Bo Jeanes
I am an software engineer who has lived and worked in New York, Brisbane, Chicago, San Francisco, and now Melbourne. I primarily work in Ruby though am a functional programmer at heart and a fan of programming languages in general. In particular, I love Rust and Clojure but keep my eye on many langauges all the time.
Throughout Mac OS X, almost every input field and text area have
consistent keyboard shortcuts to navigate and edit the text. However, in
both Terminal and iTerm, some of these don’t work like you might be used
to. Furthermore, both iTerm and Terminal respond differently to each
other, as well as to hour the rest of the system does it.
Three shortcuts I’ve especially missed are: Option-Left to move left one
word, Option-Right to move right one word, and Option-Backspace to
delete one word to the left.
The ideal solution to get these working as I’d like would be to
configure an inputrc file. However, after playing around for a few
hours, I still couldn’t get this working reliably using the Option key.
So, my fallback is to configure the terminal apps to do this themselves.
If you set up the terminal applications to use Option as the Meta key,
some remote systems will already work with the keyboard shortcuts I
want. However, locally only the Option-Backspace one works and even then
it only works out of the box in Terminal.
So, here’s how to get all three working in both Terminal and iTerm,
locally and on remote systems:
Terminal
First, enable Option as meta key option by going to the *Preferences >
Settings > Your Profile > Keyboard( and checking “Use option as meta
key” check box. This will get Option-Backspace working.
Next, while still in the Keyboard settings from the last step, press the
+ button. In the popup, choose “cursor left” for the Key, “option” for
the Modifier, and “send string to shell:” for the Action. In the text
field, type Esc then lowercase B key and press OK. This will get the
Option-Left shortcut working.
To enable the Option-Right shortcut, repeat the previous step, instead
choosing “cursor right” for the Key, and typing Esc then lowercase F in
the text field.
If you switch Terminal themes, you will have to repeat these steps for
the new theme.
iTerm
For the iTerm configuration, I will simply point you to this excellent
article which help
me get up and running. The sections that correspond to the shortcuts I
set up are 2.1, 2.2, and 2.4.
When writing new Cucumber scenarios, developers will usually start with
a failing scenario, incrementally making it pass, step by step. If this
is a common workflow for you, you’ll probably have noticed a trend of
having to re-run it each time a new step is failing, adding something
like Then show me the page after the last
passing step to see why the story is currently failing.
Chendo and I came up with a nifty little solution
in just a few lines that really helps this process of seeing what webrat
sees when a scenario fails. Throw this snippet of code in a file in your
features/support/ directory:
After do |scenario|
if scenario.respond_to? && scenario.status == :failed
save_and_open_page
end
end
I apologise to those who are uninterested for the rant that follows:
I can not stand nor hardly believe that the current government of my
otherwise awesome country would choose to completely ignore the advice
and expertise of industry professionals and the interests of the voters.
Today Steven Conroy, the Australian Minister of Communications, managed
to get a piece of legislation passed introduced that will require
all Australian ISPs to participate in the mandatory filtering of content
delivered to Australian citizen. Welcome to the likes of Iran and China.
The legislation has been disguised as a measure to protect children from
porn and the likes, despite repeated advice from industry professionals
that it will do little to curb illegal content, cost the tax payers
millions of dollars and ultimately make legal browsing a slower and
inferior experience than it already is.
Furthermore, the government has made a effort to actively
suppress
the contents of the block list, keeping those affected (read: everyone)
completely in the dark at to how it will affect their experience. The
block list was leaked and shown to include content about which the
government should have no say in how or why it is viewed such as some
youtube videos, topics on Euthanasia, a gay dating site, flash and
downloadable games, and sexual education material (contraceptives, etc).
