blanchet@39692
|
1 |
(* Title: HOL/Tools/ATP/atp_proof.ML
|
blanchet@39692
|
2 |
Author: Lawrence C. Paulson, Cambridge University Computer Laboratory
|
blanchet@39692
|
3 |
Author: Claire Quigley, Cambridge University Computer Laboratory
|
blanchet@39692
|
4 |
Author: Jasmin Blanchette, TU Muenchen
|
blanchet@39692
|
5 |
|
blanchet@43745
|
6 |
Abstract representation of ATP proofs and TSTP/SPASS syntax.
|
blanchet@39692
|
7 |
*)
|
blanchet@39692
|
8 |
|
blanchet@39692
|
9 |
signature ATP_PROOF =
|
blanchet@39692
|
10 |
sig
|
blanchet@39692
|
11 |
type 'a fo_term = 'a ATP_Problem.fo_term
|
blanchet@43402
|
12 |
type ('a, 'b, 'c) formula = ('a, 'b, 'c) ATP_Problem.formula
|
blanchet@43784
|
13 |
type 'a problem = 'a ATP_Problem.problem
|
blanchet@39692
|
14 |
|
blanchet@39731
|
15 |
datatype failure =
|
blanchet@43458
|
16 |
Unprovable |
|
blanchet@43458
|
17 |
IncompleteUnprovable |
|
blanchet@43458
|
18 |
ProofMissing |
|
blanchet@43751
|
19 |
ProofIncomplete |
|
blanchet@43745
|
20 |
UnsoundProof of bool * string list |
|
blanchet@43458
|
21 |
CantConnect |
|
blanchet@43458
|
22 |
TimedOut |
|
blanchet@43794
|
23 |
Inappropriate |
|
blanchet@43458
|
24 |
OutOfResources |
|
blanchet@43458
|
25 |
SpassTooOld |
|
blanchet@43458
|
26 |
VampireTooOld |
|
blanchet@43458
|
27 |
NoPerl |
|
blanchet@43458
|
28 |
NoLibwwwPerl |
|
blanchet@43458
|
29 |
NoRealZ3 |
|
blanchet@43458
|
30 |
MalformedInput |
|
blanchet@43458
|
31 |
MalformedOutput |
|
blanchet@43458
|
32 |
Interrupted |
|
blanchet@43458
|
33 |
Crashed |
|
blanchet@43458
|
34 |
InternalError |
|
blanchet@43458
|
35 |
UnknownError of string
|
blanchet@39731
|
36 |
|
blanchet@39695
|
37 |
type step_name = string * string option
|
blanchet@39692
|
38 |
|
blanchet@39693
|
39 |
datatype 'a step =
|
blanchet@39693
|
40 |
Definition of step_name * 'a * 'a |
|
blanchet@39693
|
41 |
Inference of step_name * 'a * step_name list
|
blanchet@39692
|
42 |
|
blanchet@43402
|
43 |
type 'a proof = ('a, 'a, 'a fo_term) formula step list
|
blanchet@39692
|
44 |
|
blanchet@43802
|
45 |
val strip_spaces : bool -> (char -> bool) -> string -> string
|
blanchet@41505
|
46 |
val short_output : bool -> string -> string
|
blanchet@42615
|
47 |
val string_for_failure : failure -> string
|
blanchet@39731
|
48 |
val extract_important_message : string -> string
|
blanchet@39731
|
49 |
val extract_known_failure :
|
blanchet@39731
|
50 |
(failure * string) list -> string -> failure option
|
blanchet@39731
|
51 |
val extract_tstplike_proof_and_outcome :
|
blanchet@43717
|
52 |
bool -> bool -> int -> (string * string) list -> (failure * string) list
|
blanchet@43717
|
53 |
-> string -> string * failure option
|
blanchet@39692
|
54 |
val is_same_step : step_name * step_name -> bool
|
blanchet@43802
|
55 |
val scan_general_id : string list -> string * string list
|
blanchet@43802
|
56 |
val parse_formula :
|
blanchet@43802
|
57 |
string list -> (string, 'a, string fo_term) formula * string list
|
blanchet@43784
|
58 |
val atp_proof_from_tstplike_proof : string problem -> string -> string proof
|
blanchet@39694
|
59 |
val map_term_names_in_atp_proof :
|
blanchet@39694
|
60 |
(string -> string) -> string proof -> string proof
|
blanchet@39694
|
61 |
val nasty_atp_proof : string Symtab.table -> string proof -> string proof
|
blanchet@39692
|
62 |
end;
|
blanchet@39692
|
63 |
|
blanchet@39692
|
64 |
structure ATP_Proof : ATP_PROOF =
|
blanchet@39692
|
65 |
struct
|
blanchet@39692
|
66 |
|
blanchet@39731
|
67 |
open ATP_Problem
|
blanchet@39731
|
68 |
|
blanchet@39731
|
69 |
datatype failure =
|
blanchet@43458
|
70 |
Unprovable |
|
blanchet@43458
|
71 |
IncompleteUnprovable |
|
blanchet@43458
|
72 |
ProofMissing |
