author | blanchet |
Mon, 21 May 2012 11:31:52 +0200 | |
changeset 47947 | 7b482cc7473e |
parent 47946 | 33afcfad3f8d |
child 47950 | 9cb132898ac8 |
permissions | -rw-r--r-- |
39452 | 1 |
(* Title: HOL/Tools/ATP/atp_proof.ML |
2 |
Author: Lawrence C. Paulson, Cambridge University Computer Laboratory |
|
3 |
Author: Claire Quigley, Cambridge University Computer Laboratory |
|
4 |
Author: Jasmin Blanchette, TU Muenchen |
|
5 |
||
42876
e336ef6313aa
more informative message when Sledgehammer finds an unsound proof
blanchet
parents:
42848
diff
changeset
|
6 |
Abstract representation of ATP proofs and TSTP/SPASS syntax. |
39452 | 7 |
*) |
8 |
||
9 |
signature ATP_PROOF = |
|
10 |
sig |
|
43678 | 11 |
type ('a, 'b) ho_term = ('a, 'b) ATP_Problem.ho_term |
42531
a462dbaa584f
added more rudimentary type support to Sledgehammer's ATP encoding
blanchet
parents:
42526
diff
changeset
|
12 |
type ('a, 'b, 'c) formula = ('a, 'b, 'c) ATP_Problem.formula |
42943 | 13 |
type 'a problem = 'a ATP_Problem.problem |
39452 | 14 |
|
42965
1403595ec38c
slightly gracefuller handling of LEO-II and Satallax output
blanchet
parents:
42962
diff
changeset
|
15 |
exception UNRECOGNIZED_ATP_PROOF of unit |
1403595ec38c
slightly gracefuller handling of LEO-II and Satallax output
blanchet
parents:
42962
diff
changeset
|
16 |
|
39491
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
17 |
datatype failure = |
42587
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
18 |
Unprovable | |
43050
59284a13abc4
support "metis" and "metisFT" as provers in the architecture, so they can be used for minimizing
blanchet
parents:
43029
diff
changeset
|
19 |
GaveUp | |
42587
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
20 |
ProofMissing | |
42882
391e41ac038b
make sure the Vampire incomplete proof detection code kicks in
blanchet
parents:
42876
diff
changeset
|
21 |
ProofIncomplete | |
44915
635ae0a73688
simplified unsound proof detection by removing impossible case
blanchet
parents:
44784
diff
changeset
|
22 |
UnsoundProof of bool * string list | |
42587
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
23 |
CantConnect | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
24 |
TimedOut | |
42953
26111aafab12
detect inappropriate problems and crashes better in Waldmeister
blanchet
parents:
42943
diff
changeset
|
25 |
Inappropriate | |
42587
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
26 |
OutOfResources | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
27 |
NoPerl | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
28 |
NoLibwwwPerl | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
29 |
MalformedInput | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
30 |
MalformedOutput | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
31 |
Interrupted | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
32 |
Crashed | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
33 |
InternalError | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
34 |
UnknownError of string |
39491
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
35 |
|
45551 | 36 |
type step_name = string * string list |
39452 | 37 |
|
39453
1740a2d6bef9
use the same TSTP/Vampire/SPASS parser for one-liners as for Isar proofs
blanchet
parents:
39452
diff
changeset
|
38 |
datatype 'a step = |
47774 | 39 |
Definition_Step of step_name * 'a * 'a | |
40 |
Inference_Step of step_name * 'a * string * step_name list |
|
39452 | 41 |
|
43678 | 42 |
type 'a proof = ('a, 'a, ('a, 'a) ho_term) formula step list |
39452 | 43 |
|
41259
13972ced98d9
more precise error messages in "verbose" (or "debug") mode, following this morning's permission debacle
blanchet
parents:
41222
diff
changeset
|
44 |
val short_output : bool -> string -> string |
41744 | 45 |
val string_for_failure : failure -> string |
39491
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
46 |
val extract_important_message : string -> string |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
47 |
val extract_known_failure : |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
48 |
(failure * string) list -> string -> failure option |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
49 |
val extract_tstplike_proof_and_outcome : |
43473
fb2713b803e6
deal with ATP time slices in a more flexible/robust fashion
blanchet
parents:
43465
diff
changeset
|
50 |
bool -> bool -> (string * string) list -> (failure * string) list -> string |
fb2713b803e6
deal with ATP time slices in a more flexible/robust fashion
blanchet
parents:
43465
diff
changeset
|
51 |
-> string * failure option |
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
52 |
val is_same_atp_step : step_name -> step_name -> bool |
42961 | 53 |
val scan_general_id : string list -> string * string list |
47947 | 54 |
val satallax_unsat_coreN : string |
42961 | 55 |
val parse_formula : |
43678 | 56 |
string list -> (string, 'a, (string, 'a) ho_term) formula * string list |
43481 | 57 |
val atp_proof_from_tstplike_proof : |
58 |
string problem -> string -> string -> string proof |
|
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
59 |
val clean_up_atp_proof_dependencies : string proof -> string proof |
39454 | 60 |
val map_term_names_in_atp_proof : |
61 |
(string -> string) -> string proof -> string proof |
|
62 |
val nasty_atp_proof : string Symtab.table -> string proof -> string proof |
|
39452 | 63 |
end; |
64 |
||
65 |
structure ATP_Proof : ATP_PROOF = |
|
66 |
struct |
|
67 |
||
43085
0a2f5b86bdd7
first step in sharing more code between ATP and Metis translation
blanchet
parents:
43050
diff
changeset
|
68 |
open ATP_Util |
39491
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
69 |
open ATP_Problem |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
70 |
|
42965
1403595ec38c
slightly gracefuller handling of LEO-II and Satallax output
