On Mon, 20 Jul 2015 06:50:02 -0700 (PDT)
Dimitris Papageorgiou <[email protected]> wrote:
> I have git init a working directory...here is
> it C:\Apache24\htdocs\Appointments\Administrator
>
> I now want to create git init another directory which sits *above
> *the aforementioned....C:\Apache24\htdocs\Appointments
>
> So....in essence....I want to disregard/delete the old
> branch/repo(C:\Apache24\htdocs\Appointments\Administrator)...*but
> keep the commits and transfer them to the new repo.*
>
> How am I going to do it? I hope I was clear.
Quite clear. The only problem is with the message subject which misses
the point completely: a branch is a line of development in a repository,
and your question is about creating a new repository.
As to the problem at hand, I'd go this route:
1) Use `git filter-branch` on all the relevant branches in your current
repository to modify all the commits in them to make them record
a prefix directory, "Administrator", in all their commits.
2) Move the Git database directory (".git") one level up -- into the
"Appointments" directory.
3) `git add` everything needed under this directory (except
"Administrator" -- as it will appear already tracked).
4) Record the new commit.
The end result will be the repository which contains all the initial
history plus a single commit of the "new" data.
Modifying the history in the original repository is a hard part.
Basically you'll need to filter all the relevant branches in your
source repository. Something like this:
$ cd theRepo
$ git branch tmp master
$ git filter-branch --tree-filter 'mkdir -p ./Administrator &&
find . -mindepth 1 -path ./Administrator -prune -o -print |
xargs mv -t ./Administrator' tmp
...now verify that the "tmp" branch looks OK, and replace "master"
with it.
Repeat for all the other branches.
(Note that this sinippet assumes a POSIX shell which means Git Bash in
your case.)
The problem here is that the history might have complicated graph of
intertwined histories, and that would make using `git filter-branch` an
excercise in patience and trial-and-error. In such a case you might
nominate just a single branch for conversion and sacrifice some bits of
"full" history.
If you're OK with some history traceability, there's another approach
which is known as "subtree merging":
1) Clone your source repository somewhere on the filesystem using
something like
git clone --bare src dst
2) Move the whole "Administrator" directory (including ".git" in it)
somewhere out of the tree.
3) Initialize a new repository under "Appointments".
`git add` everything needed except "Foo".
Record a commit.
4) Fetch the relevant branch from the repo created on step (1),
say
git fetch /c/path/to/dst master:dst-master
to create the local branch "dst-master" containing the commits
of source "master".
5) "Subtree-merge" that new branch under the prefix "Administrator":
git merge --no-commit dst-master
git read-tree -u --prefix=Administrator dst-master
git commit
By now, you have your original history merged, and it appears under
the directory "Administrator".
6) Verify the contents under "Administrator" is the same that in the
directory you've moved away on step (1) -- except for the ".git"
subdirectory, of course.
The problem with this approach is that if you will try to trace the
history of the "Administrator" subdirectory or any file under it --
using commands like
git log -- Administrator
git log -- Administrator/some/file
they will stop at the commit recorded on step (5) because the prefix
"Administrator" came into existence only then. You will still be able
to traverse there using more explicit means, of course. Read through
[2] for more info on these caveats.
1. http://bpeirce.me/moving-one-git-repository-into-another.html
2. https://groups.google.com/d/topic/git-users/HXoX-kpkYkM/discussion
--
You received this message because you are subscribed to the Google Groups "Git
for human beings" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
