Restore the "backward" option for graph dot2tex

[tscrim]

The change made on #31381 is not compatible with the previous behavior of backward=True for the edge options with dot2tex rendering of graphs. In particular, this was seen when viewing KR crystals such as

view(crystals.KirillovReshetikhin(['C',3,1], 3,1))

We will add back in the backward option, which supports different behavior than dir='back':

• backward=True will be a layout-only indication that a particular arrow should be oriented in the opposite direction compared to other arrows.
• dir='back' will have the arrows pointing in the backwards direction.

This also reverts the change in #31719 to get the previous behavior.

CC: @sagetrac-sage-combinat @seblabbe @fchapoton @anneschilling @dwbump @nthiery @bsalisbury1

Component: graphics

Keywords: dot2tex

Author: Travis Scrimshaw

Branch/Commit: 7e2b1de

Reviewer: Sébastien Labbé

Issue created by migration from https://trac.sagemath.org/ticket/32438

[tscrim]

Commit: 1da92af

[tscrim]

New commits:

1da92af

Reverting to "backward" option for dot2tex graph layout.

[tscrim]

Branch: public/graphs/backwards_option-32438

[tscrim]

comment:2

Green bot (essentially).

[fchapoton]

comment:3

I would prefer that Sébastien gives a positive review, if possible.

[seblabbe]

comment:4

Yes, please, let me take a look. I was overloaded by the "rentrée". Next days will be better, I check this tomorrow on Thursday.

[tscrim]

comment:5

No problem. There isn't a big rush on this. I would appreciate it if this got into 9.5. :)

[seblabbe]

comment:6

I still haven't had the opportunity the try the branch. My last update of sage broke and I need to take a look. But looking at the proposed branch, I have one comment:

the change

+        - ``"backward"`` (boolean)

needs documentation. Maybe for an English speaker it is clear, but to me, it is difficult to understand the difference between dir='back' and backward=True. That difference needs to be explained in words here. Secondly, it should explicit what it does (that is, u,v=v,u and etc.). Also, it should be explained that it is something that will work only when prog='dot' and not for the other Graphviz programs like neato, fdp, circo, etc.

This is why I still think it is kind of a hack but I don't disagree to put it back.

[sagetrac-git]

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

13a57ed	Merge branch 'public/graphs/backwards_option-32438' of git://trac.sagemath.org/sage into public/graphs/backwards_option-32438
0c38588	Adding more description to the backward=True example.

[sagetrac-git]

Changed commit from 1da92af to 0c38588

[tscrim]

comment:8

I added a bit more to the example to indicate more clearly what changes, which is in line with the rest of the documentation for that section. Everything within that the code does works for all such programs (although it may or may not have an impact, but that is irrelevant).

It is definitely not a hack as the layout order matters (for at least one program), and we have mathematical information attached to the digraph that would be wrong if we change the arrows.

[sagetrac-git]

Changed commit from 0c38588 to 7e2b1de

[sagetrac-git]

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

7e2b1de

32438: added documentation for option 'backward'

[seblabbe]

comment:11

I added some documentation to the 'backward' option. I give a positive review to the commits made before mine. I let Travis change the status to positive review if he agrees with my commit.

[seblabbe]

Reviewer: Sébastien Labbé

[tscrim]

comment:13

Let's get this in.

[vbraun]

Changed branch from public/graphs/backwards_option-32438 to 7e2b1de