blanchet
parents:
42962
diff
changeset
|
71 |
exception UNRECOGNIZED_ATP_PROOF of unit |
1403595ec38c
slightly gracefuller handling of LEO-II and Satallax output
blanchet
parents:
42962
diff
changeset
|
72 |
|
39491
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
73 |
datatype failure = |
42587
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
74 |
Unprovable | |
43050
59284a13abc4
support "metis" and "metisFT" as provers in the architecture, so they can be used for minimizing
blanchet
parents:
43029
diff
changeset
|
75 |
GaveUp | |
42587
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
76 |
ProofMissing | |
42882
391e41ac038b
make sure the Vampire incomplete proof detection code kicks in
blanchet
parents:
42876
diff
changeset
|
77 |
ProofIncomplete | |
44915
635ae0a73688
simplified unsound proof detection by removing impossible case
blanchet
parents:
44784
diff
changeset
|
78 |
UnsoundProof of bool * string list | |
42587
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
79 |
CantConnect | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
80 |
TimedOut | |
42953
26111aafab12
detect inappropriate problems and crashes better in Waldmeister
blanchet
parents:
42943
diff
changeset
|
81 |
Inappropriate | |
42587
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
82 |
OutOfResources | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
83 |
NoPerl | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
84 |
NoLibwwwPerl | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
85 |
MalformedInput | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
86 |
MalformedOutput | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
87 |
Interrupted | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
88 |
Crashed | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
89 |
InternalError | |
4fbb1de05169
fixed SPASS fact offset calculation and report unexpected unsound proofs with so-called sound encodings
blanchet
parents:
42550
diff
changeset
|
90 |
UnknownError of string |
39491
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
91 |
|
41259
13972ced98d9
more precise error messages in "verbose" (or "debug") mode, following this morning's permission debacle
blanchet
parents:
41222
diff
changeset
|
92 |
fun elide_string threshold s = |
13972ced98d9
more precise error messages in "verbose" (or "debug") mode, following this morning's permission debacle
blanchet
parents:
41222
diff
changeset
|
93 |
if size s > threshold then |
13972ced98d9
more precise error messages in "verbose" (or "debug") mode, following this morning's permission debacle
blanchet
parents:
41222
diff
changeset
|
94 |
String.extract (s, 0, SOME (threshold div 2 - 5)) ^ " ...... " ^ |
13972ced98d9
more precise error messages in "verbose" (or "debug") mode, following this morning's permission debacle
blanchet
parents:
41222
diff
changeset
|
95 |
String.extract (s, size s - (threshold + 1) div 2 + 6, NONE) |
13972ced98d9
more precise error messages in "verbose" (or "debug") mode, following this morning's permission debacle
blanchet
parents:
41222
diff
changeset
|
96 |
else |
13972ced98d9
more precise error messages in "verbose" (or "debug") mode, following this morning's permission debacle
blanchet
parents:
41222
diff
changeset
|
97 |
s |
13972ced98d9
more precise error messages in "verbose" (or "debug") mode, following this morning's permission debacle
blanchet
parents:
41222
diff
changeset
|
98 |
fun short_output verbose output = |
42060
889d767ce5f4
make Minimizer honor "verbose" and "debug" options better
blanchet
parents:
41944
diff
changeset
|
99 |
if verbose then |
889d767ce5f4
make Minimizer honor "verbose" and "debug" options better
blanchet
parents:
41944
diff
changeset
|
100 |
if output = "" then "No details available" else elide_string 1000 output |
889d767ce5f4
make Minimizer honor "verbose" and "debug" options better
blanchet
parents:
41944
diff
changeset
|
101 |
else |
889d767ce5f4
make Minimizer honor "verbose" and "debug" options better
blanchet
parents:
41944
diff
changeset
|
102 |
"" |
41259
13972ced98d9
more precise error messages in "verbose" (or "debug") mode, following this morning's permission debacle
blanchet
parents:
41222
diff
changeset
|
103 |
|
41744 | 104 |
val missing_message_tail = |
105 |
" appears to be missing. You will need to install it if you want to invoke \ |
|
106 |
\remote provers." |
|
39491
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
107 |
|
42876
e336ef6313aa
more informative message when Sledgehammer finds an unsound proof
blanchet
parents:
42848
diff
changeset
|
108 |
fun involving [] = "" |
43005
c96f06bffd90
merge timeout messages from several ATPs into one message to avoid clutter
blanchet
parents:
42998
diff
changeset
|
109 |
| involving ss = |
43029
3e060b1c844b
use helpers and tweak Quickcheck's priority to it comes second (to give Solve Direct slightly more time before another prover runs)
blanchet
parents:
43005
diff
changeset
|
110 |
"involving " ^ space_implode " " (Try.serial_commas "and" (map quote ss)) ^ |
3e060b1c844b
use helpers and tweak Quickcheck's priority to it comes second (to give Solve Direct slightly more time before another prover runs)
blanchet
parents:
43005
diff
changeset
|
111 |
" " |
42876
e336ef6313aa
more informative message when Sledgehammer finds an unsound proof
blanchet
parents:
42848
diff
changeset
|
112 |
|
47506 | 113 |
fun string_for_failure Unprovable = "The generated problem is unprovable." |
43050
59284a13abc4
support "metis" and "metisFT" as provers in the architecture, so they can be used for minimizing
blanchet
parents:
43029
diff
changeset
|
114 |
| string_for_failure GaveUp = "The prover gave up." |
41744 | 115 |
| string_for_failure ProofMissing = |
116 |
"The prover claims the conjecture is a theorem but did not provide a proof." |
|
42882
