Running doctests from within Sage doesn't work

[roed314]

Running doctests from within Sage results in any test that has output failing. For example:

sage: run_doctests(sage.algebras.algebra)
Doctesting /Users/roed/sage/sage-5.9.beta0/devel/sage/sage/algebras/algebra.py
Running doctests with ID 2013-03-17-01-02-17-09f21e8e.
Doctesting 1 file.
sage -t /Users/roed/sage/sage-5.9.beta0/devel/sage/sage/algebras/algebra.py
**********************************************************************
File "/Users/roed/sage/sage-5.9.beta0/devel/sage/sage/algebras/algebra.py", line 29, in sage.algebras.algebra.is_Algebra
Failed example:
    is_Algebra(R)
Expected:
    True
Got:
    True
**********************************************************************
File "/Users/roed/sage/sage-5.9.beta0/devel/sage/sage/algebras/algebra.py", line 31, in sage.algebras.algebra.is_Algebra
Failed example:
    sig_on_count()
Expected:
    0
Got:
    0
**********************************************************************
1 item had failures:
   2 of   4 in sage.algebras.algebra.is_Algebra
    [3 tests, 2 failures, 0.1 s]
----------------------------------------------------------------------
sage -t /Users/roed/sage/sage-5.9.beta0/devel/sage/sage/algebras/algebra.py  # 2 doctests failed
----------------------------------------------------------------------
Total time for all tests: 0.1 seconds
    cpu time: 0.1 seconds
    cumulative wall time: 0.1 seconds

This is the result of IPython coloring the output.

Apply

• attachment: 14290.patch

Depends on #12415
Depends on #14331
Depends on #13278

Component: doctest framework

Author: David Roe

Reviewer: Volker Braun

Merged: sage-5.9.beta2

Issue created by migration from https://trac.sagemath.org/ticket/14290

[roed314]

Attachment: 14290.patch.gz

[roed314]

comment:1

I couldn't figure out a way to doctest this fix (since the problem only occurred when running doctests from IPython), but it's ready for review.

[roed314]

Author: David Roe

[roed314]

Dependencies: #12415

[vbraun]

Reviewer: Volker Braun

[vbraun]

comment:3

Positive review to your patch.

I still think that it would be nice to have a note added to the diff that there were ansi escape sequences, since they don't get printed and you are otherwise just left with two identical strings which makes it non-obvious why the doctest failed. So I rescued that piece of the patch from #14359.

[roed314]

comment:4

Your patch looks good too.

[jdemeyer]

comment:5

This should be rebased to #14331 + #13278.

[jdemeyer]

Changed dependencies from #12415 to #12415, #14331, #13278

[vbraun]

comment:6

Patch applies cleanly here on top of sage-5.9.beta1 + #14331 + #13278.

[jdemeyer]

comment:7

Agreed, I must have done something wrong.

[jdemeyer]

comment:8

Concerning escapes: on OS X 10.8 I am seeing

\e[?1034h

(where \e stands for the escape character).
Could you also check for this code (no idea what it does).

[vbraun]

comment:9

That is not a color code but smm = "turn meta on". Most likely coming from readline if you have the wrong TERM set.

I'll update the patch to generate better diagnostic for escape sequences.

[vbraun]

Updated patch

[vbraun]

comment:10

Attachment: trac_14290_ansi_escapes_indication.patch.gz

The updated patch now makes all escape sequences visible (and doctestable):

EXAMPLES::

    sage: print 'This ist \x1b[1mbold\x1b[0m text'
    This ist <CSI-1m>bold<CSI-0m> text

[jdemeyer]

comment:11

What does CSI mean? I prefer to use ESC always.

Also, are you sure want to mess with \x9b? I think that won't go well with UTF-8.

[jdemeyer]

Merged: sage-5.9.beta2

[jdemeyer]

comment:12

For the escape codes, see #14375.