git.md

Version Control (Git)

Git’s Data Model

• Commits act as “snapshots” of a collection of folders and files
• Example

  <root> (tree)
  |
  +- foo (tree)
  |  |
  |  + bar.txt (blob, contents = "hello world")
  |
  +- baz.txt (blob, contents = "git is wonderful")
  

  • Git calls what we call folders trees and files as blobs
• Git history modeling, in the form of pseudocode

  // a file is a bunch of bytes
  type blob = array<byte>
  
  // a directory contains named files and directories
  type tree = map<string, tree | blob>
  
  // a commit has parents, metadata, and the top-level tree
  type commit = struct {
      parents: array<commit>
      author: string
      message: string
      snapshot: tree
  }
  

• Object is a blob, tree, or commit, all of which are stored and “addressed” using SHA-1 hash, aka a 40 character hexadecimal
  • git cat-file -p SHA-1_HASH will show contents of that object
• References are mutable (changeable) pointers to specific commits
  • master: most recent commit in main development branch
  • HEAD: where we currently are, so a new commit is generated with correct parents struct member
• Remotes have their own references too, and are maintained separately
  • To correspond local branches to remote branches: git branch --set-upstream-to=REMOTE/REMOTE_BRANCH

Configuration

• Configure git behavior with git config command or by editing ~/.gitconfig file
• .gitignore tells Git which files to disregard when committing changes