| Age | Commit message (Collapse) | Author |
|---|
|
lemmes se terminant par False ou not sur n'importe quelle formule
(cela crée trop d'incompatibilités dans les "try apply" etc.); de
toutes façons, "contradict" joue presque ce rôle (à ceci près qu'il
ne traverse pas les conjonctions) (tactics/tactics.ml).
- Quelques corrections sur RIneq.v
- le hint Rlt_not_eq avait été oublié dans la phase de restructuration,
- davantage de noms canoniques (O -> 0, etc.),
- nouvelle tentative de ramener "auto" vers Rle (avec Rle_ge) plutôt
que vers Rge qui est moins souvent associé à des hints.
- Utilisation du formateur deep_ft pour afficher les scripts de preuve afin
d'éviter le besoin d'un "Set Printing Depth" (vernacentries.ml).
- Suppression de certaines utilisations de l'Anomaly de meta_fvalue
qui ne correspondaient pas à des comportements anormaux (reductionops.ml).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10760 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
- Allow unicode superscripts characters.
- Put Program notations in scopes.
- Correct priority semantics in typeclasses eauto.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10759 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
(for instance as a consequence of the Set/Type switch of a library
such as FSets). After a "Require Export SetIsType.", Set becomes a
mere notation for Type (I love notations !).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10757 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Correction bugs commentaires pour coqdoc (ChoiceFacts.v)
Test-suite
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10756 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Hints reste à faire. (dommage que Rge et Rle ne soient pas
convertibles)
- Ajout de Nnat et Ndigits dans NArith.v
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10751 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
As for FSetAVL vs. FSetFullAVL, preservation of the balancing of trees
is not necessary anymore for just fulfilling the Map interface. But we
still want theses proofs to exists somewhere, since they ensure the
correct efficiency of the FMapAVL operations.
In addition, FSetFullAVL also contains the non-structural,
ocaml-faithful version of compare.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10748 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
* pure functions comes first, proofs are isolated in a sub-module
* lazy (&&&) = if ... then ... else true instead of strict (&&) = andb
* equal and compare done via continuations
* a more clever map2_opt suggested by B. Gregoire: no more
intermediate conversion to lists, some sub-functions can handle
trivial situations, etc.
* map2 is now done via map2_opt and another new iterator: map_option.
By the way, this map_option allows to define easily an efficient
filter function. Both map2_opt and map_option are currently not
available through FMapInterface.S, but they can be used by direct
reference to the underlying Raw module.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10745 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
alleviate some problems with delta. Better precedence in lambda
notation. Temporarily deactivate notations for relation conjunction,
equivalence and so on, while we search for a better syntax and maybe a
generalization (fixes bug #1820). Better destruct_call in
Program.Tactics.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10742 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
pas correctes
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10739 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10734 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
- Use support for abbreviations with params added by Hugo for inverse.
- Standard priorities for operators on relations.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10733 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
- Nouvel essai de prise en compte unfold dans apply (unification.ml)
- Correction bug commit précédent (constrintern.ml)
- Correction bug d'explication des evars non résolues (evarutil.ml)
- Fenêtre de query coqide plus large (command_windows.ml)
- Orthographe tauto (tauto.ml4)
- Crédits (ConstructiveEpsilon.v)
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10731 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
to "-R>" and the like. Split more precisely in inference of case
predicate between the new code which currently works only for matched
variables and the old one which works better on variables appearing in matched
terms types (the two could also be merged).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10729 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Add definitions of relational algebra in Classes/RelationClasses
including equivalence, inclusion, conjunction and disjunction. Add
PartialOrder class and show that we have a partial order on relations.
Change SubRelation to subrelation for consistency with the standard
library. The caracterization of PartialOrder is a bit original: we
require an equivalence and a preorder so that the equivalence relation
is equivalent to the conjunction of the order relation and its
inverse. We can derive antisymmetry and appropriate morphism instances
from this. Also add a fully general heterogeneous definition of
respectful from which we can build the non-dependent respectful
combinator.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10728 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
[in I] := t [return pred] in b", just as SSReflect does with let:.
