[git-users] Re: Odd performance problem with git diff
Yay! These .svn folders are causing havoc to our workspace in general. Amazing how much they slow it down. Another good reason to switch to git :) -- You received this message because you are subscribed to the Google Groups Git for human beings group. To post to this group, send email to git-users@googlegroups.com. To unsubscribe from this group, send email to git-users+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/git-users?hl=en.
[git-users] Re: Odd performance problem with git diff
In the course of messing around with this issue, I discovered something interesting. There were a number of .svn (Subversion) directories in my repository because the git repository was created from directories checked out of subversion. Also, a number of those .svn directories contained binary files such as .jpg, .gif, etc. So that content was essentially doubled in the repository. I deleted the .git repository and rebuilt it, but this time with .svn in the .gitignore file. Lo and behold the original git diff command returned in less than a second on the new repository. So even though there weren't *that* many binary files in this repository, there were enough to badly slow down the diff command. All is well now. :-) On Mar 24, 10:30 am, Dave R dran...@yahoo-inc.com wrote: Thanks for your help on this. Maybe I'll look into using git log instead of git diff. I've never seen git whatchanged before so I'll check that out too. On Mar 24, 1:33 am, Thomas Ferris Nicolaisen tfn...@gmail.com wrote: Actually this was a much better guide to the git date formats: http://www.alexpeattie.com/blog/working-with-dates-in-git/ -- You received this message because you are subscribed to the Google Groups Git for human beings group. To post to this group, send email to git-users@googlegroups.com. To unsubscribe from this group, send email to git-users+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/git-users?hl=en.
[git-users] Re: Odd performance problem with git diff
Yes, git log http://www.kernel.org/pub/software/scm/git/docs/git-log.htmlcan take two times as arguments. git log --since two weeks ago --until yesterday You might also want to consider git whatchangedhttp://www.kernel.org/pub/software/scm/git/docs/git-whatchanged.html : git whatchanged --since two weeks ago --until yesterday The arguments accepted are pretty versatile. For details, see the Specifying Revisions section of git-rev-parsehttp://www.kernel.org/pub/software/scm/git/docs/v1.5.0.7/git-rev-parse.html . I don't have any more ideas on figuring out what the performance bottleneck could be though, sorry. Maybe someone on the main git listhttp://vger.kernel.org/vger-lists.html#gitcan help more? -- You received this message because you are subscribed to the Google Groups Git for human beings group. To post to this group, send email to git-users@googlegroups.com. To unsubscribe from this group, send email to git-users+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/git-users?hl=en.
[git-users] Re: Odd performance problem with git diff
Actually this was a much better guide to the git date formats: http://www.alexpeattie.com/blog/working-with-dates-in-git/ -- You received this message because you are subscribed to the Google Groups Git for human beings group. To post to this group, send email to git-users@googlegroups.com. To unsubscribe from this group, send email to git-users+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/git-users?hl=en.
[git-users] Re: Odd performance problem with git diff
Thanks for your help on this. Maybe I'll look into using git log instead of git diff. I've never seen git whatchanged before so I'll check that out too. On Mar 24, 1:33 am, Thomas Ferris Nicolaisen tfn...@gmail.com wrote: Actually this was a much better guide to the git date formats: http://www.alexpeattie.com/blog/working-with-dates-in-git/ -- You received this message because you are subscribed to the Google Groups Git for human beings group. To post to this group, send email to git-users@googlegroups.com. To unsubscribe from this group, send email to git-users+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/git-users?hl=en.
[git-users] Re: Odd performance problem with git diff
Wow, thanks for your detailed response, Thomas. I'll try to answer some of your questions: Both repositories are on a remote mounted filesystem. Several people have told me that defeats the purpose of using git, but I claim innocence by virtue of inheriting the system as it is. :-) Also, I doubt that's a factor anyway since the fast repository is perfectly fast on that same filesystem. That seems to rule out any filesystem related issues such as fragmentation too. The files in the slow repository are mostly text (.html .css etc.) files. There are .gif .png and other such files sprinkled in. By far the biggest directory in the repository has 32509 regular files with 2208 that don't have obvious text suffixes. A few of those 2208 are probably also text files, but with suffixes I don't recognize. So, the repository certainly isn't dominated by binary files. There are currently 26 directories in all. Most of them have a few hundred files in them. Maybe a little more background might help. Changes are made to a given directory (new files, updated files, deleted files) throughout the day. Every night a cron job runs which does a git add . and git commit in that directory. Later a PHP program runs via a browser that looks at the git repository and shows all the pertinent changes. Depending on the URL, the script can say, Give me the names of all files in the repository changed since the beginning of time, but it usually says, Give me the names of all files in the repository changed between two given timestamps. It's the give me all changed files since the beginning of time that is vexing me. It's super fast for the larger repository and slow for the other one. So the system is interested in things at the file level rather than the commit level since a mass commit is done once a day. I'm running git version 1.7.0.6 on an RHEL system. I've run git gc in the past and it has no effect. Is there a way to run git that might reveal where it's spending all it's time? git log is super speedy. Is there a good way to use that to get all the files changed between two times? -- You received this message because you are subscribed to the Google Groups Git for human beings group. To post to this group, send email to git-users@googlegroups.com. To unsubscribe from this group, send email to git-users+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/git-users?hl=en.
[git-users] Re: Odd performance problem with git diff
And, just to clarify, the following text: On Mar 17, 10:32 am, Dave R dran...@yahoo-inc.com wrote: Here is the command that's slow on repository #1, and fast on repository #2: is wrong. The command is fast on repository #1 and slow on repository #2. -- You received this message because you are subscribed to the Google Groups Git for human beings group. To post to this group, send email to git-users@googlegroups.com. To unsubscribe from this group, send email to git-users+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/git-users?hl=en.
[git-users] Re: Odd performance problem with git diff
Hi Dave, Just to clarify: The normal way to check for changes in a repository is usually this: cd /path/to/repository git status There are many factors that come into play with git performance, and although it is inherently fast compared to a remote repository system like Subversion, it involves a lot of clever techniques on the file-system, using a lot of indexing and compression. This means, when it slows down, it's kind of hard to trace what's wrong without being a bit of a git/filesystem/operating-system wizard. For example, it could be due to heavy fragmentation of the disk sectors where your slow repository is stored. It could also be that your filesystem is slowing down the procedure. Which filesystem are you running? What are your hard-drive specs? Operating system? Git version? It does sound odd that the larger repository is fast compared to the small one. Can you say something about the nature of the two repos? Is there a lot of binary (images) data in one of them? How many files are in there? How big are the .git folders compared to the rest? Sometimes it helps asking git to repackage its index/repository (garbage collection): git gc If you want additional ways of asking for the diff between two points in time in the repository, the general syntax is this: git diff commit1 commit2 where the commits can be either SHA's, or tags. If you leave out the first one like this: git diff commit .. the diff is between HEAD and commit You can also use *git log* for seeing the commit log between two points in time, as well as git show, for looking at single commits. For example, here's the log with diff (-p) for the last three commits: git log -p HEAD~3 And here, showing only the changed files: git logv -p HEAD~3 --name-status -- You received this message because you are subscribed to the Google Groups Git for human beings group. To post to this group, send email to git-users@googlegroups.com. To unsubscribe from this group, send email to git-users+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/git-users?hl=en.
