FR: make log graph play well with oneline templates

[avamsi]

With more or less the default template --

@  zn avamsi 15 seconds ago d8
│  (empty) (no description set)
o  r avamsi 15 seconds ago 8
│  (empty) test3
│ o  ur avamsi 34 seconds ago d1
├─╯  (empty) test2
o  uu avamsi 37 seconds ago e
│  (empty) test1
o  zz 53 years ago 00
   (empty) (no description set)

With a oneline template I'm experimenting with --

@  d8 | (empty) (no description set) | zn
o  8 | (empty) test3 | r
│ o  d1 | (empty) test2 | ur
├─╯
o  e | (empty) test1 | uu
o  00 | (empty) (no description set) | zz

1. Adjacent changes should be connected (maybe users could replace o with ├─o or something?).
2. Branch connector shouldn't waste a whole line (like with d1 in this example).

I imagine 2 might be a bit more involved given it might require changes in the graph rendering code to not use another line and not just the connector (maybe ├─o again instead of ├─╯).

[martinvonz]

Let's say the graph initially looked like this instead:

@  zn avamsi 15 seconds ago d8
│  (empty) (no description set)
│ o  ur avamsi 34 seconds ago d1
| |  (empty) test2
o | r avamsi 15 seconds ago 8
│ |  (empty) test3
├─╯
o  uu avamsi 37 seconds ago e
│  (empty) test1
o  zz 53 years ago 00
   (empty) (no description set)

Then we can't simply replace the o by ├─o for commit r because there's not enough room. We could possibly use just ├o. Either way, however, I think that makes it look like it's not an ancestor of commit zn. So unless we find a unicode character that's a vertical line with a circle (or other symbol) on it, I think it will be misleading.

I don't think I would mind your suggestion 2, however.

[avamsi]

Either way, however, I think that makes it look like it's not an ancestor of commit zn.

Good point, sorry I didn't think of that -- of course, we could keep indenting child commits (now that I think of it, I think that's what Fig tree used to do -- and probably still does but I wouldn't know 😅) but that's not ideal.

#242 (and probably #1036) are relevant for the room issue (as you already know, but linking for posterity).

[joyously]

It all looks like gobbledygook to me, especially since the o is part of the range for the reverse alphabet IDs. Is everything below @ an ancestor of it? Is there a flag to show @ last?

[martinvonz]

It all looks like gobbledygook to me, especially since the o is part of the range for the reverse alphabet IDs.

We could change to some other symbol. I sent https://github.com/martinvonz/jj/pull/1359 to change to ●. Please comment on that PR if you have a different preference.

Is everything below @ an ancestor of it? Is there a flag to show @ last?

Yes, --reversed.