Change implementation: no longer a separate AST node, just add a case_style
annotation on Cases to indicate it (if ML was dependently typed we
could ensure that LetPatternStyle Cases have only one term to be
matched and one branch, alas...). This factors out most code and we lose
no functionality (win ! win !). Add LetPat.v test suite.
- Slight improvement of inference of return clauses for dependent
pattern matching. If matching a variable of non-dependent type
under a tycon that mentions it while giving no return clause, the
dependency will be automatically infered. Examples at the end of
DepPat. Should get rid of most explicit returns under tycons.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10727 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
d'un test non significatif
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10726 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
but with lazy behavior while (vm_)computing
- FSetAVL.split has now a dedicated output type instead of ( ,( , ))
- Some proof adaptations.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10725 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
- Better interface in constrintern w.r.t. evars used during typechecking
- Add "unsatisfiable_constraints" exception which gives back the raw
evar_map that was not satisfied during typeclass search (presentation
could be improved).
- Correctly infer the minimal sort for typeclasses declared as
definitions (everything was in type before).
- Really handle priorities in typeclass eauto: goals produced with higher
priority (lowest number) instances are tried before other of lower
priority goals, regardless of the number of subgoals.
- Change inverse to a notation for flip, now that universe polymorphic
definitions are handled correctly.
- Add EquivalenceDec class similar to SetoidDec, declaring decision
procedures for equivalences.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10724 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
context [] pattern). May break some user contribs...
Rename clsubstitute to substitute.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10716 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
types. Change (again) the semantics of bindings and the binding modifier
! in typeclasses. Inside [ ], implicit binding where only parameters
need to be given is the default, use ! to use explicit binding, which is
equivalent to regular bindings except for generalization of free
variables. Outside brackets (e.g. on the right of instance
declarations), explicit binding is the default, and implicit binding
can be used by adding ! in front. This avoids almost every use of ! in
the standard library and in other examples I have.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10713 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10712 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10711 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
- Renommage de Rlt_not_le de Fourier_util en Rlt_not_le_frac_opp pour
éviter la confusion avec le Rlt_not_le de RIneq.
- Quelques variantes de lemmes en plus dans RIneq.
- Déplacement des énoncés de sigT dans sig (y compris la complétude)
et utilisation de la notation { l:R | }.
- Suppression hypothèse inutile de ln_exists1.
- Ajout notation ² pour Rsqr.
Au passage:
- Déplacement de dec_inh_nat_subset_has_unique_least_element
de ChoiceFacts vers Wf_nat.
- Correction de l'espace en trop dans les notations de Specif.v liées à "&".
- MAJ fichier CHANGES
Note: il reste un axiome dans Ranalysis (raison technique: Ltac ne
sait pas manipuler un terme ouvert) et dans Rtrigo.v ("sin PI/2 = 1"
non prouvé).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10710 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
symmetric... classes for now, merging Relations with RelationClasses
requires some non-trivial changes on implicits but also some definitions
are different (e.g. antisymmetric in Classes is defined w.r.t an
equivalence relation and not eq). Move back to Reflexive and so on.
reflexivity-like tactics fail in the same way as before if Setoid was
not imported. There is now a scope for morphism signatures, hence "==>",
"++>" and "-->" can be given different interpretations.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10708 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
functional scheme.
- Proofs are placed in Raw.Proofs, that may someday become an opaque module.
- use Functional Scheme in this module Raw.proofs, instead of Function: less
derived objects.
- more cleanup.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10707 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Extraction is unchanged, since orb/andb are detected as un-strict
functions and inlined. This only prevents the use of some potential
clever Extract Inlined Constant andb => "(&&)" and idem for orb, but
this isn't a big deal.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10705 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
rewrite calls fail because an occurence is found under a lambda that
was not found before and there are no adequate morphism declarations to
make the rewrite succeed nor the possibility to give occurences to
rewrite (yet). Only setoid_rewrite will try rewriting under lambda's for
now. Example problem found in a port of the Random library.
