# Why is reflog so obscure? 8 messages from 2022-07-11 to 2022-07-14. Participants: Gerriko io, Thomas Guyot, Jeff King, Junio C Hamano, Sergey Organov. Thread: https://gitlist.dev/t/58138 ## Gerriko io, 2022-07-11 21:02 Subject: Why is reflog so obscure? Message-ID: URL: https://gitlist.dev/e/CALF%3D2ANoq1eL-RqK_dLaVThoxbvfhzgPJXFHyD1kX8CFCPx3XA%40mail.gmail.com ``` I am trying to make sense of reflog as I need to know this to use revert and reset properly. I cannot find the most basic information in the git documentation or online. Surely the most basic question is whether the reflog is sequential or just random logs or whether it is in time ascending or descending order. Without this information I am left baffled and have to guess what this line means out of the many: a0bbd34 (HEAD -> master, origin/master, origin/main) refs/remotes/origin/main@{0}: update by push Why does it have to be so obscure? ``` ## Thomas Guyot, 2022-07-11 22:55 Subject: Re: Why is reflog so obscure? Message-ID: URL: https://gitlist.dev/e/f0bb8ee8-9b38-45a7-a54b-24cf245bb3c8%40gmail.com In-Reply-To: ``` On 2022-07-11 17:02, Gerriko io wrote: > I am trying to make sense of reflog as I need to know this to use > revert and reset properly. > > I cannot find the most basic information in the git documentation or online. > > Surely the most basic question is whether the reflog is sequential or > just random logs or whether it is in time ascending or descending > order. > > Without this information I am left baffled and have to guess what this > line means out of the many: > > a0bbd34 (HEAD -> master, origin/master, origin/main) > refs/remotes/origin/main@{0}: update by push > > Why does it have to be so obscure? Hi, The reflog is simply a log of where you've been, in descending order (latest entry first), and is branch agnostic. It show every commit you've been on from latest to oldest, and each entry can be heads, tags, or detached commits. Some operations like rebase may even generate multiple entries if it's done in multiple steps (ex conflicts resolution). I believe its main uses are to delay garbage collect (which happens only after entries have expired form the reflog) and to allow refs selection using @{date} and @{nth} specifiers (see git-rev-parse, and note @{date} is not the commit date but rather the reflog date!)... And it's obviously quite handy to recover a dropped/rewritten ref if you realize you need to undo what you've done. Since the date is recorded for each entry you can also display it using the --date= option. For example to display reflog entries in local human-readable time, use: git reflog --date=local The git-reflog log display options are mostly the same as git-log and documented in git-log's manpage. Regards, -- Thomas ``` ## Jeff King, 2022-07-12 07:12 Subject: Re: Why is reflog so obscure? Message-ID: URL: https://gitlist.dev/e/Ys0e9MxIWQj%2FpVXx%40coredump.intra.peff.net In-Reply-To: ``` On Mon, Jul 11, 2022 at 06:55:09PM -0400, Thomas Guyot wrote: > The reflog is simply a log of where you've been, in descending order (latest > entry first), and is branch agnostic. It show every commit you've been on > from latest to oldest, and each entry can be heads, tags, or detached > commits. Some operations like rebase may even generate multiple entries if > it's done in multiple steps (ex conflicts resolution). This is right with one caveat: the HEAD reflog is branch agnostic. But there is a separate reflog for each individual branch, as well. -Peff ``` ## Thomas Guyot, 2022-07-12 09:22 Subject: Re: Why is reflog so obscure? Message-ID: URL: https://gitlist.dev/e/a8d2a61d-b86f-9b89-6391-36c58c390a12%40gmail.com In-Reply-To: ``` On 2022-07-12 03:12, Jeff King wrote: > On Mon, Jul 11, 2022 at 06:55:09PM -0400, Thomas Guyot wrote: > >> The reflog is simply a log of where you've been, in descending order (latest >> entry first), and is branch agnostic. It show every commit you've been on >> from latest to oldest, and each entry can be heads, tags, or detached >> commits. Some operations like rebase may even generate multiple entries if >> it's done in multiple steps (ex conflicts resolution). > This is right with one caveat: the HEAD