aboutsummaryrefslogtreecommitdiff
path: root/kernel
AgeCommit message (Collapse)Author
2011-03-11Tentative to make unification check types at every instanciation of anmsozeau
evar, and simultaneously make type inference with universes work better. This only exports more functions from kernel/univ, to be able to work with a set of universe variables during type inference. Universe constraints are gradually added during type checking, adding information necessary e.g. to lower the level of unknown Type variables to Prop or Set. There does not seem to be a disastrous performance hit on the stdlib, but might have one on some contribs (hence the "Tentative"). git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13905 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-03-08adding eta in the vmbgregoir
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13896 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-03-05Moving printing of module typing errors upwards to himsg.ml so as toherbelin
be able to call term printers. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13886 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-03-05Starting being more explicit on the reasons why module subtyping fails.herbelin
Note: I'm unsure about some subtyping error case apparently involving aliases of inductive types (middle of Subtyping.check_inductive); I bound it to some NotEqualInductiveAliases error, but this has to be checked. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13885 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-03-05Fixed a "feature" about subtyping records: one was expected not onlyherbelin
that the fields had the same names but that the parameters of the record had exactly the same names too. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13879 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-02-24Mod_subst: improving sharing of subst_mpsletouzey
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13854 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-02-23Some more simplification in Mod_substletouzey
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13852 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-02-17- Use transparency information all the way through unification andmsozeau
conversion. - Fix trans_fconv* to use evars correctly. - Normalize the goal with respect to evars before rewriting in [rewrite], allowing to see instanciations from other subgoals. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13844 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-02-11Mod_typing: some refactoring (common parts about MSEapply and co)letouzey
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13838 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-02-11Safe_typing: some refactoring (avoiding copy-paste of record definitions)letouzey
Many of the record definitions for new safe_environment follow the same pattern, we factorize them in a generic add_field function. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13837 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-02-11Mod_subt: some more refactoring, substitution is also separated in two tablesletouzey
This way, no more mixing of MBI / MPI. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13836 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-02-11Mod_subst: split delta_resolver in two tables (mp / kn)letouzey
This way, no more absurd mixing of mp/kn and Prefix_equiv/Equiv to consider, and hence no more anomaly or assert false left in Mod_subst. As we say here, "faut pas melanger les torchons et les serviettes" ... With these two specialized tables, efficiency might also be better. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13835 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-02-11compatibility <3.12 (Map.exists Map.singleton)pboutill
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13829 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-01-31A fine-grain control of inlining at functor application via priority levelsletouzey
As said in CHANGES: << The inlining done during application of functors can now be controlled more precisely. In addition to the "!F G" syntax preventing any inlining, we can now use a priority level to select parameters to inline : "<30>F G" means "only inline in F the parameters whose levels are <= 30". The level of a parameter can be fixed by "Parameter Inline(30) foo". When levels aren't given, the default value is 100. One can also use the flag "Set Inline Level ..." to set a level. >> Nota : the syntax "Parameter Inline(30) foo" is equivalent to "Set Inline Level 30. Parameter Inline foo.", and "Include <30>F G" is equivalent to "Set Inline Level 30. Include F G." For instance, in ZBinary, eq is @Logic.eq and should rather be inlined, while in BigZ, eq is (fun x y => [x]=[y]) and should rather not be inlined. We could achieve this behavior by setting a level such as 30 to the parameter eq, and then tweaking the current level when applying functors. This idea of levels might be too restrictive, we'll see, but at least the implementation of this change was quite simple. There might be situation where parameters cannot be linearly ordered according to their "inlinablility". For these cases, we would need to mention names to inline or not at a functor application, and this is a bit more tricky (and might be a pain to use if there are many names). No documentation for the moment, since this feature is experimental and might still evolve. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13807 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-01-28Remove the "Boxed" syntaxes and the const_entry_boxed fieldletouzey
