#14356: memleak in UniqueRepresentation / Poset
---------------------------+--------------------------
Reporter: vbraun | Owner: rlm
Type: defect | Status: needs_info
Priority: major | Milestone: sage-5.12
Component: memleak | Resolution:
Keywords: | Merged in:
Authors: | Reviewers:
Report Upstream: N/A | Work issues:
Branch: | Commit:
Dependencies: | Stopgaps:
---------------------------+--------------------------
Comment (by nbruin):
I think I agree with Simon's assessment that the code behaves as designed
and hence that strictly speaking, the memory leak is created by misuse.
However, I think we have a problematic design if it's that easy to
"misuse" it so easily.
I think this issue indicates that perhaps parents like Poset should NOT be
unique. If we go back to why "unique" parent were introduced we see:
- it means that polynomials in QQ!['x'] created in two separate places
will actually lie in the same parent
- it makes some things easier/faster in the coercion framework
Keeping strong references globally to the construction parameters of a
parent for the lifetime of the parent is OK if the parameters are
fundamentally "simpler" than the constructed object. However, that's not
the case with constructing a Poset: the entire complexity of the object is
part of the construction call.
That makes it rather unlikely that the uniquerepresentation feature will
ever make a useful contribution. It also means that constructing an
n-element Poset is at least O(n), because !UniqueRepresentation will walk
the call argument (i.e., all the labels!) to hash them, converting lists
to something hashable. Note that constructing a Poset from a set should
normally be O(1), because the only thing required is saving a reference to
the set and the comparison method.
So, I think parents that require more than O(1) [in some reasonable sense]
construction parameter complexity should probably usually not be
!UniqueRepresentation.
Poset seems to have explicitly designed to be !UniqueRepresentation, so I
assume someone really needed that. It does mean, though, that if someone
needs a Poset but doesn't particularly need it to be
!UniqueRepresentation, they should probably not use Poset, because it's an
expensive feature for parents like this.
--
Ticket URL: <http://trac.sagemath.org/ticket/14356#comment:6>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sage-trac.
For more options, visit https://groups.google.com/groups/opt_out.
