#!

After trying in vein to find decent documentation on the Merge functionality of Unison I eventually gave in and invested an hour or two trying out different Linux based merge programs in order to find something Unison, the merge program and I were happy with.

So onto the diff-merging candidates:

VimDiff (part of Vim)

Doesn’t work too well with Unison. I can’t get gnome-terminal (my terminal of choice) to execute synchronously so I was forced to fall back on XTerm instead.

After using XTerm I immediately hit upon a problem: there is no sensible way of outputting the results of a merge in VimDiff. Which is a pity because I do prefer VIM as an editor but this lack of functionality is annoying (yes I could have used a wrapper script but I’m trying to keep this simple).

Meld

Very pretty but lacks features such as ‘Use Left-hand-side-file for all changes’ and some of the other filtering options provided by KDiff3.

Also, like, VimDiff, does not provide any sensible output mechanism of the eventually merged file.

KDiff3

Not as good looking as Meld but provides pretty much everything I want in a merge program including the all important merged file output that Unison requires.

Can get a little annoying though with its insentient pop-up boxes. I wish there were a ‘Dont show this option again’ option to each. KDE is usually quite good with this so I’m not sure why KDiff3 managed to escape this unspoken standard.

Conclusion

So how to use KDiff3 with Unison? First install kdiff3:

apt-get install kidff3

Then open your Unison profiles .prf file and insert the following:

merge = Name * -> kdiff3 CURRENT1 CURRENT2 -L1 'Local File' -L2 'Remote File' -o NEW -m
confirmmerge = true

If you are comfortable with KDiff3 you could remove the ‘confirmmerge’ line which forces Unison to double check you actually want to accept the changes.

The downside to the above interactive merging process is there is no sensible way to preview the merge in Unison (the ‘View Merge’ button in the confirmation dialogue). For now though its preferable to not having this functionality at all.

Simple shell script fix this week – convert all newlines to null characters.

Since most Linux shells have a major problem with doing this in a sensible way, the following (on my box called simply ‘0′) takes any input (either from files or via a pipe) and converts the characters.

1
2
3
4
5
6
7

#!/usr/bin/perl
# From: http://hash-bang.net/2009/04/converting-nullconverting-null/
# Author: Matt Carter <m@ttcarter.com>
while (<>) {
	s/\n/\x00/g;
	print;
}

So now commands like:

find -type f | 0 | xargs -0 | mplayer

work perfectly (in the above case to play all files in directories recursively.

You may well ask why I don’t use the simple ‘-print0′ argument for the find program – because it doesn’t work when using later pipes.

For example if i wanted to sort the above, ’sort’ would see the null symbol as a regular character.

So ‘0′ works nicely in this case:

find -type f | 0 | sort | xargs -0 | mplayer

Or even:

find -type f | 0 | shuffle | xargs -0 | mplayer

To play all files recursively, in random order. See the article on shuffle for the source code of that filter.

While the Fisher-Yates shuffle sounds like 1920’s dance move, it is the standard method for shuffling larger arrays of data.

The below operates like a simple Unix filter taking an array of data (separated by newlines) and returning a quickly sorted output.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

#!/usr/bin/perl
# Usage: cat file | shuffle
# From: http://hash-bang.net/2009/04/shuffling-shell-inputshuffling-shell-input/
# Return a shuffled version of the input stream
# Author: Matt Carter <m@ttcarter.com>
 
sub fisher_yates_shuffle {
	my $array = shift;
	my $i;
	for ($i = @$array; --$i; ) {
		my $j = int rand ($i+1);
		next if $i == $j;
		@$array[$i,$j] = @$array[$j,$i];
	}
}
 
while(<>) {
	chomp;
	push @input, $_;
}
 
&fisher_yates_shuffle(\@input);
print join("\n",@input);

For example:

ls | shuffle

Returns a randomized list of files in the current directory.

Like most of the word I use Apache to serve up both my live and test websites.

The downside to having a box for development though is the constant config of apache to host these sites, most of which require absolute links (e.g. ‘/css/style.css’) to function.

The Apache snippet below sets up dynamic sub-domains which can be changed without reloading Apache every time you add, remove or update anything something. Simply drop your sites inside /var/www whenever you want a new subdomain. Alternatively you can symlink if they are hosted elsewhere on the box.

Its pretty easy to install, simply paste the below in your /etc/apache2/apache.conf file…

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

# Easy virtual host config
# See http://hash-bang.net/2009/03/easy-apache-virtualhosteasy-apache-virtualhost/
<VirtualHost *>
DocumentRoot /var/www/
ServerAdmin admin@localhost
Options +FollowSymLinks -Indexes
LogLevel debug
ServerAlias localhost
ServerName localhost
RewriteEngine  on
RewriteLogLevel 3
RewriteCond    %{HTTP_HOST}  ^localhost
RewriteRule    ^(.*)$        /www/$1 [L]
RewriteCond    %{HTTP_HOST}  ^www.*
RewriteRule    ^(.*)$        /www/$1 [L]
RewriteCond    %{HTTP_HOST}  ^(.*)\.localhost
RewriteRule    ^(.*)$        /%1/$1 [L]
</VirtualHost>

Now add your servers to the /etc/hosts file:

127.0.0.1      server.localhost

Now any directory you place in the /var/www directory automatically becomes a subdomain. You can now visit http:///server.localhost as a fully fledged webhost.

If you want to use another name than the rather boring ‘localhost’ simply replace the text in both text pastes above.

If only there were a way of removing that /etc/hosts hack it would mean zero config from then on. If anyone has any ideas do let me know. The addition to the hosts file seems to need applying if you are working from the same box the Apache server resides on. If you arn’t and the box is a local dedicated server its not needed at all and domains can be added and removed as necessary.

One of the regrettably unavoidable aspects of the Unix shell like environments is the impedance between what a user says and what a user means. While this is present in all computing environments the sheer power of Unix based command lines make a potential mistake catastrophic.

Who amongst us has not at some point done something like:

> rm * .tmp

or

> rm *>tmp

The former makes the mistake of a space between the dot and the ‘tmp’ part and the latter the accidental holding down of the shift key while pressing the intended full stop.

While searching for an alternative for the I-have-done-this-too-many-times-and-now-its-embarrassing approach of nuking everything with the ‘rm’ command its time i did something about it.

Inserting the following in your ~/.bashrc file will remap the ‘rm’ command to a slightly safer move-to-trash like behavior. Its not comprehensive but it has saved my life on more than one occasion.

1
2
3
4
5
6
7
8
9
10
11
12
13

# Trash support by Matt Carter <m@ttcarter.com
# Source and information: http://hash-bang.net/2009/03/trash-with-bash
function trash() {
	if [ -d "$HOME/.trashcan/$1" ]; then
		# Already exists in bin - remove
		rm -r "$HOME/.trashcan/$1"
	fi
	mv --target-directory=$HOME/.trashcan/ -- "$@"
}
alias "rm=trash"
alias "emptytrash=/bin/rm -rf $HOME/.trashcan/* $HOME/.trashcan/.??*"
alias 'rm!=/bin/rm -r'
mkdir ~/.trashcan 2>/dev/null

Now the command ‘rm’ moves files into ~/.trashcan. You can also use the command ‘rm!’ when you really mean delete immediately and the utility command ‘emptytrash’ to clean everything out.