1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
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
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
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
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
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
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
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
|
(************************************************************************)
(* v * The Coq Proof Assistant / The Coq Development Team *)
(* <O___,, * INRIA - CNRS - LIX - LRI - PPS - Copyright 1999-2016 *)
(* \VV/ **************************************************************)
(* // * This file is distributed under the terms of the *)
(* * GNU Lesser General Public License Version 2.1 *)
(************************************************************************)
(** This file defines the low-level monadic operations used by the
tactic monad. The monad is divided into two layers: a non-logical
layer which consists in operations which will not (or cannot) be
backtracked in case of failure (input/output or persistent state)
and a logical layer which handles backtracking, proof
manipulation, and any other effect which needs to backtrack. *)
(** {6 Exceptions} *)
(** To help distinguish between exceptions raised by the IO monad from
the one used natively by Coq, the former are wrapped in
[Exception]. It is only used internally so that [catch] blocks of
the IO monad would only catch exceptions raised by the [raise]
function of the IO monad, and not for instance, by system
interrupts. Also used in [Proofview] to avoid capturing exception
from the IO monad ([Proofview] catches errors in its compatibility
layer, and when lifting goal-level expressions). *)
exception Exception of exn
(** This exception is used to signal abortion in [timeout] functions. *)
exception Timeout
(** This exception is used by the tactics to signal failure by lack of
successes, rather than some other exceptions (like system
interrupts). *)
exception TacticFailure of exn
let _ = Errors.register_handler begin function
| Timeout -> Errors.errorlabstrm "Some timeout function" (Pp.str"Timeout!")
| Exception e -> Errors.print e
| TacticFailure e -> Errors.print e
| _ -> Pervasives.raise Errors.Unhandled
end
(** {6 Non-logical layer} *)
(** The non-logical monad is a simple [unit -> 'a] (i/o) monad. The
operations are simple wrappers around corresponding usual
operations and require little documentation. *)
module NonLogical =
struct
(* The functions in this module follow the pattern that they are
defined with the form [(); fun ()->...]. This is an optimisation
which signals to the compiler that the function is usually partially
applied up to the [();]. Without this annotation, partial
applications can be significantly slower.
Documentation of this behaviour can be found at:
https://ocaml.janestreet.com/?q=node/30 *)
include Monad.Make(struct
type 'a t = unit -> 'a
let return a = (); fun () -> a
let (>>=) a k = (); fun () -> k (a ()) ()
let (>>) a k = (); fun () -> a (); k ()
let map f a = (); fun () -> f (a ())
end)
type 'a ref = 'a Pervasives.ref
let ignore a = (); fun () -> ignore (a ())
let ref a = (); fun () -> Pervasives.ref a
(** [Pervasives.(:=)] *)
let (:=) r a = (); fun () -> r := a
(** [Pervasives.(!)] *)
let (!) = fun r -> (); fun () -> ! r
(** [Pervasives.raise]. Except that exceptions are wrapped with
{!Exception}. *)
let raise ?info = fun e -> (); fun () -> Exninfo.raise ?info (Exception e)
(** [try ... with ...] but restricted to {!Exception}. *)
let catch = fun s h -> ();
fun () -> try s ()
with Exception e as src ->
let (src, info) = Errors.push src in
h (e, info) ()
let read_line = fun () -> try Pervasives.read_line () with e ->
let (e, info) = Errors.push e in raise ~info e ()
let print_char = fun c -> (); fun () -> print_char c
let timeout = fun n t -> (); fun () ->
Control.timeout n t (Exception Timeout)
let make f = (); fun () ->
try f ()
with e when Errors.noncritical e ->
let (e, info) = Errors.push e in
Util.iraise (Exception e, info)
(** Use the current logger. The buffer is also flushed. *)
let print_debug s = make (fun _ -> Pp.msg_info s;Pp.pp_flush ())
let print_info s = make (fun _ -> Pp.msg_info s;Pp.pp_flush ())
let print_warning s = make (fun _ -> Pp.msg_warning s;Pp.pp_flush ())
let print_error s = make (fun _ -> Pp.msg_error s;Pp.pp_flush ())
let print_notice s = make (fun _ -> Pp.msg_notice s;Pp.pp_flush ())
let run = fun x ->
try x () with Exception e as src ->
let (src, info) = Errors.push src in
Util.iraise (e, info)
end
(** {6 Logical layer} *)
(** The logical monad is a backtracking monad on top of which is
layered a state monad (which is used to implement all of read/write,
read only, and write only effects). The state monad being layered on
top of the backtracking monad makes it so that the state is
backtracked on failure.
