Re: [PATCH] Add new git-graph command

On Sun, Mar 30, 2008 at 04:49:16PM -0700, Junio C Hamano wrote:

As I mentioned in my initial email, I tried using the graph in tig, but I found it very hard to read. However, going back and taking a closer look at tig, I think its graph is just plain wrong in some cases.

For example, here's a comparison of the first several lines of output of "git-graph --date-order --all" and "tig -- --all" in my repository for one of the projects I am working on. I've replaced the commit subjects with just the abbreviated hashes. (Hopefully your mail reader displays this with a monospace font.)

git-graph:

*   8076867
*   2613e2b
M    542f526
|\  
* |   642b381
| | *   e73dfa2
| \ \ 
|  \ \ 
M-. \ \   64e1d85
|\ \ \ \ 
| | * | |   836521f
| | | | | *   ce43181
| | | | | *   eaeeb08
| | | | | M    79d3db3
| | | | | |\  
| | | | | | *   da5bc9e
| * | | | | |   b947aab
| | | | | M  \   9ade1bc
| | | | | |\  | 
| | | | * | | |   8f3727b
| | | | * | | |   2d102cd
* | | | | | | |   bf5c6e3
| |/ / / / / / 
|/| / / / / / 
* | | | | | |   a570370
|/ / / / / / 
* | | | | |   dde9a00
| | | | | *   09048ce
| | | | |/ 
| | | | *   4ee2351

tig (version 0.10.git):

+ 8076867
* 2613e2b
M 542f526
* | 642b381
| | + e73dfa2
M |  \ \ 64e1d85
|`.`* | | 836521f
| | | | | + ce43181
| | | | | * eaeeb08
| | | | | M 79d3db3
| | | | | |`* da5bc9e
| * | | | | | b947aab
| | | | | M  \ 9ade1bc
| | | | * |`.`. 8f3727b
| | | | * | | | 2d102cd
* | | | | | | | bf5c6e3
*' / / / / / / a570370
*' / / / / / dde9a00
| | | | | * 09048ce
| | | | *' 4ee2351

Looking at the tig output, it seems like tig is definitely screwing up around the 3-way merge at 64e1d85. The way I read the output, it looks like it is telling me that commit 836521f is a parent of 542f526, which is incorrect.

It also doesn't seem to display things correctly when branch lines need to cross. For example, 836521f is a child of a570370, and b947aab is a child of dde9a00. However, the way I read tig's graph, it looks like it is telling me exactly the opposite--that 836521f is a child of dde9a00 and b947aab is a child of a570370.

I think part of the problem is that tig displays exactly 1 commit per line. It's hard to represent octopus merges and crossing branches if it all has to fit in 1 line. This is why git-graph sometimes takes extra lines in between commits to display where the branch lines are going.

The git-forest command that Matthieu Moy pointed out also seems to do a good job of representing the graph, while being a bit more compact than git-graph when displaying crossing branches. (I just joined the list in the last couple of days, so I wasn't aware of git-forest. I did try googling for similar tools before I started, but only came up with tig.)