|
blanchet@43751
|
73 |
ProofIncomplete |
|
blanchet@43745
|
74 |
UnsoundProof of bool * string list |
|
blanchet@43458
|
75 |
CantConnect |
|
blanchet@43458
|
76 |
TimedOut |
|
blanchet@43794
|
77 |
Inappropriate |
|
blanchet@43458
|
78 |
OutOfResources |
|
blanchet@43458
|
79 |
SpassTooOld |
|
blanchet@43458
|
80 |
VampireTooOld |
|
blanchet@43458
|
81 |
NoPerl |
|
blanchet@43458
|
82 |
NoLibwwwPerl |
|
blanchet@43458
|
83 |
NoRealZ3 |
|
blanchet@43458
|
84 |
MalformedInput |
|
blanchet@43458
|
85 |
MalformedOutput |
|
blanchet@43458
|
86 |
Interrupted |
|
blanchet@43458
|
87 |
Crashed |
|
blanchet@43458
|
88 |
InternalError |
|
blanchet@43458
|
89 |
UnknownError of string
|
blanchet@39731
|
90 |
|
blanchet@43802
|
91 |
fun strip_c_style_comment _ [] = []
|
blanchet@43802
|
92 |
| strip_c_style_comment is_evil (#"*" :: #"/" :: cs) =
|
blanchet@43802
|
93 |
strip_spaces_in_list true is_evil cs
|
blanchet@43802
|
94 |
| strip_c_style_comment is_evil (_ :: cs) = strip_c_style_comment is_evil cs
|
blanchet@43802
|
95 |
and strip_spaces_in_list _ _ [] = []
|
blanchet@43802
|
96 |
| strip_spaces_in_list true is_evil (#"%" :: cs) =
|
blanchet@43802
|
97 |
strip_spaces_in_list true is_evil
|
blanchet@43802
|
98 |
(cs |> chop_while (not_equal #"\n") |> snd)
|
blanchet@43802
|
99 |
| strip_spaces_in_list true is_evil (#"/" :: #"*" :: cs) =
|
blanchet@43802
|
100 |
strip_c_style_comment is_evil cs
|
blanchet@43802
|
101 |
| strip_spaces_in_list _ _ [c1] = if Char.isSpace c1 then [] else [str c1]
|
blanchet@43802
|
102 |
| strip_spaces_in_list skip_comments is_evil [c1, c2] =
|
blanchet@43802
|
103 |
strip_spaces_in_list skip_comments is_evil [c1] @
|
blanchet@43802
|
104 |
strip_spaces_in_list skip_comments is_evil [c2]
|
blanchet@43802
|
105 |
| strip_spaces_in_list skip_comments is_evil (c1 :: c2 :: c3 :: cs) =
|
blanchet@39692
|
106 |
if Char.isSpace c1 then
|
blanchet@43802
|
107 |
strip_spaces_in_list skip_comments is_evil (c2 :: c3 :: cs)
|
blanchet@39692
|
108 |
else if Char.isSpace c2 then
|
blanchet@39692
|
109 |
if Char.isSpace c3 then
|
blanchet@43802
|
110 |
strip_spaces_in_list skip_comments is_evil (c1 :: c3 :: cs)
|
blanchet@39692
|
111 |
else
|
blanchet@39692
|
112 |
str c1 :: (if forall is_evil [c1, c3] then [" "] else []) @
|
blanchet@43802
|
113 |
strip_spaces_in_list skip_comments is_evil (c3 :: cs)
|
blanchet@39692
|
114 |
else
|
blanchet@43802
|
115 |
str c1 :: strip_spaces_in_list skip_comments is_evil (c2 :: c3 :: cs)
|
blanchet@43802
|
116 |
fun strip_spaces skip_comments is_evil =
|
blanchet@43802
|
117 |
implode o strip_spaces_in_list skip_comments is_evil o String.explode
|
blanchet@39692
|
118 |
|
blanchet@39692
|
119 |
fun is_ident_char c = Char.isAlphaNum c orelse c = #"_"
|
blanchet@43802
|
120 |
val strip_spaces_except_between_ident_chars = strip_spaces true is_ident_char
|
blanchet@39692
|
121 |
|
blanchet@41505
|
122 |
fun elide_string threshold s =
|
blanchet@41505
|
123 |
if size s > threshold then
|
blanchet@41505
|
124 |
String.extract (s, 0, SOME (threshold div 2 - 5)) ^ " ...... " ^
|
blanchet@41505
|
125 |
String.extract (s, size s - (threshold + 1) div 2 + 6, NONE)
|
blanchet@41505
|
126 |
else
|
blanchet@41505
|
127 |
s
|
blanchet@41505
|
128 |
fun short_output verbose output =
|
blanchet@42915
|
129 |
if verbose then
|
blanchet@42915
|
130 |
if output = "" then "No details available" else elide_string 1000 output
|
blanchet@42915
|
131 |
else
|
blanchet@42915
|
132 |
""
|
blanchet@41505
|
133 |
|
blanchet@42615
|
134 |
val missing_message_tail =
|
blanchet@42615
|
135 |
" appears to be missing. You will need to install it if you want to invoke \
|
blanchet@42615
|
136 |
\remote provers."