391e41ac038b
make sure the Vampire incomplete proof detection code kicks in
blanchet
parents:
42876
diff
changeset
|
117 |
| string_for_failure ProofIncomplete = |
391e41ac038b
make sure the Vampire incomplete proof detection code kicks in
blanchet
parents:
42876
diff
changeset
|
118 |
"The prover claims the conjecture is a theorem but provided an incomplete \ |
46427 | 119 |
\(or unparsable) proof." |
44915
635ae0a73688
simplified unsound proof detection by removing impossible case
blanchet
parents:
44784
diff
changeset
|
120 |
| string_for_failure (UnsoundProof (false, ss)) = |
43823 | 121 |
"The prover found a type-unsound proof " ^ involving ss ^ |
122 |
"(or, less likely, your axioms are inconsistent). Specify a sound type \ |
|
123 |
\encoding or omit the \"type_enc\" option." |
|
44915
635ae0a73688
simplified unsound proof detection by removing impossible case
blanchet
parents:
44784
diff
changeset
|
124 |
| string_for_failure (UnsoundProof (true, ss)) = |
42876
e336ef6313aa
more informative message when Sledgehammer finds an unsound proof
blanchet
parents:
42848
diff
changeset
|
125 |
"The prover found a type-unsound proof " ^ involving ss ^ |
e336ef6313aa
more informative message when Sledgehammer finds an unsound proof
blanchet
parents:
42848
diff
changeset
|
126 |
"even though a supposedly type-sound encoding was used (or, less likely, \ |
43465 | 127 |
\your axioms are inconsistent). Please report this to the Isabelle \ |
128 |
\developers." |
|
41744 | 129 |
| string_for_failure CantConnect = "Cannot connect to remote server." |
130 |
| string_for_failure TimedOut = "Timed out." |
|
42953
26111aafab12
detect inappropriate problems and crashes better in Waldmeister
blanchet
parents:
42943
diff
changeset
|
131 |
| string_for_failure Inappropriate = |
47506 | 132 |
"The generated problem lies outside the prover's scope." |
41744 | 133 |
| string_for_failure OutOfResources = "The prover ran out of resources." |
134 |
| string_for_failure NoPerl = "Perl" ^ missing_message_tail |
|
135 |
| string_for_failure NoLibwwwPerl = |
|
136 |
"The Perl module \"libwww-perl\"" ^ missing_message_tail |
|
137 |
| string_for_failure MalformedInput = |
|
138 |
"The generated problem is malformed. Please report this to the Isabelle \ |
|
139 |
\developers." |
|
140 |
| string_for_failure MalformedOutput = "The prover output is malformed." |
|
43085
0a2f5b86bdd7
first step in sharing more code between ATP and Metis translation
blanchet
parents:
43050
diff
changeset
|
141 |
| string_for_failure Interrupted = "The prover was interrupted." |
41744 | 142 |
| string_for_failure Crashed = "The prover crashed." |
143 |
| string_for_failure InternalError = "An internal prover error occurred." |
|
144 |
| string_for_failure (UnknownError string) = |
|
145 |
"A prover error occurred" ^ |
|
41334
3cb52cbf0eed
enable E weight generation with unofficial latest version of E (tentatively called E 1.2B) -- backed by Judgment Day
blanchet
parents:
41265
diff
changeset
|
146 |
(if string = "" then ". (Pass the \"verbose\" option for details.)" |
3cb52cbf0eed
enable E weight generation with unofficial latest version of E (tentatively called E 1.2B) -- backed by Judgment Day
blanchet
parents:
41265
diff
changeset
|
147 |
else ":\n" ^ string) |
39491
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
148 |
|
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
149 |
fun extract_delimited (begin_delim, end_delim) output = |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
150 |
output |> first_field begin_delim |> the |> snd |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
151 |
|> first_field end_delim |> the |> fst |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
152 |
|> first_field "\n" |> the |> snd |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
153 |
handle Option.Option => "" |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
154 |
|
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
155 |
val tstp_important_message_delims = |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
156 |
("% SZS start RequiredInformation", "% SZS end RequiredInformation") |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
157 |
|
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
158 |
fun extract_important_message output = |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
159 |
case extract_delimited tstp_important_message_delims output of |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
160 |
"" => "" |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
161 |
| s => s |> space_explode "\n" |> filter_out (curry (op =) "") |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
162 |
|> map (perhaps (try (unprefix "%"))) |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
163 |
|> map (perhaps (try (unprefix " "))) |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
164 |
|> space_implode "\n " |> quote |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
165 |
|
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
166 |
(* Splits by the first possible of a list of delimiters. *) |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
167 |
fun extract_tstplike_proof delims output = |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
168 |
case pairself (find_first (fn s => String.isSubstring s output)) |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
169 |
(ListPair.unzip delims) of |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
170 |
(SOME begin_delim, SOME end_delim) => |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
171 |
extract_delimited (begin_delim, end_delim) output |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
172 |
| _ => "" |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
173 |
|
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
174 |