To the Rudd government:
This is not good enough. I am a voter and a tax payer and I will not
stand for this. I will actively participate in removing this ridiculous
action and I will march the streets to show my support for an
unrestricted flow of information like the rest of the world will
continue to enjoy. I will show other tax payers how you have wasted $40
million (or more) of their money on a piece of technology no one wants
and that will not work and remind them that this is done in a time of
financial uncertainty instead of focusing on real issues like our carbon
footprint as a nation. I have the knowledge and the will to undermine
this pointless act, and I will use them to do so.
In the process of going through the redesign of http://mocra.com, I keep finding better ways to make
the HTML more semantic.
Jack Chen and I have been turning the design into
HTML over the last few days and we’ve been striving to be as semantic as
possible by taking advantage of new HTML5 elements and by separating the
content further from the design code with all the new CSS3 selectors.
Here’s an example of something we just did to make the HTML cleaner and
more meaningful, whilst still maintaining the look and style.
Here’s what we are styling:
Usually your HTML (or HAML in this case) would look something like this:
%footer
%aside
= link_to 'Dr Nic Williams', 'http://drnicwilliams.com/'
,
= link_to 'Bodaniel Jeanes', 'http://bjeanes.com/'
,
= link_to 'Jack Chen', 'http://chendo.net/'
,
= link_to 'Ryan Bigg', 'http://frozenplague.net/'
However, we can make this list of people more semantic by using an
<ul> element and some CSS magic.
Here’s what the new HTML (HAML here) would look like:
%footer
%aside
%ul
%li= link_to 'Dr Nic Williams', 'http://drnicwilliams.com/'
%li= link_to 'Bodaniel Jeanes', 'http://bjeanes.com/'
%li= link_to 'Jack Chen', 'http://chendo.net/'
%li= link_to 'Ryan Bigg', 'http://frozenplague.net/'
And the CSS (Sass here):
body
footer
aside
padding-top: 6px
float: left
ul
list-style: none
margin: 0
padding: 0
li
display: inline
&:after
content: ","
&:last-child:after
content: ""
The other day we got some HTML from the designer of one of our clients
to implement as the new UI for an app we are building at
Mocra. Overall the HTML was good but implementing
it started to feel a little moist (i.e. not DRY).
For example, around all the apps form the client’s design had a rounded
white box, the HTML of which went something like this:
<div class="content-wht">
<div class="content-wht-inside">
<fieldset>
<div class="frow">
<label>Currency:</label>
<select class="fs"><option>EUR</option></select>
</div>
<div class="frow frlast">
<label>File:</label>
<input type="file" name="" />
</div>
</fieldset>
</div>
</div>
<div class="cround" align="center">
<img src="/images/round-corners-bottom.gif" alt="" />
</div>
It’s a lot of div tags to have to repeat around every form that has
this style. Luckily a nice feature of Rails partials can help out.
Create a shared partial file that will act as a layout like this:
<!-- app/views/shared/_fieldset.html.erb -->
<div class="content-wht">
<div class="content-wht-inside">
<fieldset>
<%= yield %>
</fieldset>
</div>
</div>
<div class="cround" align="center">
<img src="/images/round-corners-bottom.gif" alt="" />
</div>
Then in your views that need to have this white box, simply wrap your
content in that layout like so:
<% render :layout => 'shared/fieldset' do %>
<div class="frow">
<label>Currency:</label>
<select class="fs"><option>EUR</option></select>
</div>
<div class="frow frlast">
<label>File:</label>
<input type="file" name="" />
</div>
<% end %>
<!-- Or, if the inner content is already in a partial -->
<%= render, :partial => 'form', :layout => 'shared/fieldset' %>
Dependencies
Install the GEOS and PROJ frameworks from
here
Postgres
Download and Compile
http://ftp2.au.postgresql.org/pub/postgresql/source/v8.4.1/postgresql-8.4.1.tar.bz2 | tar xjf -
cd postgresql-8.4.1
./configure
make && sudo make install
The files are now all installed in the right places, onwards to making
them usable!
Create a postgres User and Group
This is adjusted from the instructions in the comments
here).