|
blanchet@39731
|
137 |
|
blanchet@43745
|
138 |
fun involving [] = ""
|
blanchet@43745
|
139 |
| involving ss = "involving " ^ commas_quote ss ^ " "
|
blanchet@43745
|
140 |
|
blanchet@42615
|
141 |
fun string_for_failure Unprovable =
|
blanchet@42615
|
142 |
"The problem is unprovable."
|
blanchet@42615
|
143 |
| string_for_failure IncompleteUnprovable =
|
blanchet@42615
|
144 |
"The prover gave up."
|
blanchet@42615
|
145 |
| string_for_failure ProofMissing =
|
blanchet@42615
|
146 |
"The prover claims the conjecture is a theorem but did not provide a proof."
|
blanchet@43751
|
147 |
| string_for_failure ProofIncomplete =
|
blanchet@43751
|
148 |
"The prover claims the conjecture is a theorem but provided an incomplete \
|
blanchet@43751
|
149 |
\proof."
|
blanchet@43745
|
150 |
| string_for_failure (UnsoundProof (false, ss)) =
|
blanchet@43745
|
151 |
"The prover found a type-unsound proof " ^ involving ss ^
|
blanchet@43745
|
152 |
"(or, less likely, your axioms are inconsistent). Try passing the \
|
blanchet@43745
|
153 |
\\"full_types\" option to Sledgehammer to avoid such spurious proofs."
|
blanchet@43745
|
154 |
| string_for_failure (UnsoundProof (true, ss)) =
|
blanchet@43745
|
155 |
"The prover found a type-unsound proof " ^ involving ss ^
|
blanchet@43745
|
156 |
"even though a supposedly type-sound encoding was used (or, less likely, \
|
blanchet@43745
|
157 |
\your axioms are inconsistent). You might want to report this to the \
|
blanchet@43745
|
158 |
\Isabelle developers."
|
blanchet@42615
|
159 |
| string_for_failure CantConnect = "Cannot connect to remote server."
|
blanchet@42615
|
160 |
| string_for_failure TimedOut = "Timed out."
|
blanchet@43794
|
161 |
| string_for_failure Inappropriate =
|
blanchet@43794
|
162 |
"The problem lies outside the prover's scope."
|
blanchet@42615
|
163 |
| string_for_failure OutOfResources = "The prover ran out of resources."
|
blanchet@42615
|
164 |
| string_for_failure SpassTooOld =
|
blanchet@39731
|
165 |
"Isabelle requires a more recent version of SPASS with support for the \
|
blanchet@39731
|
166 |
\TPTP syntax. To install it, download and extract the package \
|
blanchet@39731
|
167 |
\\"http://isabelle.in.tum.de/dist/contrib/spass-3.7.tar.gz\" and add the \
|
blanchet@39731
|
168 |
\\"spass-3.7\" directory's absolute path to " ^
|
wenzelm@42815
|
169 |
Path.print (Path.expand (Path.appends
|
blanchet@39731
|
170 |
(Path.variable "ISABELLE_HOME_USER" ::
|
wenzelm@42815
|
171 |
map Path.basic ["etc", "components"]))) ^
|
blanchet@39731
|
172 |
" on a line of its own."
|
blanchet@42615
|
173 |
| string_for_failure VampireTooOld =
|
blanchet@39731
|
174 |
"Isabelle requires a more recent version of Vampire. To install it, follow \
|
blanchet@39731
|
175 |
\the instructions from the Sledgehammer manual (\"isabelle doc\
|
blanchet@39731
|
176 |
\ sledgehammer\")."
|
blanchet@42615
|
177 |
| string_for_failure NoPerl = "Perl" ^ missing_message_tail
|
blanchet@42615
|
178 |
| string_for_failure NoLibwwwPerl =
|
blanchet@42615
|
179 |
"The Perl module \"libwww-perl\"" ^ missing_message_tail
|
blanchet@42615
|
180 |
| string_for_failure NoRealZ3 =
|
blanchet@41470
|
181 |
"The environment variable \"Z3_REAL_SOLVER\" must be set to Z3's full path."
|
blanchet@42615
|
182 |
| string_for_failure MalformedInput =
|
blanchet@42615
|
183 |
"The generated problem is malformed. Please report this to the Isabelle \
|
blanchet@42615
|
184 |
\developers."
|
blanchet@42615
|
185 |
| string_for_failure MalformedOutput = "The prover output is malformed."
|
blanchet@42615
|
186 |
| string_for_failure Crashed = "The prover crashed."
|
blanchet@42615
|
187 |
| string_for_failure InternalError = "An internal prover error occurred."
|
blanchet@42615
|
188 |
| string_for_failure (UnknownError string) =
|
blanchet@42615
|
189 |
"A prover error occurred" ^
|
blanchet@41582
|
190 |
(if string = "" then ". (Pass the \"verbose\" option for details.)"