fun extract_known_failure known_failures output = |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
175 |
known_failures |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
176 |
|> find_first (fn (_, pattern) => String.isSubstring pattern output) |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
177 |
|> Option.map fst |
2416666e6f94
refactoring: move ATP proof and error extraction code to "ATP_Proof" module
blanchet
parents:
39457
diff
changeset
|
178 |
|
43473
fb2713b803e6
deal with ATP time slices in a more flexible/robust fashion
blanchet
parents:
43465
diff
changeset
|
179 |
fun extract_tstplike_proof_and_outcome verbose complete proof_delims |
42848 | 180 |
known_failures output = |
42882
391e41ac038b
make sure the Vampire incomplete proof detection code kicks in
blanchet
parents:
42876
diff
changeset
|
181 |
case (extract_tstplike_proof proof_delims output, |
391e41ac038b
make sure the Vampire incomplete proof detection code kicks in
blanchet
parents:
42876
diff
changeset
|
182 |
extract_known_failure known_failures output) of |
391e41ac038b
make sure the Vampire incomplete proof detection code kicks in
blanchet
parents:
42876
diff
changeset
|
183 |
(_, SOME ProofIncomplete) => ("", SOME ProofIncomplete) |
43246 | 184 |
| ("", SOME ProofMissing) => ("", NONE) |
42882
391e41ac038b
make sure the Vampire incomplete proof detection code kicks in
blanchet
parents:
42876
diff
changeset
|
185 |
| ("", SOME failure) => |
43050
59284a13abc4
support "metis" and "metisFT" as provers in the architecture, so they can be used for minimizing
blanchet
parents:
43029
diff
changeset
|
186 |
("", SOME (if failure = GaveUp andalso complete then Unprovable |
42882
391e41ac038b
make sure the Vampire incomplete proof detection code kicks in
blanchet
parents:
42876
diff
changeset
|
187 |
else failure)) |
43246 | 188 |
| ("", NONE) => ("", SOME (UnknownError (short_output verbose output))) |
42882
391e41ac038b
make sure the Vampire incomplete proof detection code kicks in
blanchet
parents:
42876
diff
changeset
|
189 |
| (tstplike_proof, _) => (tstplike_proof, NONE) |
39452 | 190 |
|
45551 | 191 |
type step_name = string * string list |
39452 | 192 |
|
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
193 |
fun is_same_atp_step (s1, _) (s2, _) = s1 = s2 |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
194 |
|
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
195 |
val vampire_fact_prefix = "f" |
39452 | 196 |
|
197 |
fun step_name_ord p = |
|
39455 | 198 |
let val q = pairself fst p in |
39452 | 199 |
(* The "unprefix" part is to cope with remote Vampire's output. The proper |
200 |
solution would be to perform a topological sort, e.g. using the nice |
|
201 |
"Graph" functor. *) |
|
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
202 |
case pairself (Int.fromString |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
203 |
o perhaps (try (unprefix vampire_fact_prefix))) q of |
39452 | 204 |
(NONE, NONE) => string_ord q |
205 |
| (NONE, SOME _) => LESS |
|
206 |
| (SOME _, NONE) => GREATER |
|
207 |
| (SOME i, SOME j) => int_ord (i, j) |
|
208 |
end |
|
209 |
||
39453
1740a2d6bef9
use the same TSTP/Vampire/SPASS parser for one-liners as for Isar proofs
blanchet
parents:
39452
diff
changeset
|
210 |
datatype 'a step = |
47774 | 211 |
Definition_Step of step_name * 'a * 'a | |
212 |
Inference_Step of step_name * 'a * string * step_name list |
|
39452 | 213 |
|
43678 | 214 |
type 'a proof = ('a, 'a, ('a, 'a) ho_term) formula step list |
39452 | 215 |
|
47774 | 216 |
fun step_name (Definition_Step (name, _, _)) = name |
217 |
| step_name (Inference_Step (name, _, _, _)) = name |
|
39452 | 218 |
|
219 |
(**** PARSING OF TSTP FORMAT ****) |
|
220 |
||
42536 | 221 |
(* Strings enclosed in single quotes (e.g., file names) *) |
39452 | 222 |
val scan_general_id = |
47917
b287682bf917
improve parsing of Waldmeister dependencies (and kill obsolete hack)
blanchet
parents:
47787
diff
changeset
|
223 |
$$ "'" |-- Scan.repeat (~$$ "'") --| $$ "'" >> implode |
39452 | 224 |
|| Scan.repeat ($$ "$") -- Scan.many1 Symbol.is_letdig |
225 |
>> (fn (ss1, ss2) => implode ss1 ^ implode ss2) |
|
226 |
||
45235
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
227 |
val skip_term = |
45208
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
228 |
let |
45235
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
229 |
fun skip _ accum [] = (accum, []) |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
230 |
| skip 0 accum (ss as "," :: _) = (accum, ss) |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
231 |
| skip 0 accum (ss as ")" :: _) = (accum, ss) |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
232 |
| skip 0 accum (ss as "]" :: _) = (accum, ss) |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
233 |
| skip n accum ((s as "(") :: ss) = skip (n + 1) (s :: accum) ss |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
234 |
| skip n accum ((s as "[") :: ss) = skip (n + 1) (s :: accum) ss |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
235 |
| skip n accum ((s as "]") :: ss) = skip (n - 1) (s :: accum) ss |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
236 |
| skip n accum ((s as ")") :: ss) = skip (n - 1) (s :: accum) ss |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
237 |
| skip n accum (s :: ss) = skip n (s :: accum) ss |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
238 |
in skip 0 [] #>> (rev #> implode) end |
45208
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
239 |
|
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
240 |
datatype source = |
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
241 |
File_Source of string * string option | |
45209 | 242 |
Inference_Source of string * string list |
45208
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
243 |
|
45235
