How to fork a file (git cp ?)

I need to add a new thing to our project. The thing will be similar to what already exists. I'd like to "derive" the new files from the existing ones -- without altering them and by preserving the change-history.

This is not a separate branch -- the "forked" files will co-exist. I'd call this git-cp (analogous to git-mv), but it does not exist...

Any ideas? Thanks!

    -mi

Am 04.05.2011 19:56, schrieb Mikhail T.:

(Sorry for the botched previous post - fat fingers)

Am 04.05.2011 19:56, schrieb Mikhail T.:

You cannot. Git does not have such a thing as "copy-with-preserved-history".

You just cp the file and git add it. But you will not be able to follow a history of the file.

-- Hannes

----- Original Message -----

Log (and other commands) can search for copies while traversing history:

  -C[<n>]
  --find-copies[=<n>]
  Detect copies as well as renames. See also --find-copies-harder. If n is specified, it has the same meaning as for -M<n>.

  --find-copies-harder
  For performance reasons, by default, -C option finds copies only if the original file of the copy was modified in the same changeset. This flag makes the command inspect unmodified files as candidates for the source of copy. This is a very expensive operation for large projects, so use it with caution. Giving more than one -C option has the same effect.

But as I discovered a few weeks ago, the existing merge strategies don't understand copies (recursive can follow a rename, but if two files pass rename detection, I think the one with the higher similarity index wins).

Thanks, Stephen

On 04.05.2011 14:22, Johannes Sixt wrote:

Thank you for the information...

So, is this something worth adding to the wishlist, or was it omitted on purpose (and which purpose was that, then)?

Yours,

    -mi

On 4 May 2011 21:17, Mikhail T. <mi+thun@aldan.algebra.com> wrote:

Oh yes, that was intentional. This is easily one of the most debated "features" of Git, especially in the early days of Git when almost all SCM systems did it "the CVS way", by tracking the history of single files. Instead, Git tracks snapshots of the whole tree and focuses on the whole content instead of single files. Renames are tracked by detecting removal/adding of files, which can be detected later, for example using "git log --follow". The reason for this is mostly speed issues, and most of the time the history of a single file is not interesting in a project, but changes in the file tree as a whole.

From the FAQ at <https://git.wiki.kernel.org/index.php/GitFaq>:

  Git has to interoperate with a lot of different workflows, for example
  some changes can come from patches, where rename information may not
  be available. Relying on explicit rename tracking makes it impossible
  to merge two trees that have done exactly the same thing, except one
  did it as a patch (create/delete) and one did it using some other
  heuristic.

  On a second note, tracking renames is really just a special case of
  tracking how content moves in the tree. In some cases, you may instead
  be interested in querying when a function was added or moved to a
  different file. By only relying on the ability to recreate this
  information when needed, Git aims to provide a more flexible way to
  track how your tree is changing.

  However, this does not mean that Git has no support for renames. The
  diff machinery in Git has support for automatically detecting renames,
  this is turned on by the '-M' switch to the git-diff-* family of
  commands. The rename detection machinery is used by git-log(1) and
  git-whatchanged(1), so for example, 'git log -M' will give the commit
  history with rename information. Git also supports a limited form of
  merging across renames. The two tools for assigning blame,
  git-blame(1) and git-annotate(1) both use the automatic rename
  detection code to track renames.

  As a very special case, 'git log' version 1.5.3 and later has
  '--follow' option that allows you to follow renames when given a
  single path.

  Git has a rename command git mv, but that is just for convenience. The
  effect is indistinguishable from removing the file and adding another
  with different name and the same content.

This mail from Linus explains the issue in more detail and colour: <http://permalink.gmane.org/gmane.comp.version-control.git/217>.

Regards, Øyvind

Johannes Sixt <j6t@kdbg.org> writes:

Well, if you come from the mindset that a "file" has an identity (hence there is a distinction between "This file used to be called A and at one point was renamed to B which is the name we see today" and "Some time ago somebody created a file B with the same contents as A and then removed A at the same time"), "copy" would not make much sense. What identity does a new file B gets when you create it by copying from A?

The same identity, or a different one? What happens when you later refactor the redundant part from these two files to create a common third file C? What identity does C have?

Correct. You cannot follow a history of _the file_, as there is no such thing.

