Visual introduction to Git internals

Introduces the way Git works internally in a visual manner.
The goal is to present the core tenets of Git and to give the reader a grasp of what it does under the hood, which I find essential to efficiently use it.

It was presented as a lightning talk within a previous company of mine where I instigated the move from SVN to Git, to try and ease the transition for developers without them / before they dig deep into Git logic.

This a translation from the original one at https://speakerdeck.com/desseim/shi-jue-de-nagitnei-bu-ru-men .

Transcript

1. Let's understand Git Bare repository !? squash !? rebase !? refs !? detached head WTF!? Fast forward!? HEAD !? Legrand Guillaume 2012/02/22
2. Let's try and understand Git Bare repository !? refs !? detached head WTF!? HEAD !? Legrand Guillaume 2012/02/22
3. $ Working directory src_dir src_file.pl print “txet”;
4. $ git init Working directory .git/ objects/ refs/ untracked src_dir src_file.pl print “txet”;
5. $ git add ./src_dir Working directory src_dir src_file.pl print “txet”; .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; staged
6. $ git commit -m “First commit” Working directory src_dir src_file.pl print “txet”; .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; 5fs342w 0f4de32 scr_file .pl c92a25a e6fae02 User.name 2012-02-22 First commit e6fae02 5fs342w scr_dir HEAD heads/master heads/master c92a25a
7. $ Working directory src_dir src_file.pl print “txet”; .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; 5fs342w c92a25a First commit e6fae02 HEAD heads/master heads/master c92a25a
8. $ vim ./src_dir/src_file.pl Working directory src_dir src_file.pl print “text”; .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit modified 5fs342w e6fae02 HEAD heads/master heads/master c92a25a
9. $ git add ./src_dir Working directory src_dir src_file.pl print “text”; .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit staged 8e27be7 print “text”; 5fs342w e6fae02 HEAD heads/master heads/master c92a25a
10. $ git reset HEAD ./src_dir #unstage Working directory src_dir src_file.pl print “text”; .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 modified HEAD heads/master heads/master c92a25a
11. $ git checkout -- ./src_dir Working directory .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 src_dir src_file.pl print “txet”; HEAD heads/master heads/master c92a25a
12. $ git cat-file -p 8e27be7 print “text”; $ Working directory .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 src_dir src_file.pl print “txet”; HEAD heads/master heads/master c92a25a
13. $ vim ./src_dir/src_file.pl Working directory src_dir src_file.pl print “text”; .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit modified 5fs342w e6fae02 8e27be7 print “text”; HEAD heads/master heads/master c92a25a
14. $ git add ./src_dir Working directory src_dir src_file.pl print “text”; .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit staged 8e27be7 print “text”; 5fs342w e6fae02 HEAD heads/master heads/master c92a25a
15. $ git commit -m “2nd commit” Working directory .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 136b9d6 2nd commit parent 8e27be7 699022a src_dir src_file.pl print “text”; HEAD heads/master heads/master 136b9d6
16. $ git branch first c92a25a Working directory .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 136b9d6 2nd commit parent 8e27be7 699022a src_dir src_file.pl print “text”; HEAD heads/master heads/master 136b9d6 heads/first c92a25a
17. $ git checkout first Working directory .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 136b9d6 2nd commit parent 8e27be7 699022a src_dir src_file.pl print “txet”; HEAD heads/first heads/master 136b9d6 heads/first c92a25a
18. $ git branch -D master Working directory .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 136b9d6 2nd commit parent 8e27be7 699022a src_dir src_file.pl print “txet”; HEAD heads/first heads/first c92a25a
19. $ git checkout 136b9d6 Working directory .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 136b9d6 2nd commit parent 8e27be7 699022a src_dir src_file.pl print “text”; HEAD 136b9d6 heads/first c92a25a detached head
20. $ git fetch origin #assuming it had been pushed to origin previously Working directory .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 136b9d6 2nd commit parent 8e27be7 699022a src_dir src_file.pl print “text”; heads/first c92a25a remotes/origin /master 136b9d6 HEAD 136b9d6 detached head
21. $ git checkout master Working directory .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 136b9d6 2nd commit parent 8e27be7 699022a src_dir src_file.pl print “text”; heads/first c92a25a remotes/origin /master 136b9d6 HEAD Heads/master heads/master 136b9d6
22. $ echo 'c92a25a' > ./.git/refs/heads/master #don't actually do this! @see git-update-ref Working directory .git/ objects/ refs/ Blobs Trees Commits 0f4de32 print “txet”; c92a25a First commit 8e27be7 print “text”; 5fs342w e6fae02 136b9d6 2nd commit parent 8e27be7 699022a src_dir src_file.pl print “text”; heads/first c92a25a remotes/origin /master 136b9d6 HEAD Heads/master heads/master c92a25a
23. Summary • Git repo == .git/ directory – Can be backed up with a cp -R – If you have this directory, you can go back to any previous state • Git doesn't loose any data – It's COW: update implies duplication – Whatever mistake you make, if you had at least git added your work, it's not lost – It does not store diffs (opposed to e.g. subversion) • Everything is compressed with zlib
24. Summary • Analogous (esp. internally) to a file system with file history management (VFS) • Content ⇔ SHA-1 hash ⇔ history – E.g. if one does a rebase: the commit's parent changes ⇒ the commit's content changes ⇒ the commit's hash changes ⇒ the parent of the commit's children has changed ⇒ … ＝＝ history changes – SHA-1 == unique ID • Git is perfectly suitable as a backup tool – Even not just source code