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
244 |
val dummy_phi = AAtom (ATerm ("", [])) |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
245 |
val dummy_inference = Inference_Source ("", []) |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
246 |
|
47917
b287682bf917
improve parsing of Waldmeister dependencies (and kill obsolete hack)
blanchet
parents:
47787
diff
changeset
|
247 |
(* "skip_term" is there to cope with Waldmeister nonsense such as |
b287682bf917
improve parsing of Waldmeister dependencies (and kill obsolete hack)
blanchet
parents:
47787
diff
changeset
|
248 |
"theory(equality)". *) |
b287682bf917
improve parsing of Waldmeister dependencies (and kill obsolete hack)
blanchet
parents:
47787
diff
changeset
|
249 |
val parse_dependency = scan_general_id --| skip_term |
b287682bf917
improve parsing of Waldmeister dependencies (and kill obsolete hack)
blanchet
parents:
47787
diff
changeset
|
250 |
val parse_dependencies = |
b287682bf917
improve parsing of Waldmeister dependencies (and kill obsolete hack)
blanchet
parents:
47787
diff
changeset
|
251 |
parse_dependency ::: Scan.repeat ($$ "," |-- parse_dependency) |
45208
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
252 |
|
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
253 |
fun parse_source x = |
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
254 |
(Scan.this_string "file" |-- $$ "(" |-- scan_general_id -- |
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
255 |
Scan.option ($$ "," |-- scan_general_id) --| $$ ")" |
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
256 |
>> File_Source |
45209 | 257 |
|| Scan.this_string "inference" |-- $$ "(" |-- scan_general_id |
45235
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
258 |
--| skip_term --| $$ "," --| skip_term --| $$ "," --| $$ "[" |
45209 | 259 |
-- parse_dependencies --| $$ "]" --| $$ ")" |
45235
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
260 |
>> Inference_Source |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
261 |
|| skip_term >> K dummy_inference) x |
39452 | 262 |
|
42966
4e2d6c1e5392
more work on parsing LEO-II proofs without lambdas
blanchet
parents:
42965
diff
changeset
|
263 |
fun list_app (f, args) = |
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
264 |
fold (fn arg => fn f => ATerm (tptp_app, [f, arg])) args f |
42966
4e2d6c1e5392
more work on parsing LEO-II proofs without lambdas
blanchet
parents:
42965
diff
changeset
|
265 |
|
45881 | 266 |
(* We currently ignore TFF and THF types. *) |
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
267 |
fun parse_type_stuff x = |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
268 |
Scan.repeat (($$ tptp_has_type || $$ tptp_fun_type) |-- parse_arg) x |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
269 |
and parse_arg x = |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
270 |
($$ "(" |-- parse_term --| $$ ")" --| parse_type_stuff |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
271 |
|| scan_general_id --| parse_type_stuff |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
272 |
-- Scan.optional ($$ "(" |-- parse_terms --| $$ ")") [] |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
273 |
>> ATerm) x |
45881 | 274 |
and parse_term x = |
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
275 |
(parse_arg -- Scan.repeat ($$ tptp_app |-- parse_arg) >> list_app) x |
42966
4e2d6c1e5392
more work on parsing LEO-II proofs without lambdas
blanchet
parents:
42965
diff
changeset
|
276 |
and parse_terms x = |
4e2d6c1e5392
more work on parsing LEO-II proofs without lambdas
blanchet
parents:
42965
diff
changeset
|
277 |
(parse_term ::: Scan.repeat ($$ "," |-- parse_term)) x |
39452 | 278 |
|
39598 | 279 |
fun parse_atom x = |
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
280 |
(parse_term -- Scan.option (Scan.option ($$ tptp_not_infix) --| $$ tptp_equal |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
281 |
-- parse_term) |
39598 | 282 |
>> (fn (u1, NONE) => AAtom u1 |
45881 | 283 |
| (u1, SOME (neg, u2)) => |
284 |
AAtom (ATerm ("equal", [u1, u2])) |> is_some neg ? mk_anot)) x |
|
39452 | 285 |
|
286 |
(* TPTP formulas are fully parenthesized, so we don't need to worry about |
|
287 |
operator precedence. *) |
|
42605
8734eb0033b3
Vampire sometimes generates formulas with ~ (not) followed by a quantified subformula, without parentheses -- parse these correctly
blanchet
parents:
42603
diff
changeset
|
288 |
fun parse_literal x = |
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
289 |
((Scan.repeat ($$ tptp_not) >> length) |
42605
8734eb0033b3
Vampire sometimes generates formulas with ~ (not) followed by a quantified subformula, without parentheses -- parse these correctly
blanchet
parents:
42603
diff
changeset
|
290 |
-- ($$ "(" |-- parse_formula --| $$ ")" |
8734eb0033b3
Vampire sometimes generates formulas with ~ (not) followed by a quantified subformula, without parentheses -- parse these correctly
blanchet
parents:
42603
diff
changeset
|
291 |
|| parse_quantified_formula |
8734eb0033b3
Vampire sometimes generates formulas with ~ (not) followed by a quantified subformula, without parentheses -- parse these correctly
blanchet
parents:
42603
diff
changeset
|
292 |
|| parse_atom) |
8734eb0033b3
Vampire sometimes generates formulas with ~ (not) followed by a quantified subformula, without parentheses -- parse these correctly
blanchet
parents:
42603
diff
changeset
|
293 |
>> (fn (n, phi) => phi |> n mod 2 = 1 ? mk_anot)) x |
8734eb0033b3
Vampire sometimes generates formulas with ~ (not) followed by a quantified subformula, without parentheses -- parse these correctly
blanchet
parents:
42603
diff
changeset
|
294 |
and parse_formula x = |
8734eb0033b3
Vampire sometimes generates formulas with ~ (not) followed by a quantified subformula, without parentheses -- parse these correctly
