossible
*)
  type t = {
185
    mutable hash: int;
186
187
188
189
190
191
192
193
194
    atoms : Atoms.t;
    ints  : Intervals.t;
    chars : Chars.t;
    times : BoolPair.t;
    xml   : BoolPair.t;
    arrow : BoolPair.t;
    record: BoolRec.t;
    absent: bool
  }
195
  val empty: t
196
197
198
199
200
201
202
203
204
205
  val dump: Format.formatter -> t -> unit
  val check: t -> unit
  val equal: t -> t -> bool
  val hash: t -> int
  val compare:t -> t -> int
  val serialize: t Serialize.Put.f
  val deserialize: t Serialize.Get.f
end =
struct
  type t = {
206
    mutable hash: int;
207
208
209
210
211
212
213
214
215
    atoms : Atoms.t;
    ints  : Intervals.t;
    chars : Chars.t;
    times : BoolPair.t;
    xml   : BoolPair.t;
    arrow : BoolPair.t;
    record: BoolRec.t;
    absent: bool
  }
216

217
218
219
  let dump ppf _ =
    Format.fprintf ppf "<Types.Descr.t>"

220
221
222
223
224
225
226
227
228
229
230
231
  let empty = { 
    hash = 0;
    times = BoolPair.empty; 
    xml   = BoolPair.empty; 
    arrow = BoolPair.empty; 
    record= BoolRec.empty;
    ints  = Intervals.empty;
    atoms = Atoms.empty;
    chars = Chars.empty;
    absent= false;
  }

232
  let equal a b =
233
234
235
236
237
238
239
240
241
242
    (a == b) || (
      (Atoms.equal a.atoms b.atoms) &&
      (Chars.equal a.chars b.chars) &&
      (Intervals.equal a.ints  b.ints) &&
      (BoolPair.equal a.times b.times) &&
      (BoolPair.equal a.xml b.xml) &&
      (BoolPair.equal a.arrow b.arrow) &&
      (BoolRec.equal a.record b.record) &&
      (a.absent == b.absent)
    )
243
244
245
246
247
248
249
250
251
252
253
254
255

  let compare a b =
    if a == b then 0 
    else let c = Atoms.compare a.atoms b.atoms in if c <> 0 then c
    else let c = Chars.compare a.chars b.chars in if c <> 0 then c
    else let c = Intervals.compare a.ints b.ints in if c <> 0 then c
    else let c = BoolPair.compare a.times b.times in if c <> 0 then c
    else let c = BoolPair.compare a.xml b.xml in if c <> 0 then c
    else let c = BoolPair.compare a.arrow b.arrow in if c <> 0 then c
    else let c = BoolRec.compare a.record b.record in if c <> 0 then c
    else if a.absent && not b.absent then -1
    else if b.absent && not a.absent then 1
    else 0
256
      
257
  let hash a =
258
259
260
261
262
263
264
265
266
267
268
269
    if a.hash <> 0 then a.hash else (
      let accu = Chars.hash a.chars in
      let accu = 17 * accu + Intervals.hash a.ints in
      let accu = 17 * accu + Atoms.hash a.atoms in
      let accu = 17 * accu + BoolPair.hash a.times in
      let accu = 17 * accu + BoolPair.hash a.xml in
      let accu = 17 * accu + BoolPair.hash a.arrow in
      let accu = 17 * accu + BoolRec.hash a.record in
      let accu = if a.absent then accu+5 else accu in
      a.hash <- accu;
      accu
    )
270

271
272
273
274
275
276
277
278
279
280
281
  let check a =
    Chars.check a.chars;
    Intervals.check a.ints;
    Atoms.check a.atoms;
    BoolPair.check a.times;
    BoolPair.check a.xml;
    BoolPair.check a.arrow;
    BoolRec.check a.record;
    ()


282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
  let serialize t a =
    Chars.serialize t a.chars;
    Intervals.serialize t a.ints;
    Atoms.serialize t a.atoms;
    BoolPair.serialize t a.times;
    BoolPair.serialize t a.xml;
    BoolPair.serialize t a.arrow;
    BoolRec.serialize t a.record;
    Serialize.Put.bool t a.absent 

  let deserialize t =
    let chars = Chars.deserialize t in
    let ints = Intervals.deserialize t in
    let atoms = Atoms.deserialize t in
    let times = BoolPair.deserialize t in
    let xml = BoolPair.deserialize t in
    let arrow = BoolPair.deserialize t in
    let record = BoolRec.deserialize t in
    let absent = Serialize.Get.bool t in
301
    let d = { hash=0; 
302
      chars = chars; ints = ints; atoms = atoms; times = times; xml = xml;
303
304
305
      arrow = arrow; record = record; absent = absent } in
    check d;
    d
306
307
   