According to B. Gregoire, this stuff is obsolete. Fine control on when to launch the VM in conversion problems is now provided by VMcast. We were already almost never boxing definitions anymore in stdlib files. "(Un)Boxed Definition foo" will now trigger a parsing error, same with Fixpoint. The option "(Un)Set Boxed Definitions" aren't there anymore, but tolerated (as no-ops), since unknown options raise a warning instead of an error by default. Some more cleaning could be done in the vm. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13806 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-01-27Make simpl use the proper constant when folding (mutual) fixpointsletouzey
Before this commit, when simpl was finding the constant name for folding some (mutual) fixpoint, this was done via some repr_con followed by make_con. Problem: this doesn't preserve the canonical part of a Names.constant. For instance the following script was buggish: Module M. Fixpoint foo n := match n with O => O | S n => bar n end with bar n := match n with O => O | S n => foo n end. End M. Module N. Include M. (* foo, bar have here "user name" N but "canonical name" M *) Eval simpl in (fun x => bar (S x)). (* Anomaly: uncaught exception Failure "Cannot print a global reference". *) (* since simpl has produce a different bar with both user and canonical N *) TODO : check all other uses of make_con in the rest of the sources... git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13803 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-01-25Rewrite sort_universes to minimize the number of universesglondu
The whole standard library fits in 3 universes (instead of 22 with the previous algorithm). Very costy, time complexity: O(n^4) (if we assume map operations are done in constant time), with n being the number of universe variables. It takes ~35s for the whole stdlib. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13796 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-01-11In univ.ml, put universe_level primitives in its their own sub-moduleglondu
This is merely refactoring so that UniverseL{Map,Set} constructions appear nicely. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13788 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-01-11Add "Print Sorted Universes"glondu
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13786 85f007b7-540e-0410-9357-904b9bb8a0f7
2011-01-11More coherent comment orderingglondu
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13784 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-12-19Fixing bug #2454: inversion predicate strategy for inferring the typeherbelin
of "match" is not general enough; if there is a non dependent type constraint, we also try w/o inversion predicate in the return clause. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13727 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-12-18Univ: try to avoid a few lookup in the universe graphletouzey
Sometimes the same (repr g u) was done in different functions after being passed u as argument. We rather try to compute (repr g u) as soon as possible, and then pass it instead of u. Beware of sync issues : if g changes, arcu might become obsolete (cf. setlt, setleq, merge ...) Typical code around occurences of declare_univ was doing up to 3 lookups: - is u in g ? - if not we descend again in g to add it - and then later repr is called on the same u. With my safe_repr, we do one lookup if u is in g, and a lookup and an addition otherwise. Ok, declare_univ was rarely used, but it seems nicer this way. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13726 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-12-18Univ.constraints made fully abstract instead of being a Set of abstract stuffletouzey
No need to tell the world about the fact that constraints are implemented via caml's Set. Other modules just need to know about the empty and union functions (and addition functions "enforce_geq" and "enforce_eq" that were already there). git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13725 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-12-18Revert last commit 13723 on Univ : minor gains and more complex codeletouzey
The gains on contribs are quite small, around 3% max, apart from 3 small contribs where it's about 10% (corresponding to 10s each). With last patch, we add quicker lookup for universes in the graph (up to 5 times less calls to cmp_univ_level on an example), but probably more "administrative" work (i.e. addition of updated paths in the graphs, handling pairs of updated graphs and results in functions, etc), and some sharing might also have been lost since graphs changed more. Anyway, little gain and more complex code, let's remove this patch for now ... until the next attempt to speed-up the universe layer. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13724 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-12-17Univ: an attempt to lazily compact chains of Equiv in a functionnal wayletouzey
We'll see experimentally if this helps... A few more functions could be adapted (e.g. between), and an occurence of compare just discard the compacted graph (in compare_greater) git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13723 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-12-16Univ: two improvements (speed + space)letouzey
- The "compare" function on universes (the one answering EQ|LT|LE|NLE) was launching "collect" for creating the transitive upward closure of u, and then checking if v is in it. We now proceed more lazily, by stopping creating the transitive closure as soon as v is found. - In univ_entry, the first arg u of Equiv(u,v) is removed. It can indeed be retrieved from the key of the universe graph leading to this Equiv. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13719 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-12-01* Kernel/Term:regisgia