Also improved the required_library error message.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10704 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10701 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
(Cachan/IntMap)
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10699 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10698 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
names, gives the ability to specify qualified classes in instance
declarations. Use that in the class_tactics code.
Refine the implementation of classes. For singleton classes the
implementation of the class becomes a regular definition (into Type or
Prop). The single method becomes a 'trivial' projection that allows to
launch typeclass resolution. Each instance is just a definition as
usual. Examples in theories/Classes/RelationClasses. This permits to
define [Class reflexive A (R : relation A) := refl : forall x, R x
x.]. The definition of [reflexive] that is generated is the same as the
original one. We just need a way to declare arbitrary lemmas as
instances of a particular class to retrofit existing reflexivity lemmas
as typeclass instances of the [reflexive] class.
Also debug rewriting under binders in setoid_rewrite to allow rewriting
with lemmas which capture the bound variables when applied (works only
with setoid_rewrite, as rewrite first matches the lemma with the entire,
closed term). One can rewrite with [H : forall x, R (f x) (g x)] in the goal
[exists x, P (f x)].
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10697 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Adds some instances that incur again a small performance penalty because they cannot
be discriminated against for now (the discrimination net is considering
the head constructor to be Morphism).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10694 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
rewriting, in some sense bootstrapping the system. Remove redundant
morphisms for now to test for completeness (small performance penalty).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10689 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
resolution. Add [relation] and Setoid's [equiv] as such objects.
Considerably simplify resolve_all_evars for typeclass resolution, adding
a further refinement (and hack): evars get classified as non-resolvable
(using the evar_extra dynamic field) if they are turned into a
goal. This makes it possible to perform nested typeclass resolution
without looping. We
take advantage of that in Classes/Morphisms where [subrelation_tac] is
added to the [Morphism] search procedure and calls the apply tactic which
itself triggers typeclass resolution. Having [subrelation_tac] as a tactic
instead of an instance, we can actually force that it is applied only
once in each search branch and avoid looping.
We could get rid of the hack when we have real goals-as-evars
functionality (hint hint).
Also fix some test-suite scripts which were still calling [refl]
instead of [reflexivity].
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10687 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
check_required_library where needed (fixes bug #1797).
Remove spurious messages from ring when not in verbose mode.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10685 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
setoid rewrite. Refine and use the new unification flags setup by Hugo
to do a little bit of delta in clenv_unify/w_unify. Moved from a boolean
indicating conversion is wanted to a Cpred representing the
constants one wants to get unfolded to have more precise control. Add
corresponding test-suite file.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10684 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Program.Syntax along with stronger implicits. Update dependent induction
test file using unicode syntax.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10682 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
implicits for left, inl or eq, hence some theories had to be changed
again. It should make some user contribs compile again too. Also
do not import functional extensionality when importing Program.Basics,
add a Combinators file for proofs requiring it and a Syntax file for the
implicit settings. Move Classes.Relations to Classes.RelationClasses to
avoid name clash warnings as advised by Hugo and Pierre.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10681 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
[id] instead of their expansions. Seems to slow things down a bit
(1s. on Ring_polynom).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10680 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
* FSetAVL is greatly lightened : modulo a minor change in bal, formal
proofs of avl invariant is not needed for building a functor
implementing FSetInterface.S (even if this invariant is still true)
* Non-structural functions (union, subset, compare, equal) are
transformed into efficient structural versions
* Both proofs of avl preservation and non-structural functions are
moved to a new file FSetFullAVL, that can be ignored by the average
user.
Same for FMapAVL (still work in progress...)
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10679 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
works in inductive type definitions and fixpoints. The semantics of
an implicit inductive parameter is maybe a bit weird: it is implicit in the
inductive definition of constructors and the contructor types but not in
the inductive type itself (ie. to model the fact that one rarely wants A
in vector A n to be implicit but in vnil yes). Example in test-suite/
Also, correct the handling of the implicit arguments across sections.
