Pagination for Commits Page about gitlist

[klaussilveira]

I don't think git log allows me to give an offset along with it's limit parameter. Any ideas on how we can paginate without storing data in cache or a database? I would like to keep GitList simple and just parsing raw git output.

[kulbirsaini]

There exist options for git log which can help us achieve pagination. Look for the -number, --skip, --max-count and --reverse options in Commit Listing section of the man page.

[klaussilveira]

Thanks. That was useful. This command never crossed my mind! git log --skip=20 --max-count=10. 10 commits per page, starting at offset 20.

However, i have to make two git log runs: one to discover the total amount of commits available and another to get the actual commits from that page. Otherwise, the pager won't be very smart (the next button will be available to the infinite).

I'm thinking about leaving pagination to Javascript and load the entire thing there. The problem is that this would be an interface-oriented change and the load would still be high. Any opinions/suggestions? I can make the pagination very easily, but the next button will go on, and on, and on. Just not very intelligent.

[klaussilveira]

Please, reopen with all the issues the current pagination system has. This way the discussion will be more focused towards improving pagination, rather than having it.

[kulbirsaini]

I don't think that pushing all the commits to browser and then let javascript decide paging will be a good idea because you won't be able to get all the commits from git repo in the first place (for large repos like Linux). So, we have to figure out a way to count the total commits on the fly. There must be some hidden git feature to do this. Let's explore.

[klaussilveira]

Git exploring never stops to amuse me. GitList made me read so many man pages and i'm still finding new stuff...

I can make the git log output as slim as possible, but depending on repository size, it can be useless and will be a visible overhead.

[kulbirsaini]

To count all the commits in a repo, we can use git rev-list --all --count. This seems pretty fast. Take ~7-8 seconds to count commits for Linux on my VPS with 1GB RAM, Intel Xeon 2.33GHz. There are like 302939 commits in the repo I have.

[kulbirsaini]

Everything is awesome after your latest commit :-) Now, we need some way to cache the count. May be using session variable or something. If we can do that, we can finally browse Linux repo commits without pain :-)