I had some issues in finding the correct parameters. In case this also can help others, you can find the command here.

In case you are not yet familiar with the possibilities of `git filter-branch`, I suggest you look into the  Git book: Rewriting history, which has a good introduction to the command.

Now, to start you should clone the repository you want to change (since you will completely rewrite history in the new repos).

In short, without further explanation:

ORIG=path_to_original_repos
DEST=destination_of_rewritten_repos
DIR=directory_you_want_to_extract_from_ORIG

git clone $ORIG $DEST && \
cd $DEST && \
git filter-branch --prune-empty --tag-name-filter cat --subdirectory-filter $DIR -- --all

A few notes:

• --prune-empty will remove all commits that has no relation to the folder you extract
• --tag-name-filter cat will also rewrite any tags you put on the commits that are rewritten
• --subdirectory-filter takes a foldername as parameter, but I needed to add -- --all to make it work

I like the projects where you have a `master` branch and a few feature branches, which are actively being worked upon. Only a few branches are present at a given time, and this is less confusing for the contributors in the repository.

A couple of times I have been on a project where old branches were not cleaned up.

I found that it’s really easy to list these branches and clean them up.

Step 1: List all branches that are already merged:

$ git branch -r --merged

Lists all remote branches which are already merged. In case you want to find the old ones, you can sort them by commit date with the flag `–sort=committerdate`

Example:

$ git branch -r --sort committerdate

Committerdate flag will put the most recent branches last (you can reverse with minus: `--sort=-committerdate`.

Step 2: List only relevant branches.

The list returned from above will also contain origin/HEAD and origin/master with we have to filter out:

$ git branch -r --merged | cut -d / -f 2 | grep -v master | grep -v origin

we simply remove them with `grep -v`

Step 3: Feed list into a delete command

The above command will give you a list of branches to remove. Review the list carefully by printing each line:

$ git branch -r --merged | cut -d / -f 2 | grep -v master | grep -v origin | xargs -n 1 echo

When you are satisfied, you can remove each branch with a git push command. Example

$ git push --delete origin merged_branch1

You can simply repeat this for as many times as you have branches with the above expression. So, in total your final command will be:

$ git branch -r --merged | cut -d / -f 2 | grep -v master | grep -v origin | xargs -n 1 git push --delete origin

And done :)

This post is following my series of blog posts with useful Git Tips

Here is a first naïve impementation:

# Naïve implementation, just return whatever number I need
def sum_up_a(existing)
  return 2 if existing == 0
  return 1 if existing == 1
  return 0
end

sum_up_a(0)   # => 2
sum_up_a(1)   # => 1
sum_up_a(2)   # => 0
sum_up_a(3)   # => 0
sum_up_a(10)  # => 0

A slightly improved version could investigate number of existing, and return the diff only if needed:

MIN_COUNT = 2
# slightly improved with less conditionals
def sum_up_b(existing)
  return 0 if existing >= MIN_COUNT
  MIN_COUNT - existing
end

sum_up_b(0)   # => 2
sum_up_b(1)   # => 1
sum_up_b(2)   # => 0
sum_up_b(3)   # => 0
sum_up_b(10)  # => 0

Now, my favourite. If you look at the result of “MIN_COUNT – existing”. This number is negative or zero if we dont want extra rows. This means that we can rewrite the algorithm without any conditional logic like this:

# my favorite: return the biggest number either 0 or the difference
# Works well because difference is only positive for numbers where I 
# need some extra rows
def sum_up_c(existing)
  [MIN_COUNT - existing, 0].max 
end

sum_up_c(0)   # => 2
sum_up_c(1)   # => 1
sum_up_c(2)   # => 0
sum_up_c(3)   # => 0
sum_up_c(10)  # => 0

I like the latter, since too many conditional statements tend to clutter the logic. Just a short tip and a reminder for myself next time I run into this kind of problem.

Complete example gist avalilable on Github:

The easiest parts to change are:

• commit date
• commit log message

With this, you can easily modify time stamps if, for instance, your svn server clock is wrong.

Step 1: Modify repository to accept property changes

Make sure your svn repository accepts property changes. (commit dates and messages are just properties which are basically unversioned).