Every definition which had no explicitely given implicit arguments was
treated as if we asked to do global automatic implicit arguments on
section closing. Hence some arguments were given implicit status for no
apparent reason.
Also correct and test the parsing rule which disambiguates between {wf
..} and {A ..}.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10677 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10672 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
morphisms on R --> inverse R' for any R, R' (report by N. Tabareau).
Better implementation of unfolding for impl that does it only where
necessary to keep the goal in the same shape as the user gave.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10648 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
implementation. Mostly syntax changes when declaring parametric
relations, but also some declarations were relying on "default"
relations on some carrier. Added a new DefaultRelation type class that
allows to do that, falling back to the last declared Equivalence
relation by default. This will be bound to Add Relation in the next
commit.
Also, move the "left" and "right" notations in Program.Utils to "in_left" and
"in_right" to avoid clashes with existing scripts.
Minor change to record to allow choosing the name of the argument for
the record in projections to avoid possible incompatibilities.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10639 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
* "f_equal" is now a tactic in ML (placed alongside congruence since
it uses it). Normally, it should be completely compatible with the
former Ltac version, except that it doesn't suffer anymore from the
"up to 5 args" earlier limitation.
* "revert" also becomes an ML tactic. This doesn't bring any real
improvement, just some more uniformity with clear and generalize.
* The experimental "narrow" tactic is removed from Tactics.v, and
replaced by an evolution of the old & undocumented "specialize"
ML tactic:
- when specialize is called on an hyp H, the specialization is
now done in place on H. For instance "specialize (H t u v)"
removes the three leading forall of H and intantiates them by
t u and v.
- otherwise specialize still works as before (i.e. as a kind of
generalize).
See the RefMan and test-suite/accept/specialize.v for more infos.
Btw, specialize can still accept an optional number for specifying
how many premises to instantiate. This number should normally
be useless now (some autodetection mecanism added). Hence this
feature is left undocumented. For the happy few still using
specialize in the old manner, beware of the slight incompatibities...
* finally, "contradict" is left as Ltac in Tactics.v, but it has
now a better shape (accepts unfolded nots and/or things in Type),
and also some documentation in the RefMan
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10637 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10636 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
of the fix I added an optional "by" annotation for rewrite to solve said
conditions in the same tactic call. Most of the theories have been
updated, only FSets is missing, Pierre will take care of it.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10634 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
out. The semantics of the old setoid are faithfully simulated by the new
tactic, hence no scripts involving rewrite are modified. However,
parametric morphism declarations need to be changed, but there are only a
few in the standard library, notably in FSets. The declaration and the
introduction of variables in the script need to be tweaked a bit,
otherwise the proofs remain unchanged.
Some fragile scripts not introducting their variable names explicitely
were broken. Requiring Setoid requires Program.Basics which sets stronger
implicit arguments on some constants, a few scripts benefit from that.
Ring/field have been ported but do not really use the new typeclass
architecture as well as they could. Performance should be mostly
unchanged, but will certainly improve in the near future. Size of the
vo's seems not to have changed at all.
It will certainly break some contribs using Setoid.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10631 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10630 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
binding, add "!" syntax for the new binders which require parameters and
not superclasses. Change backquotes for curly braces for user-given
implicit arguments, following tradition. This requires a hack a la
lpar-id-coloneq. Change ident to global for typeclass names in class
binders. Also requires a similar hack to distinguish between [ C t1 tn ]
and [ c : C t1 tn ]. Update affected theories.
While hacking the parsing of { wf }, factorized the two versions of fix
annotation parsing that were present in g_constr and g_vernac.
Add the possibility of the user optionaly giving the priority for resolve and
exact hints (used by type classes). Syntax not fixed yet: a natural
after the list of lemmas in "Hint Resolve" syntax, a natural after a "|"
after the instance constraint in Instance declarations (ex in
Morphisms.v).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@10628 85f007b7-540e-0410-9357-904b9bb8a0f7
|