    
308
309
310
end
and Node :
sig
311
  type t = { id : int; comp_unit: CompUnit.t; mutable descr : Descr.t }
312
313
314
315
316
317
318
  val dump: Format.formatter -> t -> unit
  val check: t -> unit
  val equal: t -> t -> bool
  val hash: t -> int
  val compare:t -> t -> int
  val serialize: t Serialize.Put.f
  val deserialize: t Serialize.Get.f
319
  val mk: int -> Descr.t -> t
320
end =
321

322
struct
323
  type t = { id : int; comp_unit : CompUnit.t; mutable descr : Descr.t }
324
325
  let check n = ()
  let dump ppf n = failwith "Types.Node.dump"
326
327
328
329
  let hash x = x.id + 17 * x.comp_unit
  let compare x y = 
    let c = x.id - y.id in
    if c = 0 then x.comp_unit - y.comp_unit else c
330
  let equal x y = x == y
331

332
333
334
335
336
337
338
339
  let serialize_memo = Inttbl.create ()
  let counter_serialize = ref 0

  let () =
    CompUnit.close_serialize_ref := 
    (fun () ->
       Inttbl.clear serialize_memo;
       counter_serialize := 0)
340

341
  let serialize t n = 
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
    if n.comp_unit == !CompUnit.current then (
      Serialize.Put.bool t true;
      try 
	let i = Inttbl.find serialize_memo n.id in
	Serialize.Put.int t i
      with Not_found ->
	let i = !counter_serialize in
	incr counter_serialize;
	Inttbl.add serialize_memo n.id i;
	Serialize.Put.int t i;
	Descr.serialize t n.descr
    ) else (
      Serialize.Put.bool t false;
      CompUnit.serialize t n.comp_unit;
      Serialize.Put.int t n.id
357
    )
358
      
359

360
361
362
363
  let deserialize_memo = Inttbl.create ()

  let find_tbl id =
    try Inttbl.find deserialize_memo id
364
    with Not_found ->
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
      let tbl = Inttbl.create () in
      Inttbl.add deserialize_memo id tbl;
      tbl

  let mk id d =
    let n = { id = id; comp_unit = !CompUnit.current; descr = d } in
    if !CompUnit.current == CompUnit.pervasives then
      Inttbl.add (find_tbl CompUnit.pervasives) n.id n;
    n

  let deserialize t = 
    if Serialize.Get.bool t then
      let i = Serialize.Get.int t in
      let tbl = find_tbl !CompUnit.current in
      try Inttbl.find tbl i
      with Not_found ->
	let n = { id = i; comp_unit = !CompUnit.current;
		  descr = Descr.empty } in
	Inttbl.add tbl i n;
	n.descr <- Descr.deserialize t;
	n
    else
      let cu = CompUnit.deserialize t in
      let i = Serialize.Get.int t in
      try Inttbl.find (Inttbl.find deserialize_memo cu) i 
      with Not_found -> assert false

392
393
end

394
(* It is also possible to use Boolean instead of Bool here;
395
   need to analyze when each one is more efficient *)
396
397
398
399
400
401
and BoolPair : Bool.S with type elem = Node.t * Node.t = 
Bool.Make(Custom.Pair(Node)(Node))

and BoolRec : Bool.S with type elem = bool * Node.t label_map =
Bool.Make(Custom.Pair(Custom.Bool)(LabelSet.MakeMap(Node)))

402
403
404
405
module DescrHash = Hashtbl.Make(Descr)
module DescrMap = Map.Make(Descr)
module DescrSet = Set.Make(Descr)
module DescrSList = SortedList.Make(Descr)
406

407
408
409
type descr = Descr.t
type node = Node.t
include Descr
410
411
412

let hash_cons = DescrHash.create 17000  

413
414
415
416
417
418
419
420
421
422
let count = State.ref "Types.count" 0
		
let () =
  Stats.register Stats.Summary
    (fun ppf -> Format.fprintf ppf "Allocated type nodes:%i@\n" !count)
      
let make () = 
  incr count; 
  Node.mk !count empty

423
424
425
426
427
428
429
let define n d = 
  DescrHash.add hash_cons d n; 
  n.Node.descr <- d
let cons d = 
  try DescrHash.find hash_cons d 
  with Not_found ->
    incr count; 
430
    let n = Node.mk !count d in
431
432
    DescrHash.add hash_cons d n; n  

