Git Graphs, Hashes, and Compression, Oh My

Transcript

Git Graphs, Hashes, and Compression, Oh

[email protected] github.com/training @matthewmccull matthewmccullough

None

GIT - the stupid content tracker "git" can mean anything,

depending on your mood. * random three-letter combination that is pronounceable, and not actually used by any common UNIX command. The fact that it is a mispronunciation of "get" may or may not be relevant. * stupid. contemptible and despicable. simple. Take your pick from the dictionary of slang. * "global information tracker": you're in a good mood, and it actually works for you. Angels sing, and a light suddenly fills the room. * "goddamn idiotic truckload of sh*t": when it breaks Git is a fast, scalable, distributed revision control system with an unusually rich command set that provides both high-level operations and full access to internals. Git is an Open Source project covered by the GNU General Public License. It was originally written by Linus Torvalds with help of a group of hackers around the net. It is currently maintained by Junio C Hamano. “ ”

2005 git

None

Architecture Hashes

centralized VCSs use sequential revision numbers

Git uses a SHA-1 hash

None

40 hex characters (20 bytes)

9AB223D28B1AA46EF1780B22F304982E39872C34

9AB223D28B1AA46EF1780B22F304982E39872C34 <html> <body> <p>This is a test</p> <img src="http://ai.com/icon.gif"> </body>

</html> 9AB223D28B1AA46EF1780B22F304982E39872C34

Hashed Content Git commits without Git

# Green field project $ git init newproject $ cd

newproject # ...start coding

.git !"" COMMIT_EDITMSG !"" HEAD !"" MERGE_RR !"" config !""

description !"" hooks # !"" pre-commit.sample # $"" update.sample !"" index !"" info # $"" exclude !"" logs # !"" HEAD # $"" refs # $"" heads # $"" master !"" objects # !"" 54 # # $"" 3b9bebdc6bd5c4b22136034a95dd097a57d3dd # !"" info # $"" pack !"" refs !"" heads # $"" master $"" tags

$ printf "blob 12\000Hello World\n" | shasum

Architecture Hash Shortcuts

use as little of it as is unique

9AB223D28B1AA46EF 1780B22F304982E39 872C34 a certain commit

9AB22F a certain commit

git rev-parse 9AB22F expand the commit ref

Architecture Storage

Typical SCMs use delta storage

CVS / Subversion / darcs / Mercurial

v1 v2 v3 v4 File A File B File C

File A File B File B File C v5 File A File B File B Δ Δ Δ Δ Δ Δ Δ

Checkin Checkin Checkin Checkin Checkin Checkin Checkin Checkin Checkin Checkin

Checkin Delta storage gets slower as the history of a ﬁle gets longer

Directed Acyclic Graph

None

Copy of the entire tree per checkin

cp -r srcfolder srcfolder.prev

Why?

v1 v2 v3 v4 File A File B File C

File A File B File B File C v5 File A File B File B File A File A File C File C File C

hard link to existing identical blobs

v1 v2 v3 v4 File A File B File C

File A File B File B File C v5 File A File B File B File A File A File C File C File C ß

zlib deﬂates each blob at commit

v1 v2 v3 v4 File A File B File C

zlib deﬂates the entire repo

v1 v2 v3 v4 File A File B File C

2100 MB became 205 MB Act I

Architecture Hash relationships

‣Blob ‣Tree ‣Commit ‣Tag

tree tree: 7e8b1 web blob: 9ab16 index.html a10b3 tree blob:

8d162 logo.jpg blob: 51d22 draw.js 7e8b1 commit tree: a10b3 parent: nil author: Fird committer: Matthew message: Major refactoring of the web content. c67db blob <html> <body></body> </html> 9ab16 blob //Some more javascript var renderSize 51d22 blob 7D 8D B3 7F BD 12 9F E9 7B 78 9D 3F 5C A6 72 CB 8d162

tree tree: 7e8b1 web blob: 9ab16 index.html a10b3 tree blob:

v1 v2 v3 commit tree: 9a87b parent: nil author: Fird

committer: Matthew message: Major refactoring of the Javascript rendering engine. c67db commit tree: b22c1 parent: c67db author: Tim committer: Fird message: Minor update to HTML 9bd21 commit tree: b22c1 parent: 9bd21 author: Johnny committer: Joe message: New language transations 1c2d7