Failing doctests in src/sage/modular/modform_hecketriangle

[GMS103]

Is there an existing issue for this?

• I have searched the existing issues for a bug report that matches the one I want to file, without success.

Did you read the documentation and troubleshoot guide?

• I have read the documentation and troubleshoot guide

Environment

- OS:  macOS (not relevant I think)
- Sage Version:  10.0.beta4

Steps To Reproduce

make ptestlong gives

sage -t --long --warn-long 56.6 --random-seed=452785398975339820822274837493975258 src/sage/modular/modform_hecketriangle/abstract_space.py  # 3 doctests failed
sage -t --long --warn-long 56.6 --random-seed=452785398975339820822274837493975258 src/sage/modular/modform_hecketriangle/readme.py  # 1 doctest failed
sage -t --long --warn-long 56.6 --random-seed=452785398975339820822274837493975258 src/sage/modular/modform_hecketriangle/space.py  # 2 doctests failed

Expected Behavior

All tests passed!

Actual Behavior

sage -t --long --warn-long 56.6 --random-seed=452785398975339820822274837493975258 src/sage/modular/modform_hecketriangle/abstract_space.py
**********************************************************************
File "src/sage/modular/modform_hecketriangle/abstract_space.py", line 1268, in sage.modular.modform_hecketriangle.abstract_space.FormsSpace_abstract.F_basis
Failed example:
    MF.F_basis(1)
Expected:
    q - 13071/(640000*d^2)*q^3 + O(q^4)
Got:
    q + (13071/(-640000*d^2))*q^3 + O(q^4)
**********************************************************************
File "src/sage/modular/modform_hecketriangle/abstract_space.py", line 1444, in sage.modular.modform_hecketriangle.abstract_space.FormsSpace_abstract.quasi_part_gens
Failed example:
    MF.quasi_part_gens(r=0)
Expected:
    [q - 34743/(640000*d^2)*q^3 + O(q^4), q^2 - 69/(200*d)*q^3 + O(q^4)]
Got:
    [q + (34743/(-640000*d^2))*q^3 + O(q^4), q^2 - 69/(200*d)*q^3 + O(q^4)]
**********************************************************************
File "src/sage/modular/modform_hecketriangle/abstract_space.py", line 1446, in sage.modular.modform_hecketriangle.abstract_space.FormsSpace_abstract.quasi_part_gens
Failed example:
    MF.quasi_part_gens(r=1)
Expected:
    [q - 9/(200*d)*q^2 + 37633/(640000*d^2)*q^3 + O(q^4),
     q^2 + 1/(200*d)*q^3 + O(q^4)]
Got:
    [q + (9/(-200*d))*q^2 + (-37633/(-640000*d^2))*q^3 + O(q^4),
     q^2 + 1/(200*d)*q^3 + O(q^4)]
**********************************************************************

sage -t --long --warn-long 56.6 --random-seed=452785398975339820822274837493975258 src/sage/modular/modform_hecketriangle/readme.py
**********************************************************************
File "src/sage/modular/modform_hecketriangle/readme.py", line 999, in sage.modular.modform_hecketriangle.readme
Failed example:
    MF.F_basis(1)
Expected:
    q - 13071/(640000*d^2)*q^3 + O(q^4)
Got:
    q + (13071/(-640000*d^2))*q^3 + O(q^4)
**********************************************************************

sage -t --long --warn-long 56.6 --random-seed=452785398975339820822274837493975258 src/sage/modular/modform_hecketriangle/space.py
**********************************************************************
File "src/sage/modular/modform_hecketriangle/space.py", line 399, in sage.modular.modform_hecketriangle.space.QuasiCuspForms.gens
Failed example:
    MF.gens()
Expected:
    [q - 17535/(262144*d^2)*q^3 + O(q^4),
     q^2 - 47/(128*d)*q^3 + O(q^4),
     q - 9/(128*d)*q^2 + 15633/(262144*d^2)*q^3 + O(q^4),
     q^2 - 7/(128*d)*q^3 + O(q^4),
     q - 23/(64*d)*q^2 - 3103/(262144*d^2)*q^3 + O(q^4),
     q - 3/(64*d)*q^2 - 4863/(262144*d^2)*q^3 + O(q^4),
     q - 27/(64*d)*q^2 + 17217/(262144*d^2)*q^3 + O(q^4)]
Got:
    [q + (17535/(-262144*d^2))*q^3 + O(q^4),
     q^2 - 47/(128*d)*q^3 + O(q^4),
     q + (9/(-128*d))*q^2 + (-15633/(-262144*d^2))*q^3 + O(q^4),
     q^2 - 7/(128*d)*q^3 + O(q^4),
     q - 23/(64*d)*q^2 - 3103/(262144*d^2)*q^3 + O(q^4),
     q - 3/(64*d)*q^2 - 4863/(262144*d^2)*q^3 + O(q^4),
     q - 27/(64*d)*q^2 + 17217/(262144*d^2)*q^3 + O(q^4)]
**********************************************************************
File "src/sage/modular/modform_hecketriangle/space.py", line 817, in sage.modular.modform_hecketriangle.space.CuspForms.gens
Failed example:
    MF.gens()
Expected:
    [q + 296888795/(10319560704*d^3)*q^4 + O(q^5),
     q^2 + 6629/(221184*d^2)*q^4 + O(q^5),
     q^3 - 25/(96*d)*q^4 + O(q^5)]
Got:
    [q + 296888795/(10319560704*d^3)*q^4 + O(q^5),
     q^2 + (-6629/(-221184*d^2))*q^4 + O(q^5),
     q^3 - 25/(96*d)*q^4 + O(q^5)]
**********************************************************************