433
let any =  {
434
  hash = 0;
435
436
437
  times = BoolPair.full; 
  xml   = BoolPair.full; 
  arrow = BoolPair.full; 
438
  record= BoolRec.full; 
439
440
441
  ints  = Intervals.any;
  atoms = Atoms.any;
  chars = Chars.any;
442
  absent= false;
443
}
444
445

let non_constructed =
446
447
448
  { any with  
      hash = 0;
      times = empty.times; xml = empty.xml; record = empty.record }
449
     
450
	     
451
452
453
454
let interval i = { empty with hash = 0; ints = i }
let times x y = { empty with hash = 0; times = BoolPair.atom (x,y) }
let xml x y = { empty with hash = 0; xml = BoolPair.atom (x,y) }
let arrow x y = { empty with hash = 0; arrow = BoolPair.atom (x,y) }
455
let record label t = 
456
457
  { empty with hash = 0; 
      record = BoolRec.atom (true,LabelMap.singleton label t) }
458
let record' (x : bool * node Ident.label_map) =
459
460
461
  { empty with hash = 0; record = BoolRec.atom x }
let atom a = { empty with hash = 0; atoms = a }
let char c = { empty with hash = 0; chars = c }
462
      
463
464
let cup x y = 
  if x == y then x else {
465
    hash = 0;
466
467
468
    times = BoolPair.cup x.times y.times;
    xml   = BoolPair.cup x.xml y.xml;
    arrow = BoolPair.cup x.arrow y.arrow;
469
    record= BoolRec.cup x.record y.record;
470
471
472
    ints  = Intervals.cup x.ints  y.ints;
    atoms = Atoms.cup x.atoms y.atoms;
    chars = Chars.cup x.chars y.chars;
473
    absent= x.absent || y.absent;
474
475
476
477
  }
    
let cap x y = 
  if x == y then x else {
478
    hash = 0;
479
480
    times = BoolPair.cap x.times y.times;
    xml   = BoolPair.cap x.xml y.xml;
481
    record= BoolRec.cap x.record y.record;
482
    arrow = BoolPair.cap x.arrow y.arrow;
483
484
485
    ints  = Intervals.cap x.ints  y.ints;
    atoms = Atoms.cap x.atoms y.atoms;
    chars = Chars.cap x.chars y.chars;
486
    absent= x.absent && y.absent;
487
488
489
490
  }
    
let diff x y = 
  if x == y then empty else {
491
    hash = 0;
492
493
494
    times = BoolPair.diff x.times y.times;
    xml   = BoolPair.diff x.xml y.xml;
    arrow = BoolPair.diff x.arrow y.arrow;
495
    record= BoolRec.diff x.record y.record;
496
497
498
    ints  = Intervals.diff x.ints  y.ints;
    atoms = Atoms.diff x.atoms y.atoms;
    chars = Chars.diff x.chars y.chars;
499
    absent= x.absent && not y.absent;
500
501
  }
    
502

503

504

505
506
507
508
509
510
511
512
(* TODO: optimize disjoint check for boolean combinations *)
let trivially_disjoint a b =
  (Chars.disjoint a.chars b.chars) &&
  (Intervals.disjoint a.ints b.ints) &&
  (Atoms.disjoint a.atoms b.atoms) &&
  (BoolPair.trivially_disjoint a.times b.times) &&
  (BoolPair.trivially_disjoint a.xml b.xml) &&
  (BoolPair.trivially_disjoint a.arrow b.arrow) &&
513
514
  (BoolRec.trivially_disjoint a.record b.record) &&
  (not (a.absent && b.absent))
515

516

517

518
let descr n = n.Node.descr
519
let internalize n = n
520
let id n = n.Node.id
521
522


523
524
525
526
527
528
529
530
531
532
533
534
535
let rec constant = function
  | Integer i -> interval (Intervals.atom i)
  | Atom a -> atom (Atoms.atom a)
  | Char c -> char (Chars.atom c)
  | Pair (x,y) -> times (const_node x) (const_node y)
  | Xml (x,y) -> times (const_node x) (const_node y)
  | Record x -> record' (false ,LabelMap.map const_node x)
  | String (i,j,s,c) ->
      if U.equal_index i j then constant c
      else 
	let (ch,i') = U.next s i in
	constant (Pair (Char (Chars.V.mk_int ch), String (i',j,s,c)))
and const_node c = cons (constant c)
536

537
538
let neg x = diff any x

539
540
let any_node = cons any

541
module LabelS = Set.Make(LabelPool)
542

543
544
545
let any_or_absent = { any with hash=0; absent = true } 
let only_absent = { empty with hash=0; absent = true }

