Engineers use version control. I did not use different versions of my work much in college. Therefore, I was not an engineer in college. But now, I do use VCS (git) extensively and enjoy the advantages of being an adequately skilled user of it. Still, sometimes I wish to do some more complex commands. Sometimes, I want to know the details of a particular git-call. These reasons motivated me to read Pro Git about 2 months ago. I read it, learned a lot of cool stuff, and then took a hiatus from my computer so all those cool features I learned have drifted off into some malfunctioning neurons in my brain. Therefore, I am going to reread this book and this time highlight things for future reference as I embark on my journey to becoming an advanced VCS (git) user.
1. Getting Started
- Types of Version Control
- local: copy files from one directory to another
- centralized: one server contains all versioned files (CVS, Subversion, and Perforce)
- problem is SPOF (single point of failure)
- good thing is everyone is on the same page most of the time
- distributed: everyone has their own copies, central server(s) has its copy. People push and pull occasionally to be in sync (Git, Mercurial, Bazaar, or Darcs)
- Git came out of Linux kernel when the VCS linus torvalds & team used started charging (bitkeeper)
- Git unique because it is not a delta-based VCS.
Every commit, Git basically takes a picture of what all your files look like at that moment and stores a reference to that snapshot. Link to previous snapshot of file if not changed in next commit. Git thinks about its data more like a stream of snapshots.
-
Git operations are mostly local so network latency unlike other vcs
- git computes checksums with SHA-1 of your commits which means files can't change or get corrupted without git knowing about it
- 3 main stages:
- modified: changed file but not committed to your database yet
- staged: marked modified file to go into next snapshot
- committed: data stored in database
- Working directory --> staging area --> .git repository
- .git repo --> checkout to working directory. make changes and stage them on a file which is stored in .git repo
- git config
- .git/config (local, trumps all other configs "--local")
- ~/.gitconfig or ~/.config/git/config (global, "--global")
- /etc/gitconfig: (system, "--system")
- view all settings and where they are coming
- git config --list --show-origin
- list all settings vanilla
- check what git thinks a config value is:
- find origin of a config value:
- git config --show-origin user.name
2. Git Basics
- git clone <url> [target directory name]
- Git can use several transfer protocol (for url). It can be through https:// or git:// or user@server:path/to/repo/.git (SSH)
- Short status
- Delete file from filesystem and git
- Delete file from git but not filesystem
- git rm --cached <filename>
- .gitignore file syntax
- start pattern with '/hello' to avoid recursivity
- end pattern with 'hello/' to specify directory
- negate pattern with '!hello'
- '*' matches 0 or more characters
- '?ello' matches single character
- 'a/**/d' to match multiple directories (i.e. would match 'a/b/c/d' and 'a/c/b/d')
- .gitignore files apply recursively from their root directory. So you can have multiple overlaying each other. Rules apply top to bottom.
- git diff [filename...] does staged files versus changed but not staged
- git diff --staged (or --cached) does staged versus file in previous commit
- git commit -am "My commit" would stage all tracked files and commit automatically
- git mv <original filename> <new filename> simplifies 3 git commands in 1
- git log has many options. this is probably the best
- git log --pretty=format:"%h %s" --graph [-x]where x is number of commits back
- Undoing things
- un-stage something you accidentally staged: git reset HEAD <filename>
- (Danger) undoing a modification to a file: git checkout <filename>
- working with remotes
- get data from your remote projects but do not merge: git fetch <remote>
- git pull origin master automatically fetches and then merges from origin/master branch into your local master branch
- git remote show origin helps inspect a remote and give tracking information about all branches on a remote
- git push origin <tagname> or git push origin master --tags
- git tag -d <tagname> to remove local tag
- git push origin --delete <tagname> to remove tag from remote
3. Git branching
 |
| Commit object, tree object, 3 blob objects |
- branching is the killer feature of git because it is lightweight and instantaneous, branching is recommended in git rather than other VCS
- Every commit looks like above, taking SHA-1 of subdirectories. Each new commit has pointer to previous commit through a parent key in the commit object (not shown above)
- branch is a moveable pointer to one of these commits
- HEAD is the pointer to which branch/tag you are currently checked out on
Comments
Post a Comment