Backtracking differs from regular exception in that, writing (+)
for exception catching and (>>=) for bind, we require the
following extra distributivity laws:
x+(y+z) = (x+y)+z
zero+x = x
x+zero = x
(x+y)>>=k = (x>>=k)+(y>>=k) *)
(** A view type for the logical monad, which is a form of list, hence
we can decompose it with as a list. *)
type ('a, 'b) list_view =
| Nil of Exninfo.iexn
| Cons of 'a * 'b
module type Param = sig
(** Read only *)
type e
(** Write only *)
type w
(** [w] must be a monoid *)
val wunit : w
val wprod : w -> w -> w
(** Read-write *)
type s
(** Update-only. Essentially a writer on [u->u]. *)
type u
(** [u] must be pointed. *)
val uunit : u
end
module Logical (P:Param) =
struct
(** All three of environment, writer and state are coded as a single
state-passing-style monad.*)
type state = {
rstate : P.e;
ustate : P.u;
wstate : P.w;
sstate : P.s;
}
(** Double-continuation backtracking monads are reasonable folklore
for "search" implementations (including the Tac interactive
prover's tactics). Yet it's quite hard to wrap your head around
these. I recommand reading a few times the "Backtracking,
Interleaving, and Terminating Monad Transformers" paper by
O. Kiselyov, C. Shan, D. Friedman, and A. Sabry. The peculiar
shape of the monadic type is reminiscent of that of the
continuation monad transformer.
The paper also contains the rationale for the [split] abstraction.
An explanation of how to derive such a monad from mathematical
principles can be found in "Kan Extensions for Program
Optimisation" by Ralf Hinze.
A somewhat concrete view is that the type ['a iolist] is, in fact
the impredicative encoding of the following stream type:
[type 'a _iolist' = Nil of exn | Cons of 'a*'a iolist'
and 'a iolist = 'a _iolist NonLogical.t]
Using impredicative encoding avoids intermediate allocation and
is, empirically, very efficient in Ocaml. It also has the
practical benefit that the monadic operation are independent of
the underlying monad, which simplifies the code and side-steps
the limited inlining of Ocaml.
In that vision, [bind] is simply [concat_map] (though the cps
version is significantly simpler), [plus] is concatenation, and
[split] is pattern-matching. *)
type rich_exn = Exninfo.iexn
type 'a iolist =
{ iolist : 'r. state -> (rich_exn -> 'r NonLogical.t) ->
('a -> state -> (rich_exn -> 'r NonLogical.t) -> 'r NonLogical.t) ->
'r NonLogical.t }
include Monad.Make(struct
type 'a t = 'a iolist
let return x =
{ iolist = fun s nil cons -> cons x s nil }
let (>>=) m f =
{ iolist = fun s nil cons ->
m.iolist s nil (fun x s next -> (f x).iolist s next cons) }
let (>>) m f =
{ iolist = fun s nil cons ->
m.iolist s nil (fun () s next -> f.iolist s next cons) }
let map f m =
{ iolist = fun s nil cons -> m.iolist s nil (fun x s next -> cons (f x) s next) }
end)
let zero e =
{ iolist = fun _ nil cons -> nil e }
let plus m1 m2 =
{ iolist = fun s nil cons -> m1.iolist s (fun e -> (m2 e).iolist s nil cons) cons }
let ignore m =
{ iolist = fun s nil cons -> m.iolist s nil (fun _ s next -> cons () s next) }
let lift m =
{ iolist = fun s nil cons -> NonLogical.(m >>= fun x -> cons x s nil) }
(** State related *)
let get =
{ iolist = fun s nil cons -> cons s.sstate s nil }
let set (sstate : P.s) =
{ iolist = fun s nil cons -> cons () { s with sstate } nil }
let modify (f : P.s -> P.s) =
{ iolist = fun s nil cons -> cons () { s with sstate = f s.sstate } nil }
let current =
{ iolist = fun s nil cons -> cons s.rstate s nil }
let local e m =
{ iolist = fun s nil cons ->
m.iolist { s with rstate = e } nil
(fun x s' next -> cons x {s' with rstate = s.rstate} next) }
let put w =
{ iolist = fun s nil cons -> cons () { s with wstate = P.wprod s.wstate w } nil }
let update (f : P.u -> P.u) =
{ iolist = fun s nil cons -> cons () { s with ustate = f s.ustate } nil }
(** List observation *)
let once m =
{ iolist = fun s nil cons -> m.iolist s nil (fun x s _ -> cons x s nil) }
let break f m =
{ iolist = fun s nil cons ->
m.iolist s nil (fun x s next -> cons x s (fun e -> match f e with None -> next e | Some e -> nil e))
}
(** For [reflect] and [split] see the "Backtracking, Interleaving,
and Terminating Monad Transformers" paper. *)
type 'a reified = ('a, rich_exn -> 'a reified) list_view NonLogical.t
let rec reflect (m : ('a * state) reified) : 'a iolist =
{ iolist = fun s0 nil cons ->
let next = function
| Nil e -> nil e
| Cons ((x, s), l) -> cons x s (fun e -> (reflect (l e)).iolist s0 nil cons)
in
NonLogical.(m >>= next)
}
let split m : ('a, rich_exn -> 'a t) list_view t =
let rnil e = NonLogical.return (Nil e) in
let rcons p s l = NonLogical.return (Cons ((p, s), l)) in
{ iolist = fun s nil cons ->
let open NonLogical in
m.iolist s rnil rcons >>= begin function
| Nil e -> cons (Nil e) s nil
| Cons ((x, s), l) ->
let l e = reflect (l e) in
cons (Cons (x, l)) s nil
end }
let run m r s =
let s = { wstate = P.wunit; ustate = P.uunit; rstate = r; sstate = s } in
let rnil e = NonLogical.return (Nil e) in
let rcons x s l =
let p = (x, s.sstate, s.wstate, s.ustate) in
NonLogical.return (Cons (p, l))
in
m.iolist s rnil rcons
let repr x = x
end
|