michael/lean2
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
7866 commits 3 branches 0 tags 28 MiB
Commit graph

1117 commits

Author SHA1 Message Date
Daniel Selsam
413989afd6 feat(library/blast/backward): backward chaining strategy 2015-11-18 17:48:39 -08:00
Leonardo de Moura
d5d8ac8b44 fix(kernel/expr): warning when compiling using clang on OSX 2015-11-16 18:50:34 -08:00
Leonardo de Moura
92a7c38260 feat(kernel/expr): add mk_app that takes a list of arguments 2015-11-13 14:01:15 -08:00
Leonardo de Moura
76032eea90 feat(kernel/expr): change dummy expression used in default constructor 2015-11-08 14:05:02 -08:00
Leonardo de Moura
a446c54b87 fix(tests/kernel/max_sharing): max_sharing on OSX 2015-10-03 12:12:21 -07:00
Leonardo de Moura
d324d54d21 fix(kernel/expr): add global expressions storing Prop and Type.{1} to 
hash-consing cache when max-sharing is enabled
 2015-10-02 15:04:10 -07:00
Leonardo de Moura
3cf11dac87 feat(kernel/instantiate): make sure instantiate_type_univ_params and instantiate_value_univ_params caches are reset when we enable max-sharing 2015-10-01 15:42:33 -07:00
Leonardo de Moura
45594e86c7 feat(kernel/abstract): make sure Pi/Fun cache is reset when we enable max-sharing 2015-10-01 15:28:55 -07:00
Leonardo de Moura
cb2f93ee6a fix(kernel/expr,kernel/level): performance problem 
The method unordered_set::clear takes time O(n) where n is the number of
buckets in the hashtable used to implement the set.
Moreover, it does not reduce the number of buckets to 0.
That is, it keeps `n` empty buckets.

This creates performance problems if the number of buckets gets really
huge at one point.
The tst6 at tests/kernel/expr.cpp demonstrates the problem
 2015-09-30 17:25:13 -07:00
Leonardo de Moura
d07bbe7e8c fix(kernel/expr): must take binder annotations into account in the hash-consing tables 2015-09-30 17:24:41 -07:00
Leonardo de Moura
57035d3162 feat(kernel/level,library/blast/expr): add universe level hash consing support in the kernel, simplify blast/expr even more 2015-09-30 13:31:42 -07:00
Leonardo de Moura
c5603e456a feat(kernel/expr): replace "opportunistic" caching with precise caching 
We also removed compilation option LEAN_CACHE_EXPRS
 2015-09-30 12:29:43 -07:00
Leonardo de Moura
4cfebe7f1c refactor(kernel): move instantiate_univ_cache to separate .h file 2015-09-21 16:08:34 -07:00
Leonardo de Moura
9c09b0750b chore(kernel/expr): remove dead code 2015-09-21 15:43:51 -07:00
Leonardo de Moura
a7c2d798de refactor(kernel): move replace_cache to separate .h file 2015-09-21 15:24:45 -07:00
Leonardo de Moura
df3100d2cd fix(library/local_context): bug in abstract_locals procedure 2015-09-12 17:17:13 -07:00
Leonardo de Moura
45163acf25 refactor(kernel/inductive): use local constants to represent introduction rules 2015-08-25 03:46:28 -07:00
Leonardo de Moura
edb4c09bc1 fix(frontends/lean,kernel/inductive): compilation errors in Debug mode 2015-08-16 19:02:48 -07:00
Leonardo de Moura
5f5642c4ce fix(kernel/inductive): compilation error with clang++ 2015-08-15 15:06:57 -07:00
Leonardo de Moura
7bc8673786 feat(library/module): efficient inductive_reader 
Do not check imported inductive declarations when trust level is greater than 0.
 2015-08-15 14:48:49 -07:00
Leonardo de Moura
e80d9685e5 refactor(kernel/inductive): add certified_inductive_decl object 
We will use this object to implement a more efficient import procedure
 2015-08-15 13:26:38 -07:00
Leonardo de Moura
6c934229f7 feat(kernel,library/module): only module reader can add declarations without type-checking them 2015-08-14 18:37:17 -07:00
Leonardo de Moura
498afc1e6f feat(CMakeLists): add shared library 2015-08-13 11:21:05 -07:00
Leonardo de Moura
8048cbd6f2 feat(kernel): do not hide semi-constructive axioms 2015-07-28 18:15:25 -07:00
Leonardo de Moura
667f647b27 feat(kernel/expr_eq_fn): add small optimization 2015-07-01 09:19:36 -07:00
Leonardo de Moura
e7c3c887b6 feat(kernel/type_checker): add 'check' procedure that uses 'opaque_hints' 
The hints only affect performance. If a declaration type checks assuming
some constants are opaque, it should also type check without this
assumption. Of course, it may be much slower.
 2015-06-30 13:12:34 -07:00
Leonardo de Moura
772ed111e5 refactor(kernel): move extra_opaque_converter to kernel, and rename it to hint_converter 2015-06-30 12:59:28 -07:00
Leonardo de Moura
e635d9be9f refactor(kernel): rename get_weight to get_height at declaration 
Motivation:
- It is the standard name for the concept: declaration height
- Avoid confusion with the expression weight
 2015-06-30 12:59:10 -07:00
Leonardo de Moura
1e6550eda6 feat(kernel/default_converter): take into account the 'theorem' annotations in the converability checker 
The idea is that we should seldon need to unfold theorems.
The convertability checker should use that.
When the convertability checker was implemented, theorems were opaque in
Lean. So, this hint was not needed.