For your repository, you need to make sure the hook file `pre-revprop-change` exists, is executable and has exit code 0. An easy way is to write a file which always exists with code 0:

# write this file to `repository_directory/hooks/pre-revprop-change`

#!/usr/bin/env sh

exit 0

 Step 2: change date on the revision you want:

In this example I will set the date for revision 1 to May 1st, 1981.

svn propset --revprop -r 1 svn:date "1981-05-01T10:00:00.000000Z"

`svn propset` command will set certain properties. You can set many properties, and in this example we manipulate `svn:date`. If we want to change the log message, we can use `svn:log`.

I deliberately chose a date from before SVN was made. Subversion is from 2001, so this commit is now made 20 years before SVN was out.

All combined

I created a combined script, which has has the steps mentioned above, and also iterates to change all dates in a repository:

The result

The result is a fine modified commit history which could look like this

$ svn log

r10 | jesper | 1990-05-01 12:00:00 +0200 (Tue, 01 May 1990) | 1 line
commit number 10 (originally created 2015-04-15 22:21:35)
————————————————————————
r9 | jesper | 1989-05-01 12:00:00 +0200 (Mon, 01 May 1989) | 1 line
commit number 9 (originally created 2015-04-15 22:21:34)
————————————————————————
r8 | jesper | 1988-05-01 12:00:00 +0200 (Sun, 01 May 1988) | 1 line
commit number 8 (originally created 2015-04-15 22:21:33)
————————————————————————
r7 | jesper | 1987-05-01 12:00:00 +0200 (Fri, 01 May 1987) | 1 line
commit number 7 (originally created 2015-04-15 22:21:32)
————————————————————————
r6 | jesper | 1986-05-01 12:00:00 +0200 (Thu, 01 May 1986) | 1 line
commit number 6 (originally created 2015-04-15 22:21:31)
————————————————————————
r5 | jesper | 1985-05-01 12:00:00 +0200 (Wed, 01 May 1985) | 1 line
commit number 5 (originally created 2015-04-15 22:21:30)
————————————————————————
r4 | jesper | 1984-05-01 12:00:00 +0200 (Tue, 01 May 1984) | 1 line
commit number 4 (originally created 2015-04-15 22:21:29)
————————————————————————
r3 | jesper | 1983-05-01 12:00:00 +0200 (Sun, 01 May 1983) | 1 line
commit number 3 (originally created 2015-04-15 22:21:28)
————————————————————————
r2 | jesper | 1982-05-01 12:00:00 +0200 (Sat, 01 May 1982) | 1 line
commit number 2 (originally created 2015-04-15 22:21:27)
————————————————————————
r1 | jesper | 1981-05-01 12:00:00 +0200 (Fri, 01 May 1981) | 1 line
commit number 1 (originally created 2015-04-15 22:21:26)
————————————————————————

More info

• SVN book chapter on “pre-revprop-change”
• Rooot.net “Change a SVN commit log message or date“
• Stackoverflow: Can I go back and edit comments on an SVN checkin?
• Stackoverflow: Change the timestamp of a SVN revision

Now:

Before

The trick is really easy. CSS border radius can be applied in ellipsis shapes. Each corner has a  vertical and a horizontal radius. So you really just have to change the vertical radius of each corner like this:

And voila, the 60% and 40% vertical radius will do the trick.

I can highly recommend this really interesting presentation from Lea Verou on border-radius. It’s 14 minutes and highly worth the time.

Also, CSS Tricks has a collection of pre-cooked css shapes for inspiration: CSS-Tricks.com:ShapesOfCSS

This made Git history look horrible, with incorrect email, image. By accident I even wrote an incorrect name at one point.

Note the missing image.

Now it turns out my git repository was full of EMAIL instead of my correct email address.

$git log --pretty=fuller -1 9461d8b
commit 9461d8b
Author:     Jesper Rønn-Jensen <EMAIL>
AuthorDate: Tue Jan 13 09:56:13 2015 +0100
Commit:     Jesper Rønn-Jensen <EMAIL>
CommitDate: Tue Jan 13 09:56:13 2015 +0100
    adding install of hipchat and sourcetree

I had to rewrite the history in order to fix the email addresses. Turns out `git filter-branch` could help:

NEW_MAIL=”jesp….@gmail.com”
git filter-branch –env-filter ‘
> if [[ “$GIT_AUTHOR_EMAIL” = “EMAIL” || “$GIT_AUTHOR_EMAIL” = “$NEW_MAIL” ]]
> then
> if [[ “$GIT_AUTHOR_NAME” = “Ben Alman” ]]
> then
> GIT_AUTHOR_NAME=”Jesper Rønn-Jensen”
> GIT_COMMITTER_NAME=”Jesper Rønn-Jensen”
> export GIT_AUTHOR_NAME
> export GIT_COMMITTER_NAME
> fi
>
> GIT_AUTHOR_EMAIL=”$NEW_MAIL”
> GIT_COMMITTER_EMAIL=”$NEW_MAIL”
> export GIT_AUTHOR_EMAIL
> export GIT_COMMITTER_EMAIL
> fi
> ‘ — –all

10 seconds later, I could push the changed repository up back on the server. The result was much better:

Nice result :) All of this trouble was because I wanted to script the setup of a new machine. I used my clone of Ben Alman’s dotfiles, and this way his name slipped in before I figured out how to configure the .gitconfig file.

I really like that Git enables you to do almost anything. Especially that you can make mistakes and fix them later :)

Here are the adjusted tip, which works for me.

This Apple support  thread mentioned 3 tips:

1. Upgrade RAM
2. Turn off visual bling in overview pages
3. Vacuum iPhoto database file 

Could not do anything with tip 1 today (and I presume I cannot update the RAM on this 2012 macbook air)

Tip2 works: Go into settings > “appearance” and turn off “Outline” and “Drop shadow” in the overviews.

Tip3 needs adjustment. Originally *.db files were only in top folder ‘ ~/Pictures/iPhoto\ Library/’

Adjusted command to vacuum iPhoto database files:

cd ~/Pictures/iPhoto\ Library/ &&  for dbase in `find . -name '*.db'`; do sqlite3 $dbase "vacuum;"; done

css file:

.

In order to reduce the number of HTTP requests, I wanted to change it to using CSS sprites instead.

CSS file now has changed to lines like this:

.er_16{ background-position: -20px 0; }
.es_16{ background-position: -20px -20px; }
.et_16{ background-position: -20px -40px; }
.fi_16{ background-position: -20px -60px; }

> Generation of sprite: http://spritegen.website-performance.org/

> * Upload zip file with flags (in my case all the 16×16 px flags)
> * ignore duplicate images
> * build direction vertical
> * offset 4px (since the images are 16×14, this will make a 20-20 grid)
> * background/transparency colour: #ffffff
> * output format: PNG
> * Compress Image with OptiPNG: YES
> * Number of colours: True colour
> * css suffix: _16 (to indicate 16px icons)

Oh, before creatign the zip file, I removed all countries not matching a two-letter code.

cd 16/

find . -type f -name ‘???????*’ |xargs -I{} rm {}

This would remove countries named “_African Union.png”, “_United Nations.png”, “Olimpic Movement.png”  and similar.

For the following sizes of flags, I used these offset values:

16px, offset 4px will make a 20px grid

24px, offset 6px will make a 30px grid

32px, offset 8px will make a 40pxgrid

48px, offset 2px will make a 50px grid

64px, offset 6px will make a 70px grid

128px, offset 2px will make a 130px grid

git checkout `git rev-list -n 1 --before="2012-03-01" master`

Then, I ran cloc to count lines of code:

cloc .

Its really easy to register the point in time as a git branch so that you can get back and count lines of code later:

git co -b 2012-03

You can always verify latest revision like so:

Git log -1

The combination of git and cloc makes it really simple to keep history of codebase changes.

Since phantomJs is just released this week, I needed to build it myself.

I post this link to the binary in case it can benefit others.

PhantomJs 1.7.0 built on Linux Redhat Fedora 64 bit

PhantomJS is a fantastic tool for automating tests, since it is a headless browser which we wrapped in our automation framework to run javascript based tests very fast and easily.

I followed the build instructions (the instructions described for Amazon EC2 also worked on the fedora I am running).

sudo yum install gcc gcc-c++ make git openssl-devel freetype-devel fontconfig-devel
git clone git://github.com/ariya/phantomjs.git
cd phantomjs
git checkout 1.7
./build.sh --jobs 1

#(added step below by me)
sudo yum install upx

After finishing the build, run deploy/package.sh which creates PhantomJS binary tarball that can be moved around and/or extracted to a different location. This is the correct way to install the built binary into an arbitrary directory, i.e. do not simply copy the executable as it won't work.

The link I posted is the actual outcome of ./deploy/package.sh

For other operating systems, you could probably just get the binary at the official PhantomJs download page.