Polyglot: Support Python 3 too

[ralsina]

Attempt to make the codebase py2/py3 compatible

• Support installing in py2 and py3
• Make tests run with py2 and py3 in travis
• Keep tests passing in py2

And then progressively fix tests in py3 until they all pass (or close enough)

[ralsina]

This breaks math (test_slides and test_issue_363)

[ralsina]

Woohoo, 144 failures.

[ralsina]

142!

[ralsina]

140!

[ralsina]

111!

[ralsina]

Extensions are broken in py3, which is unsurprising since they do manual, obscure import things, relevant tests are test_extensions and test_fancytitles

[ralsina]

110!

The tests are still failing ... hang on :)

[ralsina]

They are going to keep failing for a while, I am done for today ;-)

[ralsina]

109!

[ralsina]

105!

[ralsina]

101! (with svglib instead of svg2rlg)

[ralsina]

100!

[ralsina]

With python 3.7 literally 1 (one) ONE test passes, compared to a couple hundred with 3.6 so ... that needs looking at :-)

Looks like document frames are broken on python 3.7. Printing frames for a very basic document in python 2.7:

[[56.6929133858, 70.8661417323, 481.88976378, 700.157480315, 6, 6, 6, 6,]]

In 3.7:

[[2.0, 7.0, 100.0, 100.0, 6, 6, 6, 6,]]

I am guessing I broke something when removing the map() calls?

[ralsina]

Looks like a bug in adjustUnits

[ralsina]

In python 2.7:

>>> re.split(r'(-?[0-9.]*)', '2cm')
['', '2', 'cm']

In python 3.7:

>>> re.split(r'(-?[0-9.]*)', '2cm')
['', '2', '', '', 'c', '', 'm', '', '']

In python 3.6: same as python 2.7

[ralsina]

Back to 100 failures.

[ralsina]

95!

[ralsina]

I think most of the remaining failures are because I changed how the target reference ids are constructed, which should be easy to fix all in one swoop.

[ralsina]

I probably broke it in 42239d3

[ralsina]

The worst part is that links include object ID, which means they change on each run :-(

[ralsina]

55!

[ralsina]

And that is it for today.

[ralsina]

I think the next problem is in syntax highlighting.

[ralsina]

Interesting one! stylesheet paths are in different order in py2 and py3:

< [INFO] styles.py:113 StylePath:['/Users/ralsina/code/rst2pdf/rst2pdf/styles', '/Users/ralsina/.rst2pdf/styles', '.']
---
> [INFO] styles.py:113 StylePath:['/Users/ralsina/.rst2pdf/styles', '/Users/ralsina/code/rst2pdf/rst2pdf/styles', '.']

[ralsina]

Still 55, but hey, that was just wrong.

[ralsina]

The basic difference seems to be that in python2 RL is inserting colors (specifically 0 0 0 rg) that it doesn't in python3.

The outputs are identical in many tests except for that.

[ralsina]

44!

[akrabat]

Rebased against master after merging #755

[oz123]

This is most exciting.

[ralsina]

I am very happy. This branch was a lot of work :)