# Find unused Group ID and User ID:
export GROUPID=`sudo dscl . -list /Groups PrimaryGroupID | ruby -e 'puts STDIN.read.scan(/\d+/m).map{|i|i.to_i}.uniq.sort.last.succ'`
export USERID=`sudo dscl . -list /Users UniqueID | ruby -e 'puts STDIN.read.scan(/\d+/m).map{|i|i.to_i}.uniq.sort.last.succ'`
# Create group:
sudo dscl . -create /Groups/_postgres
sudo dscl . -create /Groups/_postgres PrimaryGroupID $GROUPID
sudo dscl . -append /Groups/_postgres RecordName postgres
# Create user:
sudo dscl . -create /Users/_postgres
sudo dscl . -create /Users/_postgres UniqueID $USERID
sudo dscl . -create /Users/_postgres PrimaryGroupID $GROUPID
sudo dscl . -create /Users/_postgres UserShell /bin/bash
sudo dscl . -create /Users/_postgres RealName "PostgreSQL Server"
sudo dscl . -create /Users/_postgres NFSHomeDirectory /usr/local/pgsql
sudo dscl . -append /Users/_postgres RecordName postgres
That’s the clean way of making a user (so it doesn’t show up in your
Accounts preference pane, etc.
Clean up
sudo touch /var/log/psql.log
sudo mkdir /usr/local/pgsql/data
sudo chown -R postgres:postgres /usr/local/pgsql/data /var/log/psql.log
# Put Postgres in your path
export $PATH="/usr/local/pgsql/bin:$PATH"
sudo sh -c 'echo /usr/local/pgsql/bin > /etc/paths.d/pgsql'
Thanks to Apple we have a nice way of setting the PATH across the entire
system so every app should now know how to find the Postgres binaries
and the data directory is now set up to store the database files — now,
we just need to create them.
Initialise the Database
# Become the postgres user
sudo su - postgres
# If this throws next commant throws an error about shared memory,
# try putting "these lines":http://gist.github.com/224815 into
# /etc/sysctl.conf, rebooting, and trying again:
initdb -D /usr/local/pgsql/data -E UTF8
# Start the database
postgres -D /usr/local/pgsql/data >/var/log/psql.log 2>&1 &
# Create a test database and check you can connect to it
createdb test
psql test
# Just create a superuser for general dev tasks
createuser postgres
# we don't want to be the postgres user anymore
exit
Your done with the Postgres setup now!
PostGIS
curl http://postgis.refractions.net/download/postgis-1.3.7SVN.tar.gz | tar xzf -
cd postgis-1.3.7SVN
./configure --with-geosconfig=/Library/Frameworks/GEOS.framework/unix/bin/geos-config --with-projdir=/Library/Frameworks/PROJ.framework/unix
make && sudo make install
Phew! That one was easy.
Postgres Ruby Gem
This is the part that everyone cares about the most. If this works your
Rails apps should start working like a charm.
sudo env ARCHFLAGS='-arch x86_64' gem install pg
Dr Nic’s latest post, hash bang
cucumber,
reminded me of a piece of hax I whipped up a few weeks ago with Ruby and
Fish Shell.
Fish has an event system that allows you to register functions to be
auto-run after or during certain events (such as when a particular
environment variable is changed). One of this events is called
fish_command_not_found. It is triggered whenever you type a
non-existant command into the prompt.
With this in mind, you can trigger certain commands to run by matching
what fish couldn’t manage to run automatically by catching this event.
For instance:
function __fish_method_missing --on-event fish_command_not_found
method_missing $argv
end
funcsave method_missing
Given that function is created (simply paste the whole code block into
your Fish terminal), I can now create a command called
method_missing (or whatever you call inside your
__fish_method_missing) and place it somewhere in your $PATH — I like
~/.config/fish/bin.