|
blanchet@41582
|
191 |
else ":\n" ^ string)
|
blanchet@39731
|
192 |
|
blanchet@39731
|
193 |
fun extract_delimited (begin_delim, end_delim) output =
|
blanchet@39731
|
194 |
output |> first_field begin_delim |> the |> snd
|
blanchet@39731
|
195 |
|> first_field end_delim |> the |> fst
|
blanchet@39731
|
196 |
|> first_field "\n" |> the |> snd
|
blanchet@39731
|
197 |
handle Option.Option => ""
|
blanchet@39731
|
198 |
|
blanchet@39731
|
199 |
val tstp_important_message_delims =
|
blanchet@39731
|
200 |
("% SZS start RequiredInformation", "% SZS end RequiredInformation")
|
blanchet@39731
|
201 |
|
blanchet@39731
|
202 |
fun extract_important_message output =
|
blanchet@39731
|
203 |
case extract_delimited tstp_important_message_delims output of
|
blanchet@39731
|
204 |
"" => ""
|
blanchet@39731
|
205 |
| s => s |> space_explode "\n" |> filter_out (curry (op =) "")
|
blanchet@39731
|
206 |
|> map (perhaps (try (unprefix "%")))
|
blanchet@39731
|
207 |
|> map (perhaps (try (unprefix " ")))
|
blanchet@39731
|
208 |
|> space_implode "\n " |> quote
|
blanchet@39731
|
209 |
|
blanchet@39731
|
210 |
(* Splits by the first possible of a list of delimiters. *)
|
blanchet@39731
|
211 |
fun extract_tstplike_proof delims output =
|
blanchet@39731
|
212 |
case pairself (find_first (fn s => String.isSubstring s output))
|
blanchet@39731
|
213 |
(ListPair.unzip delims) of
|
blanchet@39731
|
214 |
(SOME begin_delim, SOME end_delim) =>
|
blanchet@39731
|
215 |
extract_delimited (begin_delim, end_delim) output
|
blanchet@39731
|
216 |
| _ => ""
|
blanchet@39731
|
217 |
|
blanchet@39731
|
218 |
fun extract_known_failure known_failures output =
|
blanchet@39731
|
219 |
known_failures
|
blanchet@39731
|
220 |
|> find_first (fn (_, pattern) => String.isSubstring pattern output)
|
blanchet@39731
|
221 |
|> Option.map fst
|
blanchet@39731
|
222 |
|
blanchet@43717
|
223 |
fun extract_tstplike_proof_and_outcome verbose complete res_code proof_delims
|
blanchet@43717
|
224 |
known_failures output =
|
blanchet@43751
|
225 |
case (extract_tstplike_proof proof_delims output,
|
blanchet@43751
|
226 |
extract_known_failure known_failures output) of
|
blanchet@43751
|
227 |
(_, SOME ProofIncomplete) => ("", SOME ProofIncomplete)
|
blanchet@43751
|
228 |
| ("", SOME failure) =>
|
blanchet@43751
|
229 |
("", SOME (if failure = IncompleteUnprovable andalso complete then Unprovable
|
blanchet@43751
|
230 |
else failure))
|
blanchet@43751
|
231 |
| ("", NONE) =>
|
blanchet@43751
|
232 |
("", SOME (if res_code = 0 andalso output = "" then ProofMissing
|
blanchet@43751
|
233 |
else UnknownError (short_output verbose output)))
|
blanchet@43751
|
234 |
| (tstplike_proof, _) => (tstplike_proof, NONE)
|
blanchet@39692
|
235 |
|
blanchet@39695
|
236 |
type step_name = string * string option
|
blanchet@39692
|
237 |
|
blanchet@39695
|
238 |
fun is_same_step p = p |> pairself fst |> op =
|
blanchet@39692
|
239 |
|
blanchet@39692
|
240 |
fun step_name_ord p =
|
blanchet@39695
|
241 |
let val q = pairself fst p in
|
blanchet@39692
|
242 |
(* The "unprefix" part is to cope with remote Vampire's output. The proper
|
blanchet@39692
|
243 |
solution would be to perform a topological sort, e.g. using the nice
|
blanchet@39692
|
244 |
"Graph" functor. *)
|
blanchet@39692
|
245 |
case pairself (Int.fromString o perhaps (try (unprefix "f"))) q of
|
blanchet@39692
|
246 |
(NONE, NONE) => string_ord q
|
blanchet@39692
|
247 |
| (NONE, SOME _) => LESS
|
blanchet@39692
|
248 |
| (SOME _, NONE) => GREATER
|
blanchet@39692
|
249 |
| (SOME i, SOME j) => int_ord (i, j)
|
blanchet@39692
|
250 |
end
|