546
547
let get_record r =
  let labs accu (_,r) = 
548
549
    List.fold_left 
      (fun accu (l,_) -> LabelS.add l accu) accu (LabelMap.get r) in
550
  let extend descrs labs (o,r) =
551
552
553
554
555
    let rec aux i labs r =
      match labs with
	| [] -> ()
	| l1::labs ->
	    match r with
556
	      | (l2,x)::r when l1 == l2 -> 
557
558
559
		  descrs.(i) <- cap descrs.(i) (descr x);
		  aux (i+1) labs r
	      | r ->
560
561
		  if not o then 
		    descrs.(i) <- cap descrs.(i) only_absent; (* TODO:OPT *)
562
563
		  aux (i+1) labs r
    in
564
    aux 0 labs (LabelMap.get r);
565
566
567
568
    o
  in
  let line (p,n) =
    let labels = 
569
570
      List.fold_left labs (List.fold_left labs LabelS.empty p) n in
    let labels = LabelS.elements labels in
571
    let nlab = List.length labels in
572
    let mk () = Array.create nlab any_or_absent in
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587

    let pos = mk () in
    let opos = List.fold_left 
		 (fun accu x -> 
		    (extend pos labels x) && accu)
		 true p in
    let p = (opos, pos) in

    let n = List.map (fun x ->
			let neg = mk () in
			let o = extend neg labels x in
			(o,neg)
		     ) n in
    (labels,p,n)
  in
588
  List.map line (BoolRec.get r)
589
   
590

591

592
593
594
595
596
597
598


(* Subtyping algorithm *)

let diff_t d t = diff d (descr t)
let cap_t d t = cap d (descr t)
let cup_t d t = cup d (descr t)
599
let cap_product any_left any_right l =
600
601
  List.fold_left 
    (fun (d1,d2) (t1,t2) -> (cap_t d1 t1, cap_t d2 t2))
602
    (any_left,any_right)
603
    l
604
let any_pair = { empty with hash = 0; times = any.times }
605

606

607
608
609
let rec exists max f =
  (max > 0) && (f (max - 1) || exists (max - 1) f)

610
exception NotEmpty
611

612
613
614
615
616
617
618
619
620
621
622
623
type slot = { mutable status : status; 
	       mutable notify : notify;
	       mutable active : bool }
and status = Empty | NEmpty | Maybe
and notify = Nothing | Do of slot * (slot -> unit) * notify

let slot_empty = { status = Empty; active = false; notify = Nothing }
let slot_not_empty = { status = NEmpty; active = false; notify = Nothing }

let rec notify = function
  | Nothing -> ()
  | Do (n,f,rem) -> 
624
      if n.status == Maybe then (try f n with NotEmpty -> ());
625
626
627
628
629
630
631
632
633
634
      notify rem

let rec iter_s s f = function
  | [] -> ()
  | arg::rem -> f arg s; iter_s s f rem


let set s =
  s.status <- NEmpty;
  notify s.notify;
635
  s.notify <- Nothing; 
636
637
638
639
640
641
642
  raise NotEmpty

let rec big_conj f l n =
  match l with
    | [] -> set n
    | [arg] -> f arg n
    | arg::rem ->
643
644
645
	let s = 
	  { status = Maybe; active = false; 
	    notify = Do (n,(big_conj f rem), Nothing) } in
646
647
648
	try 
	  f arg s;
	  if s.active then n.active <- true
649
	with NotEmpty -> if n.status == NEmpty then raise NotEmpty
650

651
652
let guard a f n =
  match a with
653
    | { status = Empty } -> ()
654
655
656
    | { status = Maybe } as s -> 
	n.active <- true; 
	s.notify <- Do (n,f,s.notify)
657
    | { status = NEmpty } -> f n
658

659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756

(* Fast approximation *)

module ClearlyEmpty = 
struct

let memo = DescrHash.create 33000
let marks = ref [] 

let rec slot d =
  if not ((Intervals.is_empty d.ints) && 
	  (Atoms.is_empty d.atoms) &&
	  (Chars.is_empty d.chars) &&
	  (not d.absent)) then slot_not_empty 
  else try DescrHash.find memo d
  with Not_found ->
    let s = { status = Maybe; active = false; notify = Nothing } in
    DescrHash.add memo d s;
    (try
       iter_s s check_times (BoolPair.get d.times);  
       iter_s s check_xml (BoolPair.get d.xml); 
       iter_s s check_arrow (BoolPair.get d.arrow);
       iter_s s check_record (get_record d.record);
       if s.active then marks := s :: !marks else s.status <- Empty;
     with
	 NotEmpty -> ());
    s