My method_missing binary is simply something like this:
#!/usr/bin/env ruby
command = ARGV.shift
def run(cmd)
puts "Running #{cmd.inspect} instead"
system(cmd)
end
case command
when /^git(@|:\/\/).*\.git$/
run("git clone #{command.inspect}")
when /^(?:ftp|https?):\/\/.+\.t(?:ar\.)?gz$/
run("curl #{command.inspect} | tar xzv")
else
$stderr.puts "No default action defined in #{__FILE__.inspect}"
abort
end
Each command you want to register just becomes a new when
statement. For instance, to implement the functionality Dr Nic was
trying to achieve, I simply modify the case block as such:
case command
when /^[a-z0-9_\-\/]+\.feature(:\d+)?$/
run("cucumber #{command}")
# ...
end
The existing entries I have in my method_missing command will
auto-clone the repository of a pasted Git URL and download and expand a
URL for a tar file, respectively. Not too shabby, and dead easy to
implement
Apologies readers — I just realised that the code formatting is kinda
screwed up in RSS feeds because I am using some HTML5 elements to
represent my code blocks and using CSS to style them appropriately on
the site. The RSS feeds obviously lack these stylesheets and so look
pretty crap. I am not sure what the best solution is yet. Perhaps I
embed a <style> in each RSS post or simply have the RSS feeds
only show an excerpt of the full post.
Any ideas?
For a project I am working on at the moment, I needed a Hash
that returned a different default value, i.e. not nil.
Specifically, I needed it to return another Hash, and for that
internal Hash’s default value to be an Array, like so:
hash = # ...
p hash.class #=> Hash
p hash[:non_existant_key].class #=> Hash
p hash[:another][:non_existant_key].class #=> Array
Naïvely, I threw this into my code, and went along my merry way:
entries = Hash.new(Hash.new([])). A few days later, I came back
to this code to finish it and realised that something was awry with the
values I was getting out of the Hash. It took me a while to figure it
out because the values were all integers being summed together and I was
never looking or setting the values directly.
However, eventually I saw a pattern — many of the values were the same.
For instance, given the following assignments, I should expect
[ruby]p entries[153][:monday] #=> [12.hours]:
entries[153][:monday] << 12.hours
entries[153][:tuesday] << 3.hours
entries[87][:monday] << 7.hours
entries[87][:tuesday] << 2.5.hours
However, what I found was the following:
p entries[153][:monday] #=> [12.hours, 3.hours, 7.hours, 2.5.hours]
p entries[153][:tuesday] #=> [12.hours, 3.hours, 7.hours, 2.5.hours]
p entries[87][:monday] #=> [12.hours, 3.hours, 7.hours, 2.5.hours]
p entries[87][:tuesday] #=> [12.hours, 3.hours, 7.hours, 2.5.hours]
# And in fact:
p entries[:any][:thing] #=> [12.hours, 3.hours, 7.hours, 2.5.hours]
Let that soak in for a second. Hash.new({}) uses the same
instance of the internal Array as the default value for each
of the inner-most keys. It also doesn’t set the key you want so
p entries still printed {}. In retrospect, it is
blatantly obvious that it does this, but the ramifications are still
huge.
The way to get the intended effect is of course to use the block syntax
of [ruby]Hash.new which, while much uglier, definitely works as
expected:
entries = Hash.new {|hash, key| hash[key] = Hash.new {|h, k| h[k] = [] }}
entries[153][:monday] << 12.hours
entries[153][:tuesday] << 3.hours
entries[87][:monday] << 7.hours
entries[87][:tuesday] << 2.5.hours
p entries #=> {87=>{:monday=>[7.hours], :tuesday=>[2.5.hours]}, 153=>{:monday=>[12.hours], :tuesday=>[3.hours]}}
Wow. Obvious, but good to remember. I am sure most of you have had this
issue before, but I think it is still worthy of mention…
Today I casually tweeted a hypothesis that
@chendo drained my battery. The image below
shows the amusing responses that followed:
