Use the Sage displayhook in doctests

[vbraun]

The doctests output should look like the Sage commandline output, which it currently does not quite. Differences include:

• Dictionaries will be sorted by key, makes output more reproducable
• Sets are printed as {1,2} instead of set([1,2])
• Empty set is set() instead of set([])
• Types are <Foo at 0x...> instead of <Foo object at 0x...>

Other changes:

• Normal tracebacks can now be distinguished from tracebacks in __repr__ (IPython extension):

sage: f(*args)
<repr(<sage.interfaces.gap.GapElement at 0x...>) failed: ValueError: The session in which this object was defined is no longer running.>

• Plots now print themselves in doctest mode:

sage: A.plot()
Graphics object consisting of 1 graphics primitive

Depends on #16992

CC: @dkrenn

Component: doctest framework

Author: Volker Braun

Branch/Commit: 3070715

Reviewer: François Bissey

Issue created by migration from https://trac.sagemath.org/ticket/16746

[vbraun]

Branch: u/vbraun/sort_dicts_in_doctests

[vbraun]

New commits:

338c5c3

Use Sage display hook in doctests

[vbraun]

Commit: 338c5c3

[vbraun]

Author: Volker Braun

[vbraun]

comment:5

The first commit changes the doctest displayhook to the one we want. This causes various doctest changes that need to be fixed.

[vbraun]

comment:8

I noticed that the fix for printing types "our way" (#14466) was incomplete, inside lists it still uses IPython style:

sage: type(1)
<type 'sage.rings.integer.Integer'>
sage: [type(1)]
[sage.rings.integer.Integer]

[sagetrac-git]

Changed commit from 338c5c3 to d363558

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

8c5f2b0	Directly overwrite with fixed doctests
da59f20	call git directly and from the right directory
e397b3e	Add newline at end of file and fix cmdline help
9135b6c	Merge #16992 (Directly apply --fixdoctests) branch
89945c9	Fix printing of types
d363558	Refactor our displayhook IPython integration

[vbraun]

Dependencies: #16992

[vbraun]

comment:11

I've refactored our displayhook machinery so we can pick & choose which parts we want to take from IPython. Now works ok, still have to fixup old doctests.

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

1b99c68	Also fix the cmdline doctest
bdbdd5b	Merge #16992 (Directly apply --fixdoctests) branch
a246e17	Fix printing of types
eb6a66b	Refactor our displayhook IPython integration
b579b92	Update doctests that depended on the old displayhook

[sagetrac-git]

Changed commit from d363558 to b579b92

[kcrisman]

comment:15

Wow, that is a massive set of changes... just out of curiosity, what does the documentation now look like for graphics? I don't know whether it is more obvious before or after this change that one needs to evaluate to get them. One reason for the spaces before (I mean the ones after plot doctests) was to "leave room" in the (live) doc for evaluating these.

[jdemeyer]

comment:16

I see why you needed #16992 now...

[jdemeyer]

comment:17

In src/sage/graphs/generic_graph.py, this was probably intended as a doctest (although certainly something went wrong around those lines):

sage: g.layout(layout="acyclic_dummy", save_po=True)
{('0', 0): [0.3356712660780875, 0],
 ('0', 1): [0.33563875939602245, 1],
 ('1', 0): [0.6751346747788185, 0],
 ('1', 1): [0.6715689173983604, 1]}

[sagetrac-git]

Changed commit from b362aa6 to 7de159a

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

7de159a

Fix more random test failures

[cheuberg]

comment:29

Now, all doctests pass on my linux mint 64 bit desktop.

[vbraun]

comment:30

All doctest pass on the buildbot...

[dimpase]

comment:31

Did you try this on ARM?

[vbraun]

comment:32

ARM is part of the buildbot now, so yes ;-)

[kiwifb]

comment:33

Since it pass on the build bot and I don;t see any obvious problem with the new code itself I will give it a positive review. I expect ARM to be irrelevant in this context as we are only talking about re-formatting current doctests.
Plus it will fix broken doctests in sage-on-gentoo (dictionaries displayed in a different order).

[kiwifb]

Reviewer: François Bissey

[sagetrac-git]

Changed commit from 7de159a to f9af4ee

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1 and set ticket back to needs_review. New commits:

a25e286	trac #17013: new class WordDatatype_char
398b33c	trac #17013: added is_square in WordDatatype_char
a563343	trac #17013: doctests improvements + bug fix
c00964d	trac #17013: fix Python C API declaration
f9af4ee	Merge #17013 to resolve conflict

[vbraun]

comment:36

surprisingly only one merge conflict ;-)

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1 and set ticket back to needs_review. New commits:

3070715

fix another doctest

[sagetrac-git]

Changed commit from f9af4ee to 3070715

[kiwifb]

comment:38

At this stage I think you have done all the work that should be done in one ticket. I suggest we have a follow up ticket for any new discoveries.

[vbraun]

Changed branch from u/vbraun/sort_dicts_in_doctests to 3070715