call function in sage.libs.mpmath.utils doesn't handle parent=parent(float)

[benjaminfjones]

In multiple places in the Sage library the function sage.libs.mpmath.utils.call is called and inputs are passed along with their parents. If the parent of the input doesn't have a prec method, an AttributeError is raised:

sage: import sage.libs.mpmath.all as a
sage: z = float(4)
sage: a.call(a.e1, z)
0.00377935240984891
sage: a.call(a.e1, z, parent=parent(z))
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)

/Users/jonesbe/sage/sage-4.7.2.alpha2/devel/sage-test/sage/<ipython console> in <module>()

/Users/jonesbe/sage/latest/local/lib/python2.6/site-packages/sage/libs/mpmath/utils.so in sage.libs.mpmath.utils.call (sage/libs/mpmath/utils.c:5277)()

AttributeError: type object 'float' has no attribute 'prec'

This can be fixed with a simple check in in the call code in sage/libs/mpmath/utils.pyx.

---

Apply trac_11885_v2.patch to the Sage library.

[benjaminfjones]

adds check that parent has attribute 'prec' before calling

[burcin]

Thanks for the patch. It looks good, though I have minor suggestions:

• this needs to be doctested
• using

  try:
      prec = parent.prec()
  except AttributeError:
      pass
  

might be faster than performing a test with hasattr() first, then looking up the prec() method again.

[benjaminfjones]

adds an exception check in call()

[benjaminfjones]

Good point! I made the change and added a doctest to call() to illustrate that the bug is fixed.

The change does give a speedup. The best case I found was running

sage: timeit('mputils.call(a.ei, 1.0r, parent=float)')

Before the old patch, I was getting about 87 micro seconds. With the new patch, the best time I got was about 60 micro seconds.

[leif]

Replying to benjaminfjones:

Good point! I made the change and added a doctest to call() to illustrate that the bug is fixed.

The change does give a speedup. The best case I found was running

sage: timeit('mputils.call(a.ei, 1.0r, parent=float)')

Before the old patch, I was getting about 87 micro seconds. With the new patch, the best time I got was about 60 micro seconds.

I'm not sure whether that's a deterministic result.

But Burcin is right, using try: ... is (at least) faster in case parent does have a prec() method.

[burcin]

Thanks for the changes. Positive review.

[jdemeyer]

Milestone sage-4.7.3 deleted