michael/lean2
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
7502 commits 3 branches 0 tags 28 MiB
Commit graph

193 commits

Author SHA1 Message Date
Leonardo de Moura
08169c5ac2 fix(library/unifier): fixes #809 
Daniel is correct when he says the interaction between choice
case-splits, delta case-splits, and coercions can be subtle.

I believe the following condition
https://github.com/leanprover/lean/blob/master/src/frontends/lean/elaborator.cpp#L111
reduces counter-intuitive behavior. Example, the coercion should not
influence the resulting type.
BTW, by removing this condition, many files in the library broke when I
tried to compile from scratch

      make clean-olean
      make

I used the following workaround. Given a delta-delta constraint

           f a =?= f b

If the terms are types, and no case-split will be performed, then
the delta-delta constraint is eagerly solved.
In principle, we don't need the condition that the terms are types.
However, many files break if we remove it. The problem is that many files in the standard
library are abusing the higher-order unification procedure. The
elaboration problems are quite tricky to solve.
I use the extra condition "the terms are types" because usually if they
are, "f" is morally injective, and we don't really want to unfold it.

Note that the following two cases do not work

     check '{1, 2, 3}
     check insert 1 (insert 2 (insert 3 empty))

Well, they work if we the num namespace is open, and they are
interpreted as having type (finset num)
 2015-08-31 17:59:30 -10:00
Leonardo de Moura
dc2e702373 feat(library/unifier): generate approximate solution for universe constraints of the form (max u ?m =?= max u v) 
closes #777
 2015-08-08 09:29:59 -07:00
Leonardo de Moura
5c7a20e5bd fix(library/unifier): crash when unifying constraints of the form (pr t =?= s) 
where pr is a projection and t is a stuck term

see issue #737
 2015-07-24 11:52:46 -07:00
Leonardo de Moura
1886b71c17 fix(library/unifier): fixes #705 2015-06-26 19:10:46 -07:00
Leonardo de Moura
c61e6f6595 feat(library/unifier): add new rule for constraints of the form (pr ...) =?= t, where (pr ...) is a "stuck" projection application 2015-06-26 17:18:29 -07:00
Leonardo de Moura
0b7859f387 feat(library,frontends/lean): add support for projections in the elaborator 
The idea is to simulate the computation rules such as