blanchet
parents:
42603
diff
changeset
|
295 |
(parse_literal |
43163 | 296 |
-- Scan.option ((Scan.this_string tptp_implies |
297 |
|| Scan.this_string tptp_iff |
|
298 |
|| Scan.this_string tptp_not_iff |
|
299 |
|| Scan.this_string tptp_if |
|
300 |
|| $$ tptp_or |
|
301 |
|| $$ tptp_and) -- parse_formula) |
|
39452 | 302 |
>> (fn (phi1, NONE) => phi1 |
43163 | 303 |
| (phi1, SOME (c, phi2)) => |
304 |
if c = tptp_implies then mk_aconn AImplies phi1 phi2 |
|
305 |
else if c = tptp_iff then mk_aconn AIff phi1 phi2 |
|
306 |
else if c = tptp_not_iff then mk_anot (mk_aconn AIff phi1 phi2) |
|
307 |
else if c = tptp_if then mk_aconn AImplies phi2 phi1 |
|
308 |
else if c = tptp_or then mk_aconn AOr phi1 phi2 |
|
309 |
else if c = tptp_and then mk_aconn AAnd phi1 phi2 |
|
310 |
else raise Fail ("impossible connective " ^ quote c))) x |
|
42605
8734eb0033b3
Vampire sometimes generates formulas with ~ (not) followed by a quantified subformula, without parentheses -- parse these correctly
blanchet
parents:
42603
diff
changeset
|
311 |
and parse_quantified_formula x = |
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
312 |
(($$ tptp_forall >> K AForall || $$ tptp_exists >> K AExists) |
42605
8734eb0033b3
Vampire sometimes generates formulas with ~ (not) followed by a quantified subformula, without parentheses -- parse these correctly
blanchet
parents:
42603
diff
changeset
|
313 |
--| $$ "[" -- parse_terms --| $$ "]" --| $$ ":" -- parse_literal |
8734eb0033b3
Vampire sometimes generates formulas with ~ (not) followed by a quantified subformula, without parentheses -- parse these correctly
blanchet
parents:
42603
diff
changeset
|
314 |
>> (fn ((q, ts), phi) => |
42966
4e2d6c1e5392
more work on parsing LEO-II proofs without lambdas
blanchet
parents:
42965
diff
changeset
|
315 |
(* We ignore TFF and THF types for now. *) |
45881 | 316 |
AQuant (q, map (fn ATerm (s, _) => (s, NONE)) ts, phi))) x |
39452 | 317 |
|
318 |
val parse_tstp_extra_arguments = |
|
45235
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
319 |
Scan.optional ($$ "," |-- parse_source --| Scan.option ($$ "," |-- skip_term)) |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
320 |
dummy_inference |
39452 | 321 |
|
47927
c35238d19bb9
repair the Waldmeister endgame only for Waldmeister proofs
blanchet
parents:
47926
diff
changeset
|
322 |
val waldmeister_conjecture_name = "conjecture_1" |
42943 | 323 |
|
42536 | 324 |
val tofof_fact_prefix = "fof_" |
41203
1393514094d7
fixed more issues with the Vampire output parser, and added support for Vampire's TSTP output (--proof tptp)
blanchet
parents:
41201
diff
changeset
|
325 |
|
42943 | 326 |
fun is_same_term subst tm1 tm2 = |
327 |
let |
|
328 |
fun do_term_pair _ NONE = NONE |
|
329 |
| do_term_pair (ATerm (s1, tm1), ATerm (s2, tm2)) (SOME subst) = |
|
42998
1c80902d0456
fully support all type system encodings in typed formats (TFF, THF)
blanchet
parents:
42975
diff
changeset
|
330 |
case pairself is_tptp_variable (s1, s2) of |
42943 | 331 |
(true, true) => |
332 |
(case AList.lookup (op =) subst s1 of |
|
333 |
SOME s2' => if s2' = s2 then SOME subst else NONE |
|
334 |
| NONE => |
|
335 |
if null (AList.find (op =) subst s2) then SOME ((s1, s2) :: subst) |
|
336 |
else NONE) |
|
337 |
| (false, false) => |
|
338 |
if s1 = s2 andalso length tm1 = length tm2 then |
|
339 |
SOME subst |> fold do_term_pair (tm1 ~~ tm2) |
|
340 |
else |
|
341 |
NONE |
|
342 |
| _ => NONE |
|
343 |
in SOME subst |> do_term_pair (tm1, tm2) |> is_some end |
|
344 |
||
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
345 |
fun is_same_formula comm subst (AQuant (q1, xs1, phi1)) (AQuant (q2, xs2, phi2)) = |
42943 | 346 |
q1 = q2 andalso length xs1 = length xs2 andalso |
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
347 |
is_same_formula comm ((map fst xs1 ~~ map fst xs2) @ subst) phi1 phi2 |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
348 |
| is_same_formula comm subst (AConn (c1, phis1)) (AConn (c2, phis2)) = |
42943 | 349 |
c1 = c2 andalso length phis1 = length phis2 andalso |
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
350 |
forall (uncurry (is_same_formula comm subst)) (phis1 ~~ phis2) |
47926 | 351 |
| is_same_formula comm subst (AAtom (tm1 as ATerm ("equal", [tm11, tm12]))) |
352 |
(AAtom tm2) = |
|
353 |
is_same_term subst tm1 tm2 orelse |
|
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
354 |
(comm andalso is_same_term subst (ATerm ("equal", [tm12, tm11])) tm2) |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
355 |
| is_same_formula _ subst (AAtom tm1) (AAtom tm2) = is_same_term subst tm1 tm2 |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
356 |
| is_same_formula _ _ _ _ = false |
42943 | 357 |
|
358 |
fun matching_formula_line_identifier phi (Formula (ident, _, phi', _, _)) = |
|
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
359 |
if is_same_formula true [] phi phi' then SOME (ident, phi') else NONE |
42943 | 360 |
| matching_formula_line_identifier _ _ = NONE |
361 |
||
362 |
fun find_formula_in_problem problem phi = |
|
363 |
problem |> maps snd |> map_filter (matching_formula_line_identifier phi) |
|
45551 | 364 |
|> try (single o hd) |> the_default [] |
42943 | 365 |
|
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
366 |
fun commute_eq (AAtom (ATerm (s, tms))) = AAtom (ATerm (s, rev tms)) |
47946
33afcfad3f8d
add an experimental "aggressive" mode to Sledgehammer, to experiment with more complete translations of higher-order features without breaking "metis"
blanchet
parents:
47927
diff
changeset
|
367 |
| commute_eq _ = raise Fail "expected equation" |
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
368 |
|
42962
3b50fdeb6cfc
started adding support for THF output (but no lambdas)
blanchet
parents:
42961
diff
