Re: [PATCH 1/2] revision: Denote root commits with '#'

[Kyle Marek <kmarek@pdinc.us>]

On 1/18/21 3:33 PM, Junio C Hamano wrote:
> Junio C Hamano <gitster@pobox.com> writes:
>
>> [Footnote]
>>
>> *1* Stepping back a bit, I think concentrating too much on "is it
>>      root?" is a wrong way to think about the problem.  Suppose you
>>      have two histories, e.g. (time flows from left to right; A and X
>>      are roots)
> A shorter and more concrete example.  Start from an empty repository:
>
> 	$ git init
> 	$ git commit --allow-empty -m Aroot
> 	$ git checkout --orphan side
> 	$ git commit --allow-empty -m Xroot
> 	$ git log --all --graph --oneline
>          * a1f7cb2 (HEAD -> side) Xroot
>          * b6fb655 (master) Aroot
>
> These depict two root commits, Aroot and Xroot, and no other
> commits.  We do want to show that these two commits do not have
> parent-child relationship at all, and your (and a few proposals made
> by other in the past) solution was to show them both with "#".
>
> Continuing in the same repository:
>
> 	$ git checkout --orphan another
> 	$ git commit --allow-empty -m Oroot
> 	$ git commit --allow-empty -m A
> 	$ git log --graph --oneline ^another^ another side
>          * eddf116 (HEAD -> another) A
>          * a1f7cb2 (side) Xroot
>
> These depict two commits, A and Xroot, and no other commits.  We
> also want to show that these two commits do not have parent-child
> relationship at all, but if we paint Xroot with "#", it still makes
> it appear that A is a child of Xroot.
>
>>      And the right way to look at it is "does A have any parent in
>>      the part of the history being shown?", not "does A have any
>>      parent?"  Then 'A' will get exactly the same treatment in the
>>      two examples, and the visual problem that makes A appear as if
>>      it has parent-child relationship with unrelated commit X goes
>>      away.
> So the condition we saw in your patches, !commit->parents, which
> attempted to see if it was root, needs to be replaced with a helper
> function that checks if there is any parent that is shown in the
> output.  Perhaps
>
> 	int no_interesting_parents(struct commit *commit)
> 	{
> 		struct commit_list *parents = commit->parents;
>
> 		while (parents) {
> 			if (!(parents->object.flags & UNINTERESTING))
> 				return 0;
> 			parents = parents->next;
> 		}
> 		return 1;
> 	}
>
> or something like that should serve as a replacement, i.e.
>
> 	return !commit->parents ? "#" : "*";
>
> would become
>
> 	return no_interesting_parents(commit) ? "#" : "*";
>
> Hmm?

Okay, I see what you mean. Fixing --graph to avoid implying ancestry 
sounds like a better approach to me.

That being said, I spoke to Jason recently, and he expressed interest in 
optionally marking root commits so they are easy to search for in a 
graph with something like /# in `less`. I see value in this, too.

So would you be open to my modifying of the patch in question (patch 1+2 
squashed, I guess) to instead use "--mark-roots=<mark>" to optionally 
mark root commits with a string <mark>, and pursue fixing the --graph 
rendering issue in another series?

If so, what would you like to see out of the --left-right issue? Maybe 
"--mark-left-root=<mark>" and "--mark-right-root=<mark>", so multi-byte 
strings may be used? Can there be more than one root on either side? (so 
the option would use a plural "roots" instead of "root"?)

-- 

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
-                                                               -
- Kyle Marek                        PD Inc. http://www.pdinc.us -
- Jr. Developer                     10 West 24th Street #100    -
- +1 (443) 269-1555 x361            Baltimore, Maryland 21218   -
-                                                               -
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-