and check_times (left,right) s =
  let (accu1,accu2) = cap_product any any left in
  let single_right (t1,t2) s =
    let t1 = descr t1 and t2 = descr t2 in
    if trivially_disjoint accu1 t1 || trivially_disjoint accu2 t2 then set s 
    else
      let accu1 = diff accu1 t1 in guard (slot accu1) set s;
      let accu2 = diff accu2 t2 in guard (slot accu2) set s in
  guard (slot accu1) (guard (slot accu2) (big_conj single_right right)) s

and check_xml (left,right) s =
  let (accu1,accu2) = cap_product any any_pair left in
  let single_right (t1,t2) s =
    let t1 = descr t1 and t2 = descr t2 in
    if trivially_disjoint accu1 t1 || trivially_disjoint accu2 t2 then set s 
    else
      let accu1 = diff accu1 t1 in guard (slot accu1) set s;
      let accu2 = diff accu2 t2 in guard (slot accu2) set s in
  guard (slot accu1) (guard (slot accu2) (big_conj single_right right)) s

and check_arrow (left,right) s =
  let single_right (s1,s2) s =
    let accu1 = descr s1 and accu2 = neg (descr s2) in
    let single_left (t1,t2) s =
      let accu1 = diff_t accu1 t1 in guard (slot accu1) set s;
      let accu2 = cap_t  accu2 t2 in guard (slot accu2) set s
    in
    guard (slot accu1) (big_conj single_left left) s
  in
  big_conj single_right right s

and check_record (labels,(oleft,left),rights) s =
  let rec single_right (oright,right) s = 
    let next =
      (oleft && (not oright)) ||
      exists (Array.length left)
	(fun i -> trivially_disjoint left.(i) right.(i))
    in
    if next then set s
    else
      for i = 0 to Array.length left - 1 do
	let di = diff left.(i) right.(i) in guard (slot di) set s
      done
  in
  let rec start i s =
    if (i < 0) then big_conj single_right rights s
    else guard (slot left.(i)) (start (i - 1)) s
  in
  start (Array.length left - 1) s


let is_empty d =
  let s = slot d in
  List.iter 
    (fun s' -> 
       if s'.status == Maybe then s'.status <- Empty; s'.notify <- Nothing) 
    !marks;
  marks := [];
  s.status == Empty
end

let clearly_disjoint t1 t2 =
(*
  if trivially_disjoint t1 t2 then true
  else
    if ClearlyEmpty.is_empty (cap t1 t2) then
      (Printf.eprintf "!\n"; true) else false
*)
  trivially_disjoint t1 t2 || ClearlyEmpty.is_empty (cap t1 t2) 

757
758
(* TODO: need to invesigate when ClearEmpty is a good thing... *)

759
760
761
762
let memo = DescrHash.create 33000
let marks = ref [] 

let rec slot d =
763
764
  if not ((Intervals.is_empty d.ints) && 
	  (Atoms.is_empty d.atoms) &&
765
766
	  (Chars.is_empty d.chars) &&
	  (not d.absent)) then slot_not_empty 
767
768
769
770
771
  else try DescrHash.find memo d
  with Not_found ->
    let s = { status = Maybe; active = false; notify = Nothing } in
    DescrHash.add memo d s;
    (try
772
       iter_s s check_times (BoolPair.get d.times);  
773
       iter_s s check_xml (BoolPair.get d.xml); 
774
       iter_s s check_arrow (BoolPair.get d.arrow);
775
776
777
778
779
780
781
782
783
       iter_s s check_record (get_record d.record);
       if s.active then marks := s :: !marks else s.status <- Empty;
     with
	 NotEmpty -> ());
    s

and check_times (left,right) s =
  let rec aux accu1 accu2 right s = match right with
    | (t1,t2)::right ->
784
785
786
	let t1 = descr t1 and t2 = descr t2 in
	if trivially_disjoint accu1 t1 || 
	   trivially_disjoint accu2 t2 then (
787
788
	     aux accu1 accu2 right s )
	else (
789
          let accu1' = diff accu1 t1 in 
790
	  guard (slot accu1') (aux accu1' accu2 right) s;
791
792

          let accu2' = diff accu2 t2 in 
793
	  guard (slot accu2') (aux accu1 accu2' right) s  
794
	)
795
796
    | [] -> set s
  in
797
  let (accu1,accu2) = cap_product any any left in
798
  guard (slot accu1) (guard (slot accu2) (aux accu1 accu2 right)) s
799
800
801
802
803

and check_xml (left,right) s =
  let rec aux accu1 accu2 right s = match right with
    | (t1,t2)::right ->
	let t1 = descr t1 and t2 = descr t2 in
804
	if clearly_disjoint accu1 t1 || 
805
806
807
808
	   trivially_disjoint accu2 t2 then (
	     aux accu1 accu2 right s )
	else (
          let accu1' = diff accu1 t1 in 
809
	  guard (slot accu1') (aux accu1' accu2 right) s;
810
811

          let accu2' = diff accu2 t2 in 
812
	  guard (slot accu2') (aux accu1 accu2' right) s  
813
814
815
816
	)
    | [] -> set s
  in
  let (accu1,accu2) = cap_product any any_pair left in
817
  guard (slot accu1) (guard (slot accu2) (aux accu1 accu2 right)) s
818