changeset
|
369 |
(* Syntax: (cnf|fof|tff|thf)\(<num>, <formula_role>, |
3b50fdeb6cfc
started adding support for THF output (but no lambdas)
blanchet
parents:
42961
diff
changeset
|
370 |
<formula> <extra_arguments>\). |
39452 | 371 |
The <num> could be an identifier, but we assume integers. *) |
42943 | 372 |
fun parse_tstp_line problem = |
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
373 |
((Scan.this_string tptp_cnf || Scan.this_string tptp_fof |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
374 |
|| Scan.this_string tptp_tff || Scan.this_string tptp_thf) -- $$ "(") |
42943 | 375 |
|-- scan_general_id --| $$ "," -- Symbol.scan_id --| $$ "," |
45235
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
376 |
-- (parse_formula || skip_term >> K dummy_phi) -- parse_tstp_extra_arguments |
7187bce94e88
more robust parsing of TSTP sources -- Vampire has nonstandard "introduced()" tags and Waldmeister(OnTPTP) has weird "theory(...)" dependencies
blanchet
parents:
45209
diff
changeset
|
377 |
--| $$ ")" --| $$ "." |
42943 | 378 |
>> (fn (((num, role), phi), deps) => |
379 |
let |
|
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
380 |
val ((name, phi), rule, deps) = |
42943 | 381 |
(* Waldmeister isn't exactly helping. *) |
382 |
case deps of |
|
45208
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
383 |
File_Source (_, SOME s) => |
47927
c35238d19bb9
repair the Waldmeister endgame only for Waldmeister proofs
blanchet
parents:
47926
diff
changeset
|
384 |
(if s = waldmeister_conjecture_name then |
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
385 |
case find_formula_in_problem problem (mk_anot phi) of |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
386 |
(* Waldmeister hack: Get the original orientation of the |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
387 |
equation to avoid confusing Isar. *) |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
388 |
[(s, phi')] => |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
389 |
((num, [s]), |
47926 | 390 |
phi |> not (is_same_formula false [] (mk_anot phi) phi') |
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
391 |
? commute_eq) |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
392 |
| _ => ((num, []), phi) |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
393 |
else |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
394 |
((num, [s |> perhaps (try (unprefix tofof_fact_prefix))]), |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
395 |
phi), |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
396 |
"", []) |
45208
9a00f9cc8707
marginally cleaner proof parsing, that doesn't stumble upon LEO-II's E-step proofs
blanchet
parents:
45203
diff
changeset
|
397 |
| File_Source _ => |
47921
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
398 |
(((num, phi |> find_formula_in_problem problem |> map fst), |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
399 |
phi), "", []) |
fc26d5538868
ensure the "show" equation is not reoriented by Waldmeister
blanchet
parents:
47917
diff
changeset
|
400 |
| Inference_Source (rule, deps) => (((num, []), phi), rule, deps) |
47787 | 401 |
fun mk_step () = |
402 |
Inference_Step (name, phi, rule, map (rpair []) deps) |
|
42943 | 403 |
in |
404 |
case role of |
|
405 |
"definition" => |
|
406 |
(case phi of |
|
407 |
AConn (AIff, [phi1 as AAtom _, phi2]) => |
|
47774 | 408 |
Definition_Step (name, phi1, phi2) |
42943 | 409 |
| AAtom (ATerm ("equal", _)) => |
410 |
(* Vampire's equality proxy axiom *) |
|
47774 | 411 |
Inference_Step (name, phi, rule, map (rpair []) deps) |
47787 | 412 |
| _ => mk_step ()) |
413 |
| _ => mk_step () |
|
42943 | 414 |
end) |
39452 | 415 |
|
416 |
(**** PARSING OF SPASS OUTPUT ****) |
|
417 |
||
418 |
(* SPASS returns clause references of the form "x.y". We ignore "y", whose role |
|
419 |
is not clear anyway. *) |
|
420 |
val parse_dot_name = scan_general_id --| $$ "." --| scan_general_id |
|
421 |
||
422 |
val parse_spass_annotations = |
|
423 |
Scan.optional ($$ ":" |-- Scan.repeat (parse_dot_name |
|
424 |
--| Scan.option ($$ ","))) [] |
|
425 |
||
426 |
(* It is not clear why some literals are followed by sequences of stars and/or |
|
427 |
pluses. We ignore them. *) |
|
39602 | 428 |
fun parse_decorated_atom x = |
429 |
(parse_atom --| Scan.repeat ($$ "*" || $$ "+" || $$ " ")) x |
|
39452 | 430 |
|
431 |
fun mk_horn ([], []) = AAtom (ATerm ("c_False", [])) |
|
42943 | 432 |
| mk_horn ([], pos_lits) = foldr1 (uncurry (mk_aconn AOr)) pos_lits |
433 |
| mk_horn (neg_lits, []) = mk_anot (foldr1 (uncurry (mk_aconn AAnd)) neg_lits) |
|
39452 | 434 |
| mk_horn (neg_lits, pos_lits) = |
42943 | 435 |
mk_aconn AImplies (foldr1 (uncurry (mk_aconn AAnd)) neg_lits) |
436 |
(foldr1 (uncurry (mk_aconn AOr)) pos_lits) |
|
39452 | 437 |
|
39645 | 438 |
fun parse_horn_clause x = |
439 |
(Scan.repeat parse_decorated_atom --| $$ "|" --| $$ "|" |
|
440 |
-- Scan.repeat parse_decorated_atom --| $$ "-" --| $$ ">" |
|
441 |
-- Scan.repeat parse_decorated_atom |
|
442 |
>> (mk_horn o apfst (op @))) x |
|
39452 | 443 |
|
46427 | 444 |
fun resolve_spass_num (SOME names) _ _ = names |
445 |
| resolve_spass_num NONE spass_names num = |
|
446 |
case Int.fromString num of |
|
447 |
SOME j => if j > 0 andalso j <= Vector.length spass_names then |
|
448 |
Vector.sub (spass_names, j - 1) |
|
449 |
else |
|
450 |
[] |
|
451 |
| NONE => [] |
|
43481 | 452 |
|
46390 | 453 |
val parse_spass_debug = |
454 |
Scan.option ($$ "(" |-- Scan.repeat (scan_general_id --| Scan.option ($$ ",")) |
|
455 |
--| $$ ")") |
|
456 |
||
46427 | 457 |
(* Syntax: <num>[0:<inference><annotations>] <atoms> || <atoms> -> <atoms>. |
458 |
derived from formulae <ident>* *) |
|
45162 | 459 |
fun parse_spass_line spass_names = |
46390 | 460 |
parse_spass_debug |-- scan_general_id --| $$ "[" --| $$ "0" --| $$ ":" |