Additional Information

Why plus/minus signs are not simplified in fractions?

[jhpalmieri]

I see this with 10.0.beta4 but not 10.0.beta2 (both built today on the same machine), so it's not caused by recent homebrew upgrades, for example. (Unless I'm confused, which is certainly possible.)

[jhpalmieri]

Perhaps caused by #35045.

[jhpalmieri]

I tried with a linux virtual machine and doctests passed, so somehow it seems to be platform-dependent.

[jhpalmieri]

Marking as a blocker since we get failing doctests.

[mezzarobba]

(Replying to #35045 (comment):)

I don't see how something like that can be platform-dependent. Have you compared the parents of the output on Linux and OSX, or checked if manually created expressions in the same parent failed to simplify too?

[jhpalmieri]

I don't know the mathematics, so I don't know what to try. If I do a = MF.F_basis(1), then parent(a) and type(a) are the same on OS X and Linux, but their print representations are different. If you have access to an OS X machine, you can see this with 10.0.beta4 or later. The doctest failures are not there in 10.0.beta3, but they are after applying https://patch-diff.githubusercontent.com/raw/sagemath/sage/pull/35045.diff.

[jhpalmieri]

Maybe there is some different system package, and this behavior is revealing that difference somehow.

[mezzarobba]

If I do a = MF.F_basis(1), then parent(a) and type(a) are the same on OS X and Linux, but their print representations are different.

What about the parent of a.q_expansion()? Is it Power Series Ring in q over Fraction Field of Univariate Polynomial Ring in d over Integer Ring as on Linux, or something different (e.g., ... over RationalField)? How do the print representations of manually created elements of that parent look like? of its base ring?

[jhpalmieri]

For both platforms, I see the same behavior:

sage: R = parent(a.q_expansion())
sage: R
sage: R.base_ring().inject_variables()
Defining d
sage: R(1/(-d))
1/-d

sage: R.inject_variables()
Defining q
sage: q + (13071/(-640000*d^2))*q^3
q + (13071/(-640000*d^2))*q^3

At least on OS X I see the same behavior in vanilla 10.0.beta3 — without the branch at #35045. (I haven't tried on linux.)

[jhpalmieri]

The behavior looks very strange to me, even before #35045. With 10.0.beta3:

sage: from sage.modular.modform_hecketriangle.space import WeakModularForms, CuspForms
sage: MF = WeakModularForms(n=5, k=62/3, ep=-1)
sage: a = MF.F_basis(1)
sage: a
q - 13071/(640000*d^2)*q^3 - 26203/(1500000*d^3)*q^4 + O(q^5)
sage: R = parent(a.q_expansion())
sage: R.inject_variables()
Defining q
sage: R.base_ring().inject_variables()
Defining d

sage: q - 13071/(640000*d^2)*q^3 - 26203/(1500000*d^3)*q^4 + O(q^5)
q - 13071/(640000*d^2)*q^3 - 26203/(1500000*d^3)*q^4 + O(q^5)
sage: q + (13071/(-640000*d^2))*q^3 + (26203/(-1500000*d^3))*q^4 + O(q^5)
q + (13071/(-640000*d^2))*q^3 + (26203/(-1500000*d^3))*q^4 + O(q^5)

The last two expressions evaluate as equal, so why don't they print the same way? Maybe #35045 just changed how the expression is defined, from something along the lines of q - A/B to q + A/(-B), and this catches an apparently old behavior that the signs are not simplified to convert the second form to the first.

[jhpalmieri]

Maybe it's Singular. When I use ./configure --with-system-singular=no, tests pass. On OS X, Homebrew includes Singular 4.3.2, whereas Sage includes 4.3.1.

[mezzarobba]

The last two expressions evaluate as equal, so why don't they print the same way?

Sounds like #35221 (comment), but I don't understand why the issue occurs on some systems only then. (Edit: I had missed your comment about singular.)

Maybe #35045 just changed how the expression is defined

Yes, most probably.