819
820
821
822
and check_arrow (left,right) s =
  let single_right (s1,s2) s =
    let rec aux accu1 accu2 left s = match left with
      | (t1,t2)::left ->
823
          let accu1' = diff_t accu1 t1 in 
824
	  guard (slot accu1') (aux accu1' accu2 left) s;
825
826

          let accu2' = cap_t  accu2 t2 in 
827
	  guard (slot accu2') (aux accu1 accu2' left) s
828
829
830
      | [] -> set s
    in
    let accu1 = descr s1 in
831
    guard (slot accu1) (aux accu1 (neg (descr s2)) left) s
832
833
  in
  big_conj single_right right s
834

835
and check_record (labels,(oleft,left),rights) s =
836
837
  let rec aux rights s = match rights with
    | [] -> set s
838
    | (oright,right)::rights ->
839
	let next =
840
	  (oleft && (not oright)) ||
841
	  exists (Array.length left)
842
	    (fun i -> trivially_disjoint left.(i) right.(i))
843
844
845
846
847
848
	in
	if next then aux rights s
	else
	  for i = 0 to Array.length left - 1 do
	    let back = left.(i) in
	    let di = diff back right.(i) in
849
850
	    guard (slot di) (fun s ->
			left.(i) <- di;
851
852
853
			aux rights s;
			left.(i) <- back;
		     ) s
854
(* TODO: are side effects correct ? *)
855
856
857
858
859
	  done
  in
  let rec start i s =
    if (i < 0) then aux rights s
    else
860
      guard (slot left.(i)) (start (i - 1)) s
861
862
863
864
  in
  start (Array.length left - 1) s


865
(*
866
let timer_subtype = Stats.Timer.create "Types.is_empty"
867
*)
868

869
let is_empty d =
870
(*  Stats.Timer.start timer_subtype;*)
871
872
  let s = slot d in
  List.iter 
873
874
    (fun s' -> 
       if s'.status == Maybe then s'.status <- Empty; s'.notify <- Nothing) 
875
876
    !marks;
  marks := [];
877
(*  Stats.Timer.stop timer_subtype *)
878
    (s.status == Empty)
879

880
(*
881
let is_empty d =
882
883
884
885
886
887
888
(*  let b1 = ClearlyEmpty.is_empty d in
  let b2 = is_empty d in
  assert (b2 || not b1);
  Printf.eprintf "b1 = %b; b2 = %b\n" b1 b2;
  b2  *)
  if ClearlyEmpty.is_empty d then (Printf.eprintf "!\n"; true) else is_empty d
*)  
889
890
891
892
893
894
895
896
897
898
899
900
901

let non_empty d = 
  not (is_empty d)

let subtype d1 d2 =
  is_empty (diff d1 d2)

module Product =
struct
  type t = (descr * descr) list

  let other ?(kind=`Normal) d = 
    match kind with
902
903
      | `Normal -> { d with hash = 0; times = empty.times }
      | `XML -> { d with hash = 0; xml = empty.xml }
904
905
906
907
908

  let is_product ?kind d = is_empty (other ?kind d)

  let need_second = function _::_::_ -> true | _ -> false

909
910
911
912
  let normal_aux = function
    | ([] | [ _ ]) as d -> d
    | d ->

913
914
915
916
917
918
919
    let res = ref [] in

    let add (t1,t2) =
      let rec loop t1 t2 = function
	| [] -> res := (ref (t1,t2)) :: !res
	| ({contents = (d1,d2)} as r)::l ->
	    (*OPT*) 
920
(*	    if equal_descr d1 t1 then r := (d1,cup d2 t2) else*)
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
	      
	      let i = cap t1 d1 in
	      if is_empty i then loop t1 t2 l
	      else (
		r := (i, cup t2 d2);
		let k = diff d1 t1 in 
		if non_empty k then res := (ref (k,d2)) :: !res;
		
		let j = diff t1 d1 in 
		if non_empty j then loop j t2 l
	      )
      in
      loop t1 t2 !res
    in
    List.iter add d;
    List.map (!) !res