reflog is branch agnostic. But > there is a separate reflog for each individual branch, as well. Hi Jeff, Thanks for clarifying that - I suspected it since we can do @{} although I didn't find any reference branch reflogs in the documentation. I could've missed it... Is there a way to read a branch reflog? -- Thomas ``` ## Junio C Hamano, 2022-07-12 14:23 Subject: Re: Why is reflog so obscure? Message-ID: URL: https://gitlist.dev/e/xmqqo7xuif46.fsf%40gitster.g In-Reply-To: ``` Thomas Guyot writes: > Thanks for clarifying that - I suspected it since we can do > @{} although I didn't find any reference branch reflogs > in the documentation. I could've missed it... Is there a way to read a > branch reflog? $ git reflog ;# lists entries of reflog of HEAD, starting at HEAD@{0} $ git reflog HEAD ;# same $ git reflog HEAD@{4} ;# same, starting at HEAD@{4} $ git reflog master ;# entries of reflog of "master" $ git reflog master@{0} ;# same $ git reflog master@{now} ;# same, show with timestamps $ git reflog master@{4.minutes} ;# same, starting at master@{4.minutes} For the branch that is currently checked out, you can omit the name when you use any of the @{...} notation, so $ git reflog @{0} $ git reflog @{now} are often the easiest ways to view what you did on the current branch. ``` ## Sergey Organov, 2022-07-12 15:18 Subject: Re: Why is reflog so obscure? Message-ID: <87ilo2we9o.fsf@osv.gnss.ru> URL: https://gitlist.dev/e/87ilo2we9o.fsf%40osv.gnss.ru In-Reply-To: ``` Junio C Hamano writes: > Thomas Guyot writes: > >> Thanks for clarifying that - I suspected it since we can do >> @{} although I didn't find any reference branch reflogs >> in the documentation. I could've missed it... Is there a way to read a >> branch reflog? > > $ git reflog ;# lists entries of reflog of HEAD, starting at HEAD@{0} > $ git reflog HEAD ;# same > $ git reflog HEAD@{4} ;# same, starting at HEAD@{4} > $ git reflog master ;# entries of reflog of "master" > $ git reflog master@{0} ;# same > $ git reflog master@{now} ;# same, show with timestamps > $ git reflog master@{4.minutes} ;# same, starting at master@{4.minutes} > > For the branch that is currently checked out, you can omit the name > when you use any of the @{...} notation, so > > $ git reflog @{0} > $ git reflog @{now} > > are often the easiest ways to view what you did on the current > branch. Very handy, thanks! Would be nice to have this in EXAMPLES section of corresponding manual page. Honestly, I tried (admittedly not very hard) to figure how to get dates from "git reflog" a few times, but every time I gave up, so this (along with the --dates option turning dates output on) is not very discoverable. -- Sergey Organov ``` ## Junio C Hamano, 2022-07-12 15:50 Subject: Re: Why is reflog so obscure? Message-ID: URL: https://gitlist.dev/e/xmqq4jzmib3n.fsf%40gitster.g In-Reply-To: <87ilo2we9o.fsf@osv.gnss.ru> ``` Sergey Organov writes: > Very handy, thanks! Would be nice to have this in EXAMPLES section of > corresponding manual page. > > Honestly, I tried (admittedly not very hard) to figure how to get dates > from "git reflog" a few times, but every time I gave up, so this (along > with the --dates option turning dates output on) is not very > discoverable. Heh, I wouldn't have bothered writing the message you are responding to if it were documented already ;-) Patches welcome. ``` ## Thomas Guyot, 2022-07-14 00:03 Subject: Re: Why is reflog so obscure? Message-ID: URL: https://gitlist.dev/e/c12c14a9-9948-a47c-35a0-869f95d95bb9%40gmail.com In-Reply-To: ``` On 2022-07-12 11:50, Junio C Hamano wrote: > Sergey Organov writes: > >> Very handy, thanks! Would be nice to have this in EXAMPLES section of >> corresponding manual page. >> >> Honestly, I tried (admittedly not very hard) to figure how to get dates >> from "git reflog" a few times, but every time I gave up, so this (along >> with the --dates option turning dates output on) is not very >> discoverable. > Heh, I wouldn't have bothered writing the message you are responding > to if it were documented already ;-) > > Patches welcome. Indeed these are great! I was thinking the same thing, I can work on a patch shortly. Regards, -- Thomas ```