- Fix a bug in [comp_term] (casts were ignored). - Improve the efficiency of hash table lookup. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13662 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-12-01* Kernel/Termregisgia
Remove an unsound optimization. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13661 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-12-01* Kernel/Termregisgia
Fix an efficiency bug when hash-consing deep [constr]s. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13660 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-11-15Minor fixes from Gregory Malecha. A typo fixed and a better (located) msozeau
error message in case of unnammed record parameters. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13635 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-11-03Correction bug 2427soubiran
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13617 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-11-03Remove suspiciously named "implicit" stuff from Termglondu
The lambda_implicit series of functions are used only in Indtypes, so we move them there. In the checker, they are already there... git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13615 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-11-03In Univ, unify order_request and constraint_typeglondu
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13614 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-11-02More generic Univ.dump_universesglondu
Instead of formatting directly to an output channel, provide an output function that handles formatting and I/O. This allows changing the output format. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13610 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-10-05Reintroduce kind_of_type (used by Presburger contrib)glondu
This partially reverts commit r13467. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13497 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-10-04Forgotten lifts in eta-expansionglondu
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13494 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-10-03Fixing printing of module_path names (was using a debuggingherbelin
message since r11177). git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13486 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-09-28Remove kind_of_type, kind_of_term2 (dead code)glondu
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13467 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-09-24Partial review of removed dead code (r13460)herbelin
- The code detected unused in notation.ml revealed a bug. - In term.ml, restored a (short) useless function for consistency/symmetry of the interface. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13463 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-09-24Some dead code removal, thanks to Oug analyzerletouzey
In particular, the unused lib/tlm.ml and lib/gset.ml are removed In addition, to simplify code, Libobject.record_object returning only the ('a->obj) function, which is enough almost all the time. Use Libobject.record_object_full if you really need also the (obj->'a). git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13460 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-09-23Fix inconsistency in Prop/Set conversion checkglondu
This commit fixes a bug that made the system inconsistent with proof irrelevance (the main idea being that Set = Prop by reflexivity). git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13450 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-09-20Added eta-expansion in kernel, type inference and tactic unification,herbelin
governed in the latter case by a flag since (useful e.g. for setoid rewriting which otherwise loops as it is implemented). git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13443 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-09-15Sharing is not anymore broken by traverse_module.soubiran
+commit r13412 git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13418 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-09-15Fix likely semantic typosglondu
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13417 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-08-27* kernel/Safe_typing.LightenLibrary:regisgia
Fix an incorrect initialization of the index counter. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13386 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-08-27* (checker|kernel)/Safe_typing:regisgia
Rename "lighten_*" into "traverse_*" inside the [traverse_library] function. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13385 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-08-27* (checker|kernel)/Safe_typing:regisgia
Fix a bug in traverse_library. The extraction was not satisfied with [mod_expr] set to [None]. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13384 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-08-27* checker/SafeTyping kernel/SafeTyping:regisgia
Fix typos. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13380 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-08-27* Improve documentation of LightenLibrary.regisgia
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13378 85f007b7-540e-0410-9357-904b9bb8a0f7
2010-08-27* (checker|kernel)/Safe_typing: New LightenLibrary.regisgia
This module introduces an indirection behind opaque const_body to enable the optional demarshalling of them. * library/Library checker/Check: Use LightenLibrary. git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13377 85f007b7-540e-0410-9357-904b9bb8a0f7
generated by cgit v1.2.3 (git 2.25.1) at 2026-03-01 10:05:43 +0000