Git-Fu

A few tricks about Git I'd like to remember. Write early, improve later!

1. From submodule to subdirectory

1.1. Context

One of my repositories contains many distinct subprojects that I did during a Java training. This arborescence of subprojects look like this:

20180312_project1
20180312_project2
.
.
.
20180328_project20
.
.
.
20180410_project36

…and so on. 20180328_project20 has its own repository (also named 20180328_project20 and is included as a submodule.

1.2. Goal

I want the 20180328_project20 submodule to become a subdirectory, and the submodule's history to be integrated to the main Git history. This integration should be chronological and the commit messages should be prefixed with the name of the project.

Original Git history:

[project36] Add Angular component
[project36] Create project
.
.
.
[project2] Improve logic
[project2] Create project
[project1] Add CSS
[project1] Create project

Submodule's Git history:

Add Logback to POM
Create project

Target Git history:

[project36] Add Angular component
[project36] Create project
.
.
.
[project20] Add Logback to POM
[project20] Create project
.
.
.
[project2] Improve logic
[project2] Create project
[project1] Add CSS
[project1] Create project

1.3. Steps

1.3.1. Remove all traces of the submodule

I'm used to do this manually, but there might be some better ways.

Edit the .gitmodules file to remove the submodule and stage the change.
```
git add .gitmodules
```
Remove the submodule files from version control.
```
git rm --cached 20180328_project20
```
Delete the files.
```
rm -rf 20180328_project20
```
Edit the .git/config file to remove the active submodule (only when the submodule was cloned into the main repo).
Remove the submodule from .git/modules/ directory.
```
rm -rf .git/modules/20180328_project20/
```

Commit the changes.

git commit -m "[project20] Remove submodule"

1.3.2. Include target repository as a branch in main repo

Add the distant repository to the main repository's remotes.

git remote add origin-project20 [url for 20180328_project20]

Fetch the target repo's content.
```
git fetch origin-project20
```
You should see the remote master branch when running git branch -a:
```
remotes/origin-project20/master
```
Checkout the target's repo master branch.
```
git checkout -b project20 origin-project20/master
```
git log will show you the history.

1.3.3. Rewrite target repo's history

Here we have two goals:

Put all the files (except .git) from the target repo in a subdirectory;
Prefix commit messages with the name of the project.

We also need to ensure that the commit date won't be changed by these two steps in order to keep the chrological order of commits in the final repo.

Move all files in a subdirectory.
```
git filter-branch --tree-filter "mkdir -p 20180328_project20; git mv -k * 20180328_project20" HEAD
```
(source)

Quote from the original article:

Dotfiles like .gitignore are not included, so you will need to go back to those commits when they were created and make sure they are created in the proper subdirectory, amending the commits using interactive rebase.

Use git-rebase for this.
Prefix all commits with the name of the project. At the moment I have no better solution than to reword everything with a interactive rebase.
```
git rebase -i --root
```
…then reword everything.
Change the commit date to make it equal to authoring date.
```
git filter-branch -f --env-filter 'export GIT_COMMITTER_DATE="$GIT_AUTHOR_DATE"'
```
git log --pretty=fuller should display the correct commit date.

1.3.4. Merge target branch

Now we can just merge that branch in the main repo.

git checkout master
git merge project20 --allow-unrelated-histories

1.3.5. Clean

git branch -D project20
git remote remove origin-project20

And push the result :)

1.4. Result

An arborescence with a 20180328_project20 subdirectory;
A git history containing the original target repo's history with a prefix (such as [project20] Create project), everything in chronological order.