A Git script to list objects by size in an unpacked repo.

git-blob-list.rb

#!/usr/bin/env ruby

class GitObject
  attr_reader :git_type, :hash, :size
  def initialize(hash, git_type, size)
    @hash = hash
    @git_type = git_type
    @size = size
  end

  def to_s
    "#{@hash} (#{@size})"
  end
end

object_dirs = []
Dir.foreach('.git/objects') do |dir|
  if /^[a-f0-9A-F]{2}/.match(dir)
    object_dirs << Dir.new(".git/objects/#{dir}")
  end
end

hashes = []
object_dirs.each do |dir|
  hash_prefix = dir.path.split('/')[-1]
  Dir.foreach(dir.path) do |object_filename|
    if /^[a-f0-9A-F]{38}/ =~ object_filename
      hashes << "#{hash_prefix}#{object_filename}"
    end
  end
end
 
objects = {}

hashes.each do |hash| 
  objects[hash] = GitObject.new(hash, `git cat-file -t #{hash}`.chomp, `git cat-file -s #{hash}`.chomp)
end

objects.values.select do |object|
  object.git_type == 'blob'
end.sort! do |a,b|
  b.size.to_i <=> a.size.to_i
end.each do |blob|
  puts "#{blob.hash}\t#{blob.size}"
end

Sizes are in bytes (as given by git cat-file -s in line 36).

Alternative: http://naleid.com/blog/2012/01/17/finding-and-purging-big-files-from-git-history