2022 09 22 Meeting - math-comp/analysis GitHub Wiki

Participants: Affeldt, Cohen, Roux, Stone

port to HB #698
- in topology.v:
  - find a way to port Definition product_topologicalType
    - does not seem to require any special technique, should be a standard approach, i.e., define a symbol product_... which is a Type that is endowed with a canonical structure, etc.
    - TODO: Reynald to investigate in October
  - non forgetful inheritance warning at line 3840
    - is it related to the fact that the uniform topology does not depend on the topology of the domain?
      - maybe no
    - actually, this should be an error
    - isSource was never intended to be a structure, it was a technical artifact
      - rule of thumb: ignore the warning and leave an issue?
    - TODO: Cyril to investigate
  - Section RestrictedUniformTopology.: might be connect to the non forgetful inheritance warning
  - uniform_limit_continuous and another one: ...
- in normedtype.v: depends on product topology
split packages classical #600 (this seems mostly ready, just waiting to happen)
- should be top priority
- we have added the split in nix and this should indeed be done first, the opam sync is easier
- PR not yet reviewed
- What do you do when the version is too low for the classical package to exist?
  - it becomes a dummy package that does nothing (trick borrowed from MathComp)
  - Cyril: would have wanted an error to be thrown. do not see where this is done in MathComp
    - https://github.com/NixOS/nixpkgs/blob/master/pkgs/development/coq-modules/mathcomp/default.nix#L112
    - it's a dummy package at the top not the bottom, it does introduce an extra dependency, it does not change the preexisting hierarchy
- this PR is orthogonal to the one HB but it might have a big impact
- TODO: merge what we have ASAP and try to improve later
PRs:
- https://github.com/math-comp/analysis/pull/747
  - easy one
  - NB: no unfold before an intro
- https://github.com/math-comp/analysis/pull/748
  - more interesting
  - problem with R and \bar R: we want both theories for integral in R and in \bar R
  - go to Bochner integral and integral on R would be Bochner1?
  - using a different notation would also make sense
  - Rintegral is transitory and only exist because we do not have Bochner
    - notation in ring_scope following the pattern for `| ... |
  - Bochner needs a notion of dual space
    - definition via a canonical embedding into a dual space
    - we were waiting for HB to do that
- https://github.com/math-comp/analysis/pull/737
  - report a curiousness about a need to a structure whereas the mixin only is needed
    - see isse #750
  - Cyril:
    - generalize [0,1] in IsPath
    - definition of path is not the right one
    - composition of paths is not associative
    - define path as a quotient up to remapping
  - use generic_quotient with same_path_component to define the equivalence relation
    - 1. reparametrization is an equivalence relation
    - 1. use that to build the canonical surjection
    - 1. define paths as point in the range of that space
  - TODO: redefine paths
    - Cyril: some proofs might shrink
- https://github.com/math-comp/analysis/pull/674
  - uniform space respected uniform and weak topology
  - Cyril to review?
- https://github.com/math-comp/analysis/pull/627
  - not worth merging now