blanchet@39692
|
251 |
|
blanchet@39693
|
252 |
datatype 'a step =
|
blanchet@39693
|
253 |
Definition of step_name * 'a * 'a |
|
blanchet@39693
|
254 |
Inference of step_name * 'a * step_name list
|
blanchet@39692
|
255 |
|
blanchet@43402
|
256 |
type 'a proof = ('a, 'a, 'a fo_term) formula step list
|
blanchet@39692
|
257 |
|
blanchet@39692
|
258 |
fun step_name (Definition (name, _, _)) = name
|
blanchet@39692
|
259 |
| step_name (Inference (name, _, _)) = name
|
blanchet@39692
|
260 |
|
blanchet@39692
|
261 |
(**** PARSING OF TSTP FORMAT ****)
|
blanchet@39692
|
262 |
|
blanchet@43407
|
263 |
(* Strings enclosed in single quotes (e.g., file names) *)
|
blanchet@39692
|
264 |
val scan_general_id =
|
blanchet@39692
|
265 |
$$ "'" |-- Scan.repeat (~$$ "'") --| $$ "'" >> implode
|
blanchet@39692
|
266 |
|| Scan.repeat ($$ "$") -- Scan.many1 Symbol.is_letdig
|
blanchet@39692
|
267 |
>> (fn (ss1, ss2) => implode ss1 ^ implode ss2)
|
blanchet@39692
|
268 |
|
blanchet@39692
|
269 |
(* Generalized first-order terms, which include file names, numbers, etc. *)
|
blanchet@43465
|
270 |
fun parse_annotation x =
|
blanchet@43481
|
271 |
((scan_general_id ::: Scan.repeat ($$ " " |-- scan_general_id))
|
blanchet@43481
|
272 |
-- Scan.optional parse_annotation [] >> op @
|
blanchet@43465
|
273 |
|| $$ "(" |-- parse_annotations --| $$ ")"
|
blanchet@43465
|
274 |
|| $$ "[" |-- parse_annotations --| $$ "]") x
|
blanchet@43465
|
275 |
and parse_annotations x =
|
blanchet@43465
|
276 |
(Scan.optional (parse_annotation
|
blanchet@43465
|
277 |
::: Scan.repeat ($$ "," |-- parse_annotation)) []
|
blanchet@39692
|
278 |
>> flat) x
|
blanchet@39692
|
279 |
|
blanchet@39694
|
280 |
fun parse_term x =
|
blanchet@39694
|
281 |
(scan_general_id
|
blanchet@43414
|
282 |
--| Scan.option ($$ ":" -- scan_general_id) (* ignore TFF types for now *)
|
blanchet@43613
|
283 |
-- Scan.optional ($$ "(" |-- parse_terms --| $$ ")") []
|
blanchet@39692
|
284 |
>> ATerm) x
|
blanchet@39694
|
285 |
and parse_terms x = (parse_term ::: Scan.repeat ($$ "," |-- parse_term)) x
|
blanchet@39692
|
286 |
|
blanchet@39822
|
287 |
fun parse_atom x =
|
blanchet@39822
|
288 |
(parse_term -- Scan.option (Scan.option ($$ "!") --| $$ "=" -- parse_term)
|
blanchet@39822
|
289 |
>> (fn (u1, NONE) => AAtom u1
|
blanchet@43784
|
290 |
| (u1, SOME (NONE, u2)) => AAtom (ATerm ("equal", [u1, u2]))
|
blanchet@39822
|
291 |
| (u1, SOME (SOME _, u2)) =>
|
blanchet@43784
|
292 |
mk_anot (AAtom (ATerm ("equal", [u1, u2]))))) x
|
blanchet@39692
|
293 |
|
blanchet@39692
|
294 |
fun fo_term_head (ATerm (s, _)) = s
|
blanchet@39692
|
295 |
|
blanchet@39692
|
296 |
(* TPTP formulas are fully parenthesized, so we don't need to worry about
|
blanchet@39692
|
297 |
operator precedence. *)
|
blanchet@43476
|
298 |
fun parse_literal x =
|
blanchet@43476
|
299 |
((Scan.repeat ($$ "~") >> length)
|
blanchet@43476
|
300 |
-- ($$ "(" |-- parse_formula --| $$ ")"
|
blanchet@43476
|
301 |
|| parse_quantified_formula
|
blanchet@43476
|
302 |
|| parse_atom)
|
blanchet@43476
|
303 |
>> (fn (n, phi) => phi |> n mod 2 = 1 ? mk_anot)) x
|
blanchet@43476
|
304 |
and parse_formula x =
|
blanchet@43476
|
305 |
(parse_literal
|
blanchet@39692
|
306 |
-- Scan.option ((Scan.this_string "=>" >> K AImplies
|
blanchet@39692
|
307 |
|| Scan.this_string "<=>" >> K AIff
|
blanchet@39692
|
308 |
|| Scan.this_string "<~>" >> K ANotIff
|
blanchet@39692
|
309 |
|| Scan.this_string "<=" >> K AIf
|
blanchet@43476
|
310 |
|| $$ "|" >> K AOr
|
blanchet@43476
|
311 |
|| $$ "&" >> K AAnd)
|
blanchet@39694
|
312 |
-- parse_formula)
|
blanchet@39692
|