(* Partitioning:

(t,s) - ((t1,s1) | (t2,s2) | ... | (tn,sn))
=
(t & t1, s - s1) | ... | (t & tn, s - sn) | (t - (t1|...|tn), s)

945
*)
946
  let get_aux any_right d =
947
948
    let accu = ref [] in
    let line (left,right) =
949
      let (d1,d2) = cap_product any any_right left in
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
      if (non_empty d1) && (non_empty d2) then
	let right = List.map (fun (t1,t2) -> descr t1, descr t2) right in
	let right = normal_aux right in
	let resid1 = ref d1 in
	let () = 
	  List.iter
	    (fun (t1,t2) ->
	       let t1 = cap d1 t1 in
	       if (non_empty t1) then
		 let () = resid1 := diff !resid1 t1 in
		 let t2 = diff d2 t2 in
		 if (non_empty t2) then accu := (t1,t2) :: !accu
	    ) right in
	if non_empty !resid1 then accu := (!resid1, d2) :: !accu 
    in
965
    List.iter line (BoolPair.get d);
966
    !accu
967
968
969
(* Maybe, can improve this function with:
     (t,s) \ (t1,s1) = (t&t',s\s') | (t\t',s),
   don't call normal_aux *)
970

971

972
973
  let get ?(kind=`Normal) d = 
    match kind with
974
975
      | `Normal -> get_aux any d.times
      | `XML -> get_aux any_pair d.xml
976
977
978

  let pi1 = List.fold_left (fun acc (t1,_) -> cup acc t1) empty
  let pi2 = List.fold_left (fun acc (_,t2) -> cup acc t2) empty
979
980
981
982
  let pi2_restricted restr = 
    List.fold_left (fun acc (t1,t2) -> 
		      if is_empty (cap t1 restr) then acc
		      else cup acc t2) empty
983
984

  let restrict_1 rects pi1 =
985
986
    let aux acc (t1,t2) = 
      let t1 = cap t1 pi1 in if is_empty t1 then acc else (t1,t2)::acc in
987
988
989
990
    List.fold_left aux [] rects
  
  type normal = t

991
  module Memo = Map.Make(BoolPair)
992

993
994
  (* TODO: try with an hashtable *)
  (* Also, avoid lookup for simple products (t1,t2) *)
995
  let memo = ref Memo.empty
996
  let normal_times d = 
997
998
999
    try Memo.find d !memo 
    with
	Not_found ->
1000
	  let gd = get_aux any d in
1001
	  let n = normal_aux gd in
1002
1003
(* Could optimize this call to normal_aux because one already
   know that each line is normalized ... *)
1004
1005
	  memo := Memo.add d n !memo;
	  n
1006

1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
  let memo_xml = ref Memo.empty
  let normal_xml d = 
    try Memo.find d !memo_xml
    with
	Not_found ->
	  let gd = get_aux any_pair d in
	  let n = normal_aux gd in
	  memo_xml := Memo.add d n !memo_xml;
	  n

  let normal ?(kind=`Normal) d =
    match kind with 
      | `Normal -> normal_times d.times 
      | `XML -> normal_xml d.xml


1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
  let merge_same_2 r =
    let r = 
      List.fold_left 
	(fun accu (t1,t2) ->
	   let t = try DescrMap.find t2 accu with Not_found -> empty in
	   DescrMap.add t2 (cup t t1) accu
	) DescrMap.empty r in
    DescrMap.fold (fun t2 t1 accu -> (t1,t2)::accu) r []
	 

1033
1034
1035
1036
1037
1038
1039
  let constraint_on_2 n t1 =
    List.fold_left 
      (fun accu (d1,d2) ->
	 if is_empty (cap d1 t1) then accu else cap accu d2)
      any
      n

1040
1041
  let any = { empty with hash = 0; times = any.times }
  and any_xml = { empty with hash = 0; xml = any.xml }
1042
  let is_empty d = d == []
1043
end
1044

1045
module Record = 
1046
struct
1047
1048
  let has_record d = not (is_empty { empty with hash= 0; record = d.record })
  let or_absent d = { d with hash = 0; absent = true }
1049
1050
1051
  let any_or_absent = or_absent any
  let has_absent d = d.absent

1052
  let only_absent = {empty with hash = 0; absent = true}
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
  let only_absent_node = cons only_absent

  module T = struct
    type t = descr
    let any = any_or_absent
    let cap = cap
    let cup = cup
    let diff = diff
    let is_empty = is_empty
    let empty = empty
  end
  module R = struct
    type t = descr
1066
    let any = { empty with hash = 0; record = any.record }
1067
1068
1069
1070
1071
1072
1073
1074
    let cap = cap
    let cup = cup
    let diff = diff
    let is_empty = is_empty
    let empty = empty
  end
  module TR = Normal.Make(T)(R)

