Stream: MetaCoq

Topic: ✔ `cored` and `extends_decls`

Jason Gross (Dec 06 2022 at 20:51):

I want a theorem proving

  forall (cf : checker_flags) (Σ Σ' : global_env),
  extends_decls Σ Σ' ->
  wf Σ ->
  wf Σ' ->
  forall (Γ : context) (u u' : Universes.universes_decl),
  wf_local (Σ, u) Γ ->
  wf_local (Σ', u') Γ ->
  forall x y : PCUICAst.term,
  PCUICSafeLemmata.cored Σ Γ x y -> PCUICSafeLemmata.cored Σ' Γ x y

Does it exist? If not, where should I add it?

Théo Winterhalter (Dec 06 2022 at 21:01):

So it's just weakening of cored for the global environment right? We should have it for red and since cored x y is (logically equivalent to) red y x it should work? I don't see why you would need all these hypotheses though.

Jason Gross (Dec 06 2022 at 21:11):

Yes. I don't need all the hypotheses, that's just what I have available.

Jason Gross (Dec 06 2022 at 21:12):

It looks like we only have cored logically equivalent to ∥ redp ∥ and redp a subrelation of red, though?

Jason Gross (Dec 06 2022 at 21:13):

What's the weakening lemma called for red?

Jason Gross (Dec 06 2022 at 21:14):

(I only see one for context weakening, not global env weakening...)

Jason Gross (Dec 06 2022 at 21:25):

Ah, found it in PCUICWeakeningEnvConv, weakening_env_red1

Meven Lennon-Bertrand (Dec 06 2022 at 21:29):

Looks like it is here for one-step reduction, but there does not seem to be a corresponding lemma for the transitive closure…

Meven Lennon-Bertrand (Dec 06 2022 at 21:30):

Oops, my bad, you found it already

Jason Gross (Dec 06 2022 at 21:31):

Yeah, but I sent my "found" notice to the wrong stream

Théo Winterhalter (Dec 07 2022 at 07:28):

Oh I don't know what redp is. In any case, I don't think it's desirable to have such a lemma in the library since it's just a tool for termination, we don't need to prove more stuff about it.

Yannick Forster (Dec 07 2022 at 07:32):

Mh, but we might prove more stuff about termination in future, no? If Jason proved it, I'm in favour of including it in upstream MetaCoq. It won't create issues for other code that's already there, and it might save whoever will need it in the future the time it would take to reprove it

Théo Winterhalter (Dec 07 2022 at 07:34):

I would guess not but as you say it does no harm (except in making the development grow in size).

Jason Gross (Dec 07 2022 at 22:21):

I'm not using it locally in my development, but in making normalization into a typeclass.

Théo Winterhalter (Dec 08 2022 at 09:05):

Oh you need it there? Then I guess we must accept it. :)

Jason Gross (Dec 08 2022 at 09:23):

PR at https://github.com/MetaCoq/metacoq/pull/801

The proof is needed because erasure needs normalization not just in the global environment it's passed, but also in the env that results from popping off decls.

Théo Winterhalter (Dec 08 2022 at 09:26):

I'm surprised because I thought cored would never appear in a setting where you change the global environment. Anyway we can merge it I guess.

Jason Gross (Dec 08 2022 at 09:29):

Thank you!

Notification Bot (Dec 08 2022 at 09:29):

Jason Gross has marked this topic as resolved.

Last updated: May 31 2023 at 02:31 UTC