313 |
>> (fn (phi1, NONE) => phi1
|
blanchet@43784
|
314 |
| (phi1, SOME (c, phi2)) => mk_aconn c phi1 phi2)) x
|
blanchet@43476
|
315 |
and parse_quantified_formula x =
|
blanchet@43476
|
316 |
(($$ "!" >> K AForall || $$ "?" >> K AExists)
|
blanchet@43476
|
317 |
--| $$ "[" -- parse_terms --| $$ "]" --| $$ ":" -- parse_literal
|
blanchet@43476
|
318 |
>> (fn ((q, ts), phi) =>
|
blanchet@43476
|
319 |
(* FIXME: TFF *)
|
blanchet@43476
|
320 |
AQuant (q, map (rpair NONE o fo_term_head) ts, phi))) x
|
blanchet@39692
|
321 |
|
blanchet@39692
|
322 |
val parse_tstp_extra_arguments =
|
blanchet@43465
|
323 |
Scan.optional ($$ "," |-- parse_annotation
|
blanchet@43465
|
324 |
--| Scan.option ($$ "," |-- parse_annotations)) []
|
blanchet@39692
|
325 |
|
blanchet@41451
|
326 |
val vampire_unknown_fact = "unknown"
|
blanchet@43784
|
327 |
val waldmeister_conjecture = "conjecture_1"
|
blanchet@43784
|
328 |
|
blanchet@43407
|
329 |
val tofof_fact_prefix = "fof_"
|
blanchet@41451
|
330 |
|
blanchet@43784
|
331 |
fun is_same_term subst tm1 tm2 =
|
blanchet@43784
|
332 |
let
|
blanchet@43784
|
333 |
fun do_term_pair _ NONE = NONE
|
blanchet@43784
|
334 |
| do_term_pair (ATerm (s1, tm1), ATerm (s2, tm2)) (SOME subst) =
|
blanchet@43784
|
335 |
case pairself is_atp_variable (s1, s2) of
|
blanchet@43784
|
336 |
(true, true) =>
|
blanchet@43784
|
337 |
(case AList.lookup (op =) subst s1 of
|
blanchet@43784
|
338 |
SOME s2' => if s2' = s2 then SOME subst else NONE
|
blanchet@43784
|
339 |
| NONE =>
|
blanchet@43784
|
340 |
if null (AList.find (op =) subst s2) then SOME ((s1, s2) :: subst)
|
blanchet@43784
|
341 |
else NONE)
|
blanchet@43784
|
342 |
| (false, false) =>
|
blanchet@43784
|
343 |
if s1 = s2 andalso length tm1 = length tm2 then
|
blanchet@43784
|
344 |
SOME subst |> fold do_term_pair (tm1 ~~ tm2)
|
blanchet@43784
|
345 |
else
|
blanchet@43784
|
346 |
NONE
|
blanchet@43784
|
347 |
| _ => NONE
|
blanchet@43784
|
348 |
in SOME subst |> do_term_pair (tm1, tm2) |> is_some end
|
blanchet@43784
|
349 |
|
blanchet@43784
|
350 |
fun is_same_formula subst (AQuant (q1, xs1, phi1)) (AQuant (q2, xs2, phi2)) =
|
blanchet@43784
|
351 |
q1 = q2 andalso length xs1 = length xs2 andalso
|
blanchet@43784
|
352 |
is_same_formula ((map fst xs1 ~~ map fst xs2) @ subst) phi1 phi2
|
blanchet@43784
|
353 |
| is_same_formula subst (AConn (c1, phis1)) (AConn (c2, phis2)) =
|
blanchet@43784
|
354 |
c1 = c2 andalso length phis1 = length phis2 andalso
|
blanchet@43784
|
355 |
forall (uncurry (is_same_formula subst)) (phis1 ~~ phis2)
|
blanchet@43784
|
356 |
| is_same_formula subst (AAtom (ATerm ("equal", [tm11, tm12]))) (AAtom tm2) =
|
blanchet@43784
|
357 |
is_same_term subst (ATerm ("equal", [tm11, tm12])) tm2 orelse
|
blanchet@43784
|
358 |
is_same_term subst (ATerm ("equal", [tm12, tm11])) tm2
|
blanchet@43784
|
359 |
| is_same_formula subst (AAtom tm1) (AAtom tm2) = is_same_term subst tm1 tm2
|
blanchet@43784
|
360 |
| is_same_formula _ _ _ = false
|
blanchet@43784
|
361 |
|
blanchet@43784
|
362 |
fun matching_formula_line_identifier phi (Formula (ident, _, phi', _, _)) =
|
blanchet@43784
|
363 |
if is_same_formula [] phi phi' then SOME ident else NONE
|
blanchet@43784
|
364 |
| matching_formula_line_identifier _ _ = NONE
|
blanchet@43784
|
365 |
|
blanchet@43784
|
366 |
fun find_formula_in_problem problem phi =
|
blanchet@43784
|
367 |
problem |> maps snd |> map_filter (matching_formula_line_identifier phi)
|
blanchet@43784
|
368 |
|> try hd
|
blanchet@43784
|
369 |
|
blanchet@43396
|
370 |
(* Syntax: (cnf|fof|tff)\(<num>, <formula_role>, <formula> <extra_arguments>\).