1075
  let any_record = { empty with hash = 0; record = BoolRec.full }
1076
1077
1078

  let atom o l = 
    if o && LabelMap.is_empty l then any_record else
1079
    { empty with hash = 0; record = BoolRec.atom (o,l) }
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090

  type zor = Pair of descr * descr | Any

  let aux_split d l=
    let f (o,r) =
      try
	let (lt,rem) = LabelMap.assoc_remove l r in
	Pair (descr lt, atom o rem)
      with Not_found -> 
	if o then
	  if LabelMap.is_empty r then Any else
1091
	    Pair (any_or_absent, { empty with hash=0; record = BoolRec.atom (o,r) })
1092
1093
	else
	  Pair (only_absent,
1094
		{ empty with hash = 0; record = BoolRec.atom (o,r) })
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
    in
    List.fold_left 
      (fun b (p,n) ->
	 let rec aux_p accu = function
	   | x::p -> 
	       (match f x with
		  | Pair (t1,t2) -> aux_p ((t1,t2)::accu) p
		  | Any -> aux_p accu p)
	   | [] -> aux_n accu [] n
	 and aux_n p accu = function
	   | x::n -> 
	       (match f x with
		  | Pair (t1,t2) -> aux_n p ((t1,t2)::accu) n
		  | Any -> b)
	   | [] -> (p,accu) :: b in
	 aux_p [] p)
      []
      (BoolRec.get d.record)

  let split (d : descr) l =
    TR.boolean (aux_split d l)

  let split_normal d l =
    TR.boolean_normal (aux_split d l)


  let project d l =
    let t = TR.pi1 (split d l) in
    if t.absent then raise Not_found;
    t

  let project_opt d l =
    let t = TR.pi1 (split d l) in
1128
    { t with hash = 0; absent = false }
1129
1130
1131

  let condition d l t =
    TR.pi2_restricted t (split d l)
1132

1133
1134
1135
1136
1137
(* TODO: eliminate this cap ... (reord l only_absent_node) when
   not necessary. eg. {| ..... |} \ l *)

  let remove_field d l = 
    cap (TR.pi2 (split d l)) (record l only_absent_node)
1138

1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
  let first_label d =
    let min = ref LabelPool.dummy_max in
    let aux (_,r) = 
      match LabelMap.get r with
	  (l,_)::_ -> if (l:int) < !min then min := l | _ -> () in
    BoolRec.iter aux d.record;
    !min

  let empty_cases d =
    let x = BoolRec.compute
	      ~empty:0 ~full:3 ~cup:(lor) ~cap:(land)
	      ~diff:(fun a b -> a land lnot b)
	      ~atom:(function (o,r) ->
		       assert (LabelMap.get r == []);
		       if o then 3 else 1
		    )
	      d.record in
    (x land 2 <> 0, x land 1 <> 0)

  let has_empty_record d =
    BoolRec.compute
      ~empty:false ~full:true ~cup:(||) ~cap:(&&)
      ~diff:(fun a b -> a && not b)
      ~atom:(function (o,r) ->
	       List.for_all 
	         (fun (l,t) -> (descr t).absent)
	         (LabelMap.get r)
	    )
      d.record
    

(*TODO: optimize merge
   - pre-compute the sequence of labels
   - remove empty or full { l = t }
*)

  let merge d1 d2 = 
    let res = ref empty in
    let rec aux accu d1 d2 =
      let l = min (first_label d1) (first_label d2) in
      if l = LabelPool.dummy_max then
	let (some1,none1) = empty_cases d1 
	and (some2,none2) = empty_cases d2 in
	let none = none1 && none2 and some = some1 || some2 in
	let accu = LabelMap.from_list (fun _ _ -> assert false) accu in
	(* approx for the case (some && not none) ... *)
	res := cup !res (record' (some, accu))
      else
	let l1 = split d1 l and l2 = split d2 l in
	let loop (t1,d1) (t2,d2) =
	  let t = 
	    if t2.absent 
1191
	    then cup t1 { t2 with hash = 0; absent = false } 
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
	    else t2 
	  in
	  aux ((l,cons t)::accu) d1 d2
	in
	List.iter (fun x -> List.iter (loop x) l2) l1
	  
    in
    aux [] d1 d2;
    !res

1202
  let any = { empty with hash = 0; record = any.record }