You can still follow the history of contents, though. If you did a refactor like the one in the above example, "blame -L <range>" would follow the contents just fine. The command is a 80% satisfactory implementation of Linus's grand vision expressed in one of the most important message in the git mailing list archive:

    http://thread.gmane.org/gmane.comp.version-control.git/27/focus=217

On 04.05.2011 17:02, Junio C Hamano wrote:

What I want is to signify something like: "This code was obtained from that in file A."

"copy" -- of an individual file -- makes just as much sense as "move" (rename).

    -mi

"Mikhail T." <mi+thun@aldan.algebra.com> writes:

I think that is what exactly "blame -C -C" gives you.

On Thu, May 5, 2011 at 4:14 AM, Junio C Hamano <gitster@pobox.com> wrote:

Maybe Mikhail wanted to say that if there's a git-mv as a shortcut for
  "cp old new ; rm old; add new"
then there should be a git-cp as a shortcut for
  "cp old new; add new"

Just for convenience (and symmetry with git-mv).

He did write:

Piotr Krukowiecki <piotr.krukowiecki@gmail.com> writes:

Copying and then futzing with a copy is a bad discipline to begin with. git already has a reputation of having too many commands. I am not thrilled about the idea of making things worse by adding an unnecessary command, and especially one that encourages a bad workflow.

On Thu, May 5, 2011 at 8:50 PM, Junio C Hamano <gitster@pobox.com> wrote:

I'm not saying I want to have git-cp, but could you explain what is wrong with "cp old new; add new" (why is that bad workflow? How come adding a copy of a file can be a workflow at all?)

On 04.05.2011 22:14, Junio C Hamano wrote:

For that to be useful, one has to suspect, the file was derived by copying something else... Simple "git log" will not suggest that -- unless the commit message, that adds the new copy of a file points to it...

On 05.05.2011 14:02, Piotr Krukowiecki wrote:

git-mv preserves the old's change-history in new, so it is more than the above, is not it?

Yes...

    -mi

On Thu, May 05, 2011 at 03:31:22PM -0400, Mikhail T. wrote:

I think "git log --follow" will do what you want; it uses FIND_COPIES_HARDER.

No. The renames are detected at the time of viewing, not at the time of commit.

-Peff

On Thu, May 5, 2011 at 9:31 PM, Mikhail T. <mi+thun@aldan.algebra.com> wrote:

Maybe it should be the default (performance issues?)

It's the same IMO: log with --follow will follow both "copies" and "renames".

BTW, I don't understand why 'status' shows renames but not copies:

$ cp f fcp && git add fcp && git status # Changes to be committed: # new file: fcp

$ mv f fmv && git add fmv && git rm f && git status # Changes to be committed: # renamed: f -> fmv

I would expect sth like "copied: f -> fcp". Not sure what about one file copied to multiple files, but I suppose renames have the same problem.

It should not be a problem performance-wise...

On Thu, May 5, 2011 at 10:01 PM, Piotr Krukowiecki <piotr.krukowiecki@gmail.com> wrote:

I mean similar problem - you can delete one or more identical files and add one or more new files with the same content.

On Thu, May 05, 2011 at 10:01:32PM +0200, Piotr Krukowiecki wrote:

Performance is part of it, but also the fact that "--follow" has some limitations. For example, you can't use it with arbitrary pathspecs. I hope to fix that at some point. There was a GSoC proposal, but it didn't get selected; I'm hoping to work on it myself sometime this summer.

Yeah, we probably should do copy detection. Even weirder, we seem to do rename detection for what's to be committed, but not for what's in the worktree.

If you want to do a patch, the changes would go in wt-status.c, in the functions:

  wt_status_collect_changes_worktree
  wt_status_collect_changes_index

For people running "git status" manually, no. But something like FIND_COPIES_HARDER may be expensive on a big tree for people who use "git status" output as part of their shell prompt. So probably it should be configurable.

-Peff

On 5/5/2011 1:01 PM, Piotr Krukowiecki wrote:

...

Rename detection compares new files against deleted files. Copy detection compares new files against every file in the tree, which is usually much more costly.

--Pete

Pete Harlan <pgit@pcharlan.com> writes:

Almost correct. Copy detection usually compares new paths with paths modified in the same commit. "Every file in the tree" is done only when you ask to find copies harder.