|
blanchet@39692
|
371 |
The <num> could be an identifier, but we assume integers. *)
|
blanchet@43784
|
372 |
fun parse_tstp_line problem =
|
blanchet@43784
|
373 |
((Scan.this_string "cnf" || Scan.this_string "fof" || Scan.this_string "tff")
|
blanchet@43784
|
374 |
-- $$ "(")
|
blanchet@43784
|
375 |
|-- scan_general_id --| $$ "," -- Symbol.scan_id --| $$ ","
|
blanchet@43784
|
376 |
-- parse_formula -- parse_tstp_extra_arguments --| $$ ")" --| $$ "."
|
blanchet@43784
|
377 |
>> (fn (((num, role), phi), deps) =>
|
blanchet@43784
|
378 |
let
|
blanchet@43784
|
379 |
val (name, deps) =
|
blanchet@43784
|
380 |
(* Waldmeister isn't exactly helping. *)
|
blanchet@43784
|
381 |
case deps of
|
blanchet@43784
|
382 |
["file", _, s] =>
|
blanchet@43784
|
383 |
((num,
|
blanchet@43784
|
384 |
if s = vampire_unknown_fact then
|
blanchet@43784
|
385 |
NONE
|
blanchet@43784
|
386 |
else if s = waldmeister_conjecture then
|
blanchet@43784
|
387 |
find_formula_in_problem problem (mk_anot phi)
|
blanchet@43784
|
388 |
else
|
blanchet@43784
|
389 |
SOME (s |> perhaps (try (unprefix tofof_fact_prefix)))),
|
blanchet@43784
|
390 |
[])
|
blanchet@43784
|
391 |
| ["file", _] => ((num, find_formula_in_problem problem phi), [])
|
blanchet@43784
|
392 |
| _ => ((num, NONE), deps)
|
blanchet@43784
|
393 |
in
|
blanchet@43784
|
394 |
case role of
|
blanchet@43784
|
395 |
"definition" =>
|
blanchet@43784
|
396 |
(case phi of
|
blanchet@43784
|
397 |
AConn (AIff, [phi1 as AAtom _, phi2]) =>
|
blanchet@43784
|
398 |
Definition (name, phi1, phi2)
|
blanchet@43784
|
399 |
| AAtom (ATerm ("equal", _)) =>
|
blanchet@43784
|
400 |
(* Vampire's equality proxy axiom *)
|
blanchet@43784
|
401 |
Inference (name, phi, map (rpair NONE) deps)
|
blanchet@43784
|
402 |
| _ => raise Fail "malformed definition")
|
blanchet@43784
|
403 |
| _ => Inference (name, phi, map (rpair NONE) deps)
|
blanchet@43784
|
404 |
end)
|
blanchet@39692
|
405 |
|
blanchet@39692
|
406 |
(**** PARSING OF SPASS OUTPUT ****)
|
blanchet@39692
|
407 |
|
blanchet@39692
|
408 |
(* SPASS returns clause references of the form "x.y". We ignore "y", whose role
|
blanchet@39692
|
409 |
is not clear anyway. *)
|
blanchet@39692
|
410 |
val parse_dot_name = scan_general_id --| $$ "." --| scan_general_id
|
blanchet@39692
|
411 |
|
blanchet@39692
|
412 |
val parse_spass_annotations =
|
blanchet@39692
|
413 |
Scan.optional ($$ ":" |-- Scan.repeat (parse_dot_name
|
blanchet@39692
|
414 |
--| Scan.option ($$ ","))) []
|
blanchet@39692
|
415 |
|
blanchet@39692
|
416 |
(* It is not clear why some literals are followed by sequences of stars and/or
|
blanchet@39692
|
417 |
pluses. We ignore them. *)
|
blanchet@39826
|
418 |
fun parse_decorated_atom x =
|
blanchet@39826
|
419 |
(parse_atom --| Scan.repeat ($$ "*" || $$ "+" || $$ " ")) x
|
blanchet@39692
|
420 |
|
blanchet@39692
|
421 |
fun mk_horn ([], []) = AAtom (ATerm ("c_False", []))
|
blanchet@43784
|
422 |
| mk_horn ([], pos_lits) = foldr1 (uncurry (mk_aconn AOr)) pos_lits
|
blanchet@43784
|
423 |
| mk_horn (neg_lits, []) = mk_anot (foldr1 (uncurry (mk_aconn AAnd)) neg_lits)
|
blanchet@39692
|
424 |
| mk_horn (neg_lits, pos_lits) =
|
blanchet@43784
|
425 |
mk_aconn AImplies (foldr1 (uncurry (mk_aconn AAnd)) neg_lits)
|
blanchet@43784
|
426 |
(foldr1 (uncurry (mk_aconn AOr)) pos_lits)
|
blanchet@39692
|
427 |
|
blanchet@39869
|
428 |
fun parse_horn_clause x =
|
blanchet@39869
|
429 |
(Scan.repeat parse_decorated_atom --| $$ "|" --| $$ "|"
|
blanchet@39869
|
430 |
-- Scan.repeat parse_decorated_atom --| $$ "-" --| $$ ">"
|
blanchet@39869
|
431 |
-- Scan.repeat parse_decorated_atom
|
blanchet@39869
|
432 |
>> (mk_horn o apfst (op @))) x
|
blanchet@39692
|
433 |
|
blanchet@39692
|
434 |
(* Syntax: <num>[0:<inference><annotations>]
|
blanchet@39692
|
435 |
<atoms> || <atoms> -> <atoms>. *)
|
blanchet@39869
|
436 |
fun parse_spass_line x =
|
blanchet@39869
|
437 |
(scan_general_id --| $$ "[" --| $$ "0" --| $$ ":" --| Symbol.scan_id
|
blanchet@39869
|
438 |
-- parse_spass_annotations --| $$ "]" -- parse_horn_clause --| $$ "."
