#16244: Simplify TestSuite's some_elements role and logic (no random sampling)
-------------------------------------+-------------------------------------
Reporter: nthiery | Owner:
Type: defect | Status: needs_review
Priority: major | Milestone: sage-6.2
Component: doctest framework | Resolution:
Keywords: | Merged in:
Authors: Nicolas M. Thiéry | Reviewers:
Report Upstream: N/A | Work issues:
Branch: | Commit:
u/nthiery/simplify_testsuite_s_some_elements_role_and_logic__no_random_sampling_|
2486f804d260370ee1d9a7a9b3a7962a36ea0e65
Dependencies: | Stopgaps:
-------------------------------------+-------------------------------------
Changes (by nthiery):
* cc: roed, saraedum (added)
Old description:
> Currently `TestSuite` (more precisely `InstanceTester.some_elements`)
> tries to be fancy by choosing "some elements" using a random sample. The
> random sample is built using Python's `random.sample`, which requires its
> input to be a Sequence (the i-th element can be fetched with o[i]), or
> some dict-like object. This can get brittle with other kinds of inputs
> where `__getitem__` is used for other purposes, or where unranking is
> just computationally expensive. The `some_elements` method also assumes
> `__len__` to be implemented and cheap enough.
>
> Example:
> {{{
> FF = IntegerModRing(29)
> tester = FF._tester()
> list(tester.some_elements(CartesianProduct(FF, FF, FF)))
> *bang*
> }}}
>
> In fact the following is close to fail, and just works because the error
> raised during the sample is caught.
> {{{
> tester.some_elements(FF)
> }}}
>
> This ticket reduces the role of `InstanceTester.some_elements` to just do
> some argument parsing and ensure that at most "max_run" elements are
> returned. It only requires the input to be iterable.
>
> If we want to have fancy random samples, we should define a specific
> protocol (typically P.sample()) for it, or just let parents decide on the
> appropriate strategy by defining some_elements appropriately.
>
> This was originaly analyzed in #15919.
New description:
Since #14284, `TestSuite` (more precisely `InstanceTester.some_elements`)
tries to be fancy by choosing "some elements" using a random sample. The
random sample is built using Python's `random.sample`, which requires its
input to be a Sequence (i.e. the i-th element can be fetched with o[i]),
or some dict-like object. This can get brittle with inputs where
`__getitem__` is used for other purposes, or where unranking is just
computationally expensive. The `some_elements` method also assumes
`__len__` to be implemented and cheap enough.
Example:
{{{
sage: FF = IntegerModRing(29) # needs to be >21 otherwise random.sample
uses a different strategy
sage: tester = FF._tester(elements=FF, max_runs=5)
sage: list(tester.some_elements())
...
ValueError: first letter of variable name must be a letter
}}}
This ticket reduces the role of `InstanceTester.some_elements` to just do
some argument parsing and ensure that at most "max_run" elements are
returned. It only requires the input to be iterable.
If we want to have fancy random samples, we should define a specific
protocol (typically P.sample()) for it, or just let parents decide on the
appropriate strategy by defining some_elements appropriately.
This was originaly analyzed in #15919.
--
--
Ticket URL: <http://trac.sagemath.org/ticket/16244#comment:3>
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/d/optout.