This modification is another workaround for the performance problem with
the migrate command at library/data/real/division.lean.
This solution is better than applying proof irrelevance eagerly because
it also addresses similar problems in the HoTT library which does not
support proof irrelevance.

This commit also enables the conv_opt for all theorems.
 2015-06-30 10:42:26 -07:00
Leonardo de Moura
acc0832928 feat(kernel/default_converter): add "eager" proof irrelevance optimization 2015-06-29 19:10:39 -07:00
Leonardo de Moura
a2389fb664 refactor(kernel/default_converter): move delay_check code to postpone_is_def_eq method 2015-06-22 13:59:32 -07:00
Leonardo de Moura
76477aedd1 refactor(kernel): move is_stuck predicate to converter 2015-06-22 11:37:14 -07:00
Leonardo de Moura
54496709a2 refactor(kernel/default_converter): add virtual method reduce_def_eq 
The idea is to allow us to customize the default converter outside of
the kernel.
 2015-06-22 10:04:19 -07:00
Leonardo de Moura
a518a45239 refactor(kernel/default_converter): separate lazy_delta_reduction procedure 2015-06-22 08:45:05 -07:00
Leonardo de Moura
3626bd83bf refactor(util/sexpr/format): remove format constructors using std::initializer_list 
For some reason lean.js (Lean compiled using emscripten) crashes when
this kind of constructor is used.
 2015-06-20 14:02:18 -07:00
Leonardo de Moura
accc9dc38b chore(util/sexpr/format): remove unnecessary method 2015-06-19 20:11:42 -07:00
Leonardo de Moura
70fc05294b refactor(library/local_context): avoid hack in local_context 2015-06-18 15:41:00 -07:00
Leonardo de Moura
14c33c4e01 feat(kernel/error_msgs): add workaround for issue #669 
This issue should fix the new problem reported at #669.
It only disables beta-reduction in the pretty printer for application
type mismatch at (f n) when f is a lambda expression.
 2015-06-17 12:00:25 -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
25cbc5c154 fix(kernel/expr): remove 'cast_binder_info' 
We should put it back when we decide to implement it.
We also fix the bogus comment at src/frontends/lean/parser.cpp.

see issue #667
 2015-06-11 18:11:22 -07:00
Leonardo de Moura
8b062b4448 feat(kernel/default_converter): add support for stuck terms at try_eta_expansion_core 2015-06-10 15:19:39 -07:00
Leonardo de Moura
f7dd85a935 chore(*): fix 'breif' typos 2015-06-08 16:54:55 -07:00
Leonardo de Moura
496189feed feat(kernel/default_converter): cache failures for (f t =?= f s) heuristic 
See item 4 at
https://groups.google.com/forum/?utm_medium=email&utm_source=footer#!msg/lean-discuss/oJXwW5wT2Ds/d1Dgr8B-pE0J

See also
https://github.com/leanprover/lean/pull/659
 2015-06-08 10:41:30 -07:00
Leonardo de Moura
0e099b5fd8 feat(library/tactic/rewrite_tactic): apply beta&eta reduction before rewriting steps, add option 'rewrite.beta_eta' (default true) to control new feature. 2015-06-06 20:43:52 -07:00
Leonardo de Moura
d0d3f9bb41 refactor(kernel,library,frontends/lean): add helper functions, and cleanup collect_locals 2015-06-05 17:29:36 -07:00
Leonardo de Moura
4d52d4c790 fix(library/init/quot): prove quot.exact 2015-06-05 08:04:55 -07:00
Floris van Doorn
ff01774fd7 renaming(hit): rename type_quotient to quotient, and quotient to set_quotient 
This renaming is because type_quotient is a nonstandard name.  I've had a discussion with Egbert
Rijke, Steve Awodey and Dan Licata, and the consensus for a better name was 'quotient'.  I had to
make changes in src/kernel/hits/hits.cpp, I renamed g_type_quotient* by g_hit_quotient* (to avoid
name clash the standard library quotient, although I don't know whether that name clash would
matter).
 2015-06-04 20:14:13 -04:00
Leonardo de Moura
7a0e198147 feat(kernel,frontends/lean/builtin_cmds): allow kernel extensions to report their builtin constants 2015-05-29 16:28:16 -07:00
Leonardo de Moura
0ceedbe69e fix(library/normalize): fixes #640 2015-05-29 15:58:59 -07:00
Leonardo de Moura
004ea80e65 fix(library/tactic/rewrite_tactic): apply beta reduction when selecting patterns 2015-05-24 18:44:30 -07:00