Lazily import plot.

[robertwb]

Improve startup time.

CC: @kcrisman

Component: misc

Keywords: sd32

Reviewer: Dima Pasechnik

Issue created by migration from https://trac.sagemath.org/ticket/11043

[robertwb]

comment:1

Attachment: 11043-lazy-plot.patch.gz

Before

2.039 sage.all (None)
0.410 sage.misc.all (sage.all)
0.260 twisted.persisted.styles (sage.all)
0.221 sage.rings.all (sage.all)
0.188 sage.functions.all (sage.all)
0.171 functional (sage.misc.all)
0.164 piecewise (sage.functions.all)
0.162 sage.calculus.all (piecewise)
0.155 desolvers (sage.calculus.all)
0.154 sage.plot.all (desolvers)
0.151 zope.interface (twisted.persisted.styles)
0.137 pkg_resources (zope.interface)
0.132 interpreter (sage.misc.all)
0.131 IPython.ipapi (interpreter)

After

1.745 sage.all (None)
0.405 sage.misc.all (sage.all)
0.263 twisted.persisted.styles (sage.all)
0.180 sage.rings.all (sage.all)
0.168 functional (sage.misc.all)
0.154 zope.interface (twisted.persisted.styles)
0.141 pkg_resources (zope.interface)
0.134 sage.combinat.all (sage.all)
0.128 interpreter (sage.misc.all)
0.127 IPython.ipapi (interpreter)

That's another quarter-second (12.5%) savings.

[williamstein]

comment:2

Attachment: 11043-lazy-plot-rebase4.7.1.patch.gz

[robertwb]

comment:3

This is a more conservative patch which only saves importing the plot.* modules (about a 50-100ms savings on my computer).

[robertwb]

Attachment: 11043-lazy-plot-again.patch.gz

apply only this patch

[williamstein]

comment:5

Robert, did you even test this? It seems to fail doctests all over the place. E.g.,

sage -t  devel/sage/sage/calculus/riemann.pyx
**********************************************************************
File "/mnt/usb1/scratch/wstein/sage-4.7.1-sage.math.washington.edu-x86_64-Linux/devel/sage-main/sage/calculus/riemann.pyx", line 114:
    sage: m = Riemann_Map([f, hf], [hf, hfprime], 0.5 + 0.5*I)
Exception raised:
    Traceback (most recent call last):
      File "/mnt/usb1/scratch/wstein/sage-4.7.1-sage.math.washington.edu-x86_64-Linux/local/bin/ncadoctest.p
...
ImportError: No module named riemann

It seems like using LazyImport in any nontrivial way always results in serious breakage that is revealed by doctests. I hit this numerous times myself. Hmmm.

[williamstein]

Changed keywords from none to sd32

[sagetrac-drkirkby]

comment:7

Is anything happening on this? Robert wrote the patch, William says it breaks doctests all over the place, but then it just stalls.

Dave

[dimpase]

comment:13

outdated

[dimpase]

Reviewer: Dima Pasechnik