Changes between Version 1 and Version 2 of Ticket #25393, comment 16


Ignore:
Timestamp:
05/31/18 12:04:15 (3 years ago)
Author:
egourgoulhon
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #25393, comment 16

    v1 v2  
    33>
    44
    5 Actually the type of immutatibility involved for `ContinuousMap` is that defined by Simon King in this [https://groups.google.com/d/msg/sage-devel/d8-r3uEewFU/Jb02875ACgAJ sage-devel post]: methods, like `add_expr`, may change the internal representation of the object, but in any case they must preserve the equivalence class with respect to `==` to which the object belongs at its creation.
     5Actually the type of immutatibility involved for `ContinuousMap` is that defined by Simon King in this [https://groups.google.com/d/msg/sage-devel/d8-r3uEewFU/Jb02875ACgAJ sage-devel post]: methods, like `add_expr`, may change (enhance) the internal representation of the object, but in any case they must preserve the equivalence class with respect to `==` to which the object belongs at its creation.
    66
    77> Based on your explanation though, maybe I should back off a bit on my argument that `Morphism.__hash__` should be removed (unless someone can point out other subclasses of it that are definitely ''not'' in any sense immutable).  Instead, I might just change the implementation a bit to do away with its use of `repr()`.  If the domain doesn't have generators we can just ignore that and use `None` or something.  IMO it's the `repr()` that's the real problem because that's just flakey and surprising.  So maybe I'll just change that for now and see what that affects.