git_objects.rb

require 'digest/sha1'
require 'zlib'
require 'pp'

module Git
  
  OBJECTS = {}
  
  class Object
  
    def initialize(content)
      @uncompressed_content = header(content) + content
      puts hash
      return hash
    end
    
    def hash
      @sha1 ||= Digest::SHA1.hexdigest(@uncompressed_content)
    end
    
    def compressed
      @compressed_content ||= Zlib::Deflate.deflate(@uncompressed_content)
    end
    
    def path
      @path ||= ".git/objects/" + hash[0..1] + "/" + hash[2..-1]
    end
    
    def header(type, content)
      "#{type} #{content.length}\0"
    end
    
    def save
      OBJECTS[path.to_sym] ||= compressed
    end
    
  end
  
  class Blob < Object
    
    def header(content)
      super("blob", content)
    end
    
    def hash_bytes
      eval(hash.scan(/../).map {|p| "\x" + p.upcase}.join)
    end
    
  end
  
  class Tree < Object
  
    def initialize(content)
      validate(content)
      super(content)
    end
    
    def header(content)
      super("tree", content)
    end
    
    private
    
    def validate(content)
      # must be a list of trees and blobs
      # 
    end
    
  end
  
  class Commit < Object
  
    def initialize(content)
      validate(content)
      super(content)
    end
    
    def header(content)
      super("commit", content)
    end
    
    private
    
    def validate(content)
      # must refer to a tree hash, zero or more parent commit hashes,
      # an author (with datetime), a commiter (with datetime), and a message
    end
    
  end
  
  class Tag < Object
    def initialize(content)
      validate(content)
      super(content)
    end
    
    private
    
    def validate(content)
      #tag must have an object hash, object type, tag name, tagger (with datetime), and message
      # example:
      #object 7e7b6a09dc5e466b7992fea125732c67239ac92b
      #type commit
      #tag v2.0
      #tagger Joe Schmo <[email protected]> 1430542850 -0700
      #
      #tag test
      #
    end
    
    def save
      super
      #and also add a file in .git/refs/tags whose name is the tag name and
      # whose content is the sha1 of the tag object created in the object store
    end
  end

end

# Every command in git either makes one or more objects in the database,
# changes a reference, or displays an object or reference.
# The only exception to this rule is modifying your config file.

# git add some_file.txt
# results in the following:
# a new blob is created, and added to the object store
file_name = "some_file.txt"
file_content = "some text in a file"

blob = Git::Blob.new(file_content)
blob.save
pp Git::OBJECTS

# it also adds the new blob to the index so that when you commit
# the following tree is created
tree_content = "100644 #{file_name}\x00#{blob.hash_bytes}"

tree = Git::Tree.new(tree_content)
tree.save
pp Git::OBJECTS

commit_content  = (tree      = "tree #{tree.hash}\n")
commit_content += (author    = "author Dominic Muller <[email protected]> 1430522440 -0700\n")
commit_content += (committer = "committer Dominic Muller <[email protected]> 1430522440 -0700\n")
commit_content += (message   = "\nI'm a commit message!\n")

# git commit -m "I'm a commit message!"
# this will now result in a commit object being created that points to that tree
# that the index made for you.
# The author and commiter will be grabbed from your config files
commit = Git::Commit.new(commit_content)
commit.save
pp Git::OBJECTS

# congrats!  Now the object store has 3 new key-value pairs.  One blob, one tree, and one commit
# The files are saved after being compressed with the zlib deflate compression.
# so the key is the hash of the content of the object (including the header prepended),
# and the value is the compressed content (including the header prepended).

# A branch is a reference to some commit
# When you make your first commit, Git sets the master reference to point to its hash

master = commit.hash

# Git also sets a reference called HEAD to point to some branch.
# Which in turn points at some commit hash, but we'll just use the branch name for now
# let's make another commit!

file_content = "new file here!"
file_name2 = "new_file.txt"

blob2 = Git::Blob.new(file_content)
blob2.save

# and now to make a tree which has BOTH files so far

tree_content  = (tree_file_1 = "100644 #{file_name}\x00#{blob.hash_bytes}")
tree_content += (tree_file_2 = "100644 #{file_name2}\x00#{blob2.hash_bytes}")

second_tree = Git::Tree.new(tree_content)
second_tree.save

# And of course, if we're going to make history,
# we have to tell the new commit where he came from
# which requires adding a parent line

new_commit_content  = (tree      = "tree #{second_tree.hash}\n")
new_commit_content += (parent    = "parent #{commit.hash}\n")
new_commit_content += (author    = "author Dominic Muller <[email protected]> 1430523736 -0700\n")
new_commit_content += (committer = "committer Dominic Muller <[email protected]> 1430523736 -0700\n")
new_commit_content += (message   = "\nA second commit.  I have a parent!\n")

new_commit = Git::Commit.new(new_commit_content)
new_commit.save
master = new_commit.hash
pp Git::OBJECTS

# If we want, we can go re-inflate any compressed objects

compressed_commit = Git::OBJECTS[:".git/objects/#{new_commit.hash[0..1]}/#{new_commit.hash[2..-1]}"]

puts Zlib::Inflate.inflate(compressed_commit).split("\0")[1..-1].join

# btw, that last line is exactly what `git cat-file -p HEAD` would have done