|
blanchet@39869
|
439 |
>> (fn ((num, deps), u) =>
|
blanchet@39869
|
440 |
Inference ((num, NONE), u, map (rpair NONE) deps))) x
|
blanchet@39692
|
441 |
|
blanchet@43784
|
442 |
fun parse_line problem = parse_tstp_line problem || parse_spass_line
|
blanchet@43784
|
443 |
fun parse_proof problem s =
|
blanchet@43519
|
444 |
s |> strip_spaces_except_between_ident_chars
|
blanchet@43519
|
445 |
|> raw_explode
|
blanchet@43519
|
446 |
|> Scan.finite Symbol.stopper
|
blanchet@43519
|
447 |
(Scan.error (!! (fn _ => raise Fail "unrecognized ATP output")
|
blanchet@43784
|
448 |
(Scan.repeat1 (parse_line problem))))
|
blanchet@43519
|
449 |
|> fst
|
blanchet@39692
|
450 |
|
blanchet@39692
|
451 |
fun clean_up_dependency seen dep = find_first (curry is_same_step dep) seen
|
blanchet@39692
|
452 |
fun clean_up_dependencies _ [] = []
|
blanchet@39692
|
453 |
| clean_up_dependencies seen ((step as Definition (name, _, _)) :: steps) =
|
blanchet@39692
|
454 |
step :: clean_up_dependencies (name :: seen) steps
|
blanchet@39692
|
455 |
| clean_up_dependencies seen (Inference (name, u, deps) :: steps) =
|
blanchet@39692
|
456 |
Inference (name, u, map_filter (clean_up_dependency seen) deps) ::
|
blanchet@39692
|
457 |
clean_up_dependencies (name :: seen) steps
|
blanchet@39692
|
458 |
|
blanchet@43784
|
459 |
fun atp_proof_from_tstplike_proof _ "" = []
|
blanchet@43784
|
460 |
| atp_proof_from_tstplike_proof problem s =
|
blanchet@43320
|
461 |
s ^ "$" (* the $ sign acts as a sentinel (FIXME: needed?) *)
|
blanchet@43784
|
462 |
|> parse_proof problem
|
blanchet@43320
|
463 |
|> sort (step_name_ord o pairself step_name)
|
blanchet@43320
|
464 |
|> clean_up_dependencies []
|
blanchet@39692
|
465 |
|
blanchet@39694
|
466 |
fun map_term_names_in_term f (ATerm (s, ts)) =
|
blanchet@39694
|
467 |
ATerm (f s, map (map_term_names_in_term f) ts)
|
blanchet@39694
|
468 |
fun map_term_names_in_formula f (AQuant (q, xs, phi)) =
|
blanchet@39694
|
469 |
AQuant (q, xs, map_term_names_in_formula f phi)
|
blanchet@39694
|
470 |
| map_term_names_in_formula f (AConn (c, phis)) =
|
blanchet@39694
|
471 |
AConn (c, map (map_term_names_in_formula f) phis)
|
blanchet@39694
|
472 |
| map_term_names_in_formula f (AAtom t) = AAtom (map_term_names_in_term f t)
|
blanchet@39694
|
473 |
fun map_term_names_in_step f (Definition (name, phi1, phi2)) =
|
blanchet@39694
|
474 |
Definition (name, map_term_names_in_formula f phi1,
|
blanchet@39694
|
475 |
map_term_names_in_formula f phi2)
|
blanchet@39694
|
476 |
| map_term_names_in_step f (Inference (name, phi, deps)) =
|
blanchet@39694
|
477 |
Inference (name, map_term_names_in_formula f phi, deps)
|
blanchet@39694
|
478 |
fun map_term_names_in_atp_proof f = map (map_term_names_in_step f)
|
blanchet@39694
|
479 |
|
blanchet@39694
|
480 |
fun nasty_name pool s = s |> Symtab.lookup pool |> the_default s
|
blanchet@39694
|
481 |
fun nasty_atp_proof pool =
|
blanchet@39694
|
482 |
if Symtab.is_empty pool then I
|
blanchet@39694
|
483 |
else map_term_names_in_atp_proof (nasty_name pool)
|
blanchet@39694
|
484 |
|
blanchet@39692
|
485 |
end;
|