    pr1 (mk a b) ==> a

in the elaborator
 2015-06-26 17:18:29 -07:00
Leonardo de Moura
d8620ef4c9 fix(kernel,library,frontends/lean): improve error messages 
see #669
 2015-06-14 19:44:00 -07:00
Leonardo de Moura
e3a0e62859 fix(library/unifier): try to generate approximate solution for flex-flex constraints before discarding them 
fixes #662
 2015-06-09 14:36:31 -07:00
Leonardo de Moura
f830bf54c2 refactor(*): clarify name_generator ownership 2015-05-21 14:32:36 -07:00
Leonardo de Moura
4f12409c63 fix(library/unifier): assertion violation 
This assertion violation was introduced when we added "projection
macros" to speedup the elaboration process.
 2015-05-19 10:07:31 -07:00
Leonardo de Moura
19361f0196 feat(library/unifier): do not fire type class resolution as last resort when type contains metavariables 
see discussion at #604
 2015-05-18 15:45:23 -07:00
Leonardo de Moura
57ea660963 refactor(*): start process for eliminating of opaque definitions from the kernel 
see issue #576
 2015-05-08 16:06:04 -07:00
Leonardo de Moura
eb3a236119 fix(library/unifier): typo 
fixes #588
 2015-05-07 16:30:02 -07:00
Leonardo de Moura
88cd6e9a63 chore(library/unifier): remove dead code 2015-05-07 15:49:33 -07:00
Leonardo de Moura
eb23a30626 feat(library/unifier): additional memory checks 2015-04-20 17:41:08 -07:00
Leonardo de Moura
28dad944c5 fix(library/unifier): cyclic assignment (?M <- ?M) 
This was producing nonterminating behavior on example described at issue #489
 2015-04-20 17:35:37 -07:00
Leonardo de Moura
9fe2d5c74c refactor(library/unifier): use new assign method in the unifier 2015-03-12 15:01:40 -07:00
Leonardo de Moura
8f004671a2 fix(library/unifier): typo 2015-03-12 13:15:23 -07:00
Leonardo de Moura
3e4d849a4a refactor(kernel/metavar.h): simplify API 2015-03-12 12:50:53 -07:00
Leonardo de Moura
1244f01518 chore(library/unifier): remove dead code 2015-03-09 14:07:14 -07:00
Leonardo de Moura
9a6c675908 feat(library/unifier): add option to disable nonchronological backtracking 2015-03-09 12:08:58 -07:00
Leonardo de Moura
abd238aef0 feat(*): add [quasireducible] attribute 2015-03-04 22:12:49 -08:00
Leonardo de Moura
7db6ed7c14 refactor(library/unifier): move m_pattern configuration option to unifier_config 2015-03-03 20:24:18 -08:00
Leonardo de Moura
341a9a2010 refactor(library/unifier): remove dead code 2015-03-03 20:14:17 -08:00
Leonardo de Moura
7e2f0f9a36 fix(library/unifier): in the context_check, we should not consider local constants that occur in the type of other constants 
This was a performance bug.
We were missing higher-order pattern constraints due to this bug.
 2015-03-02 17:28:56 -08:00
Leonardo de Moura
72eed42ac8 feat(library/unifier): ignore irrelevant branches when solving flex-rigid constraints 2015-02-26 13:43:54 -08:00
Leonardo de Moura
fcd67649ed refactor(kernel): expose may_reduce_later method 2015-02-07 20:36:26 -08:00
Leonardo de Moura
b57f93bad5 refactor(kernel): remove unnecessary procedures 2015-02-07 20:14:19 -08:00
Leonardo de Moura
dbc8e9e13a refactor(*): add method get_num_univ_params 2015-01-28 17:22:18 -08:00
Leonardo de Moura
2717adde94 feat(library/unifier): add option 'unifier.conservative', use option by default in the calc_assistant 2015-01-19 16:23:29 -08:00
Leonardo de Moura
2e4a2451e6 refactor(library/reducible): simplify reducible/irreducible semantics 2015-01-08 18:52:18 -08:00
Leonardo de Moura
597f7385e8 fix(library/unifier): incorrect fix 2015-01-07 16:57:02 -08:00
Leonardo de Moura
2990a6d029 fix(library/unifier): missing "set_conflict" in process_delta 2015-01-07 12:41:01 -08:00
Leonardo de Moura
2867789bec fix(library/unifier): handle missing first-order flex-flex case 2014-12-10 22:11:30 -08:00
Leonardo de Moura
050104cdfd fix(library/unifier): assertion violation 2014-12-02 12:17:59 -08:00
Leonardo de Moura
06f436840f fix(library/unifier): postpone class-instance constraints whose type could not be inferred 2014-12-01 22:27:23 -08:00
Leonardo de Moura
19d14678ef refactor(library/unifier): remove dead code 2014-12-01 21:57:34 -08:00
Leonardo de Moura
f1e915a188 feat(frontends/lean): add 'find_decl' command 2014-11-23 23:00:59 -08:00
Leonardo de Moura
8c8bf41e39 feat(frontends/lean/server): do not unfold definitions in FINDG 2014-11-23 19:03:39 -08:00
Leonardo de Moura
00df34a1c4 feat(library/unifier): generalize method process_succ_eq_max_core 2014-11-14 14:25:41 -08:00
Leonardo de Moura
51719145f9 feat(library/unifier): solved universe constraints of the form succ^k1 a = max k2 ?m (when k1 >= k2) 2014-11-12 17:28:33 -08:00
Leonardo de Moura
ea739100b3 fix(library/unifier): broken optimization in the unifier 
See new comments and tests for details.
 2014-10-28 16:09:41 -07:00
Leonardo de Moura
78bc3ef7e4 feat(library/unifier): improve FailLocal/FailCircular failures in the unifier by using normalization 
This improvements was marked as TODO, and was preventing us from
elaborating the example in the new test vector3.lean
 2014-10-27 16:49:29 -07:00
Leonardo de Moura
096c67b2e5 fix(library/unifier): occurs-check bug 2014-10-25 00:16:02 -07:00
Leonardo de Moura
8974b52f7b perf(library/unifier): avoid unnecessary wasteful computation 2014-10-16 17:16:49 -07:00
Leonardo de Moura
9ba59c6b25 feat(library/universe): improve support for universe level constraints in the unifier 2014-10-09 20:28:39 -07:00
Leonardo de Moura
8fa171cb92 refactor(library/unifier): allow general 'unify' procedure to take an initial substitution as argument 2014-10-07 17:30:57 -07:00
Leonardo de Moura
60d8369688 fix(library/unifier): missing justification when updating choice constraints 
The bug was not producing incorrect results, but really bad error
messages.
See: empty.lean.expected.out
 2014-10-04 10:40:53 -07:00
Leonardo de Moura
9f6a8827e0 refactor(*): use name_map 2014-09-28 10:23:11 -07:00
Leonardo de Moura
22e47430b5 feat(library/unifier): add 'on-demand' choice constraints, they are processed as soon as their type does not contain meta-variables anymore 2014-09-27 21:50:39 -07:00