461 |
-- Symbol.scan_id -- parse_spass_annotations --| $$ "]" |
|
462 |
-- parse_horn_clause --| $$ "." |
|
46427 | 463 |
-- Scan.option (Scan.this_string "derived from formulae " |
46451 | 464 |
|-- Scan.repeat (scan_general_id --| Scan.option ($$ " "))) |
46427 | 465 |
>> (fn ((((num, rule), deps), u), names) => |
47774 | 466 |
Inference_Step ((num, resolve_spass_num names spass_names num), u, |
467 |
rule, map (swap o `(resolve_spass_num NONE spass_names)) deps)) |
|
45162 | 468 |
|
47947 | 469 |
val satallax_unsat_coreN = "__satallax_unsat_core" (* arbitrary *) |
470 |
||
45162 | 471 |
(* Syntax: <name> *) |
45203 | 472 |
fun parse_satallax_line x = |
473 |
(scan_general_id --| Scan.option ($$ " ") |
|
47947 | 474 |
>> (fn s => Inference_Step ((s, [s]), dummy_phi, satallax_unsat_coreN, []))) |
475 |
x |
|
43481 | 476 |
|
477 |
fun parse_line problem spass_names = |
|
45203 | 478 |
parse_tstp_line problem || parse_spass_line spass_names || parse_satallax_line |
43481 | 479 |
fun parse_proof problem spass_names tstp = |
44784 | 480 |
tstp |> strip_spaces_except_between_idents |
43481 | 481 |
|> raw_explode |
482 |
|> Scan.finite Symbol.stopper |
|
483 |
(Scan.error (!! (fn _ => raise UNRECOGNIZED_ATP_PROOF ()) |
|
484 |
(Scan.repeat1 (parse_line problem spass_names)))) |
|
485 |
|> fst |
|
486 |
||
487 |
(** SPASS's FLOTTER hack **) |
|
488 |
||
489 |
(* This is a hack required for keeping track of facts after they have been |
|
490 |
clausified by SPASS's FLOTTER preprocessor. The "ATP/scripts/spass" script is |
|
491 |
also part of this hack. *) |
|
492 |
||
493 |
val set_ClauseFormulaRelationN = "set_ClauseFormulaRelation" |
|
494 |
||
495 |
fun extract_clause_sequence output = |
|
496 |
let |
|
497 |
val tokens_of = String.tokens (not o Char.isAlphaNum) |
|
498 |
fun extract_num ("clause" :: (ss as _ :: _)) = Int.fromString (List.last ss) |
|
499 |
| extract_num _ = NONE |
|
500 |
in output |> split_lines |> map_filter (extract_num o tokens_of) end |
|
39452 | 501 |
|
43481 | 502 |
fun is_head_digit s = Char.isDigit (String.sub (s, 0)) |
503 |
val scan_integer = Scan.many1 is_head_digit >> (the o Int.fromString o implode) |
|
504 |
||
505 |
val parse_clause_formula_pair = |
|
506 |
$$ "(" |-- scan_integer --| $$ "," |
|
507 |
-- (Symbol.scan_id ::: Scan.repeat ($$ "," |-- Symbol.scan_id)) --| $$ ")" |
|
508 |
--| Scan.option ($$ ",") |
|
509 |
val parse_clause_formula_relation = |
|
510 |
Scan.this_string set_ClauseFormulaRelationN |-- $$ "(" |
|
511 |
|-- Scan.repeat parse_clause_formula_pair |
|
512 |
val extract_clause_formula_relation = |
|
513 |
Substring.full #> Substring.position set_ClauseFormulaRelationN |
|
514 |
#> snd #> Substring.position "." #> fst #> Substring.string |
|
515 |
#> raw_explode #> filter_out Symbol.is_blank #> parse_clause_formula_relation |
|
516 |
#> fst |
|
39452 | 517 |
|
43481 | 518 |
fun extract_spass_name_vector output = |
519 |
(if String.isSubstring set_ClauseFormulaRelationN output then |
|
520 |
let |
|
521 |
val num_seq = extract_clause_sequence output |
|
522 |
val name_map = extract_clause_formula_relation output |
|
523 |
val name_seq = num_seq |> map (these o AList.lookup (op =) name_map) |
|
524 |
in name_seq end |
|
525 |
else |
|
526 |
[]) |
|
527 |
|> Vector.fromList |
|
528 |
||
529 |
fun atp_proof_from_tstplike_proof _ _ "" = [] |
|
530 |
| atp_proof_from_tstplike_proof problem output tstp = |
|
531 |
tstp ^ "$" (* the $ sign acts as a sentinel (FIXME: needed?) *) |
|
532 |
|> parse_proof problem (extract_spass_name_vector output) |
|
42449
494e4ac5b0f8
detect some unsound proofs before showing them to the user
blanchet
parents:
42060
diff
changeset
|
533 |
|> sort (step_name_ord o pairself step_name) |
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
534 |
|
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
535 |
fun clean_up_dependencies _ [] = [] |
47774 | 536 |
| clean_up_dependencies seen |
537 |
((step as Definition_Step (name, _, _)) :: steps) = |
|
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
538 |
step :: clean_up_dependencies (name :: seen) steps |
47774 | 539 |
| clean_up_dependencies seen (Inference_Step (name, u, rule, deps) :: steps) = |
540 |
Inference_Step (name, u, rule, |
|
541 |
map_filter (fn dep => find_first (is_same_atp_step dep) seen) deps) :: |
|
42968
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
542 |
clean_up_dependencies (name :: seen) steps |
74415622d293
more work on parsing LEO-II proofs and extracting uses of extensionality
blanchet
parents:
42966
diff
changeset
|
543 |
|
42975 | 544 |
fun clean_up_atp_proof_dependencies proof = clean_up_dependencies [] proof |
39452 | 545 |
|
39454 | 546 |
fun map_term_names_in_term f (ATerm (s, ts)) = |
547 |
ATerm (f s, map (map_term_names_in_term f) ts) |
|
548 |
fun map_term_names_in_formula f (AQuant (q, xs, phi)) = |
|
549 |
AQuant (q, xs, map_term_names_in_formula f phi) |
|
550 |
| map_term_names_in_formula f (AConn (c, phis)) = |
|
551 |
AConn (c, map (map_term_names_in_formula f) phis) |
|
552 |
| map_term_names_in_formula f (AAtom t) = AAtom (map_term_names_in_term f t) |
|
47774 | 553 |
fun map_term_names_in_step f (Definition_Step (name, phi1, phi2)) = |
554 |
Definition_Step (name, map_term_names_in_formula f phi1, |
|
555 |
map_term_names_in_formula f phi2) |
|
556 |
| map_term_names_in_step f (Inference_Step (name, phi, rule, deps)) = |
|
557 |
Inference_Step (name, map_term_names_in_formula f phi, rule, deps) |
|
39454 | 558 |
fun map_term_names_in_atp_proof f = map (map_term_names_in_step f) |
559 |
||
560 |
fun nasty_name pool s = s |> Symtab.lookup pool |> the_default s |
|
561 |
fun nasty_atp_proof pool = |
|
562 |
if Symtab.is_empty pool then I |
|
563 |
else map_term_names_in_atp_proof (nasty_name pool) |
|
564 |
||
39452 | 565 |
end; |