I've Got a Ruby and Everything Is a Web App

[igarewa]

Add github.com SSH fingerprints to known_hosts

Verify the SHA256 fingerprints with these commands:

ssh-keyscan -t rsa github.com | awk '{print $3}' | base64 -d | sha256sum -b | awk '{print $1}' | xxd -r -p | base64 ssh-keyscan -t dsa github.com | awk '{print $3}' | base64 -d | sha256sum -b | awk '{print $1}' | xxd -r -p | base64

Github Fingerprints are documented here

Save them with these commands:

ssh-keyscan -t rsa github.com >> ~/.ssh/known_hosts ssh-keyscan -t dsa github.com >> ~/.ssh/known_hosts

[igarewa]

One-liner to generate a key for signing alpine linux packages.

docker run --rm --env "PACKAGER=Your Name <[email protected]>" \ gliderlabs/alpine:3.1 \ /bin/sh -c \ "apk-install abuild && \ abuild-keygen && \ ls /root/.abuild/*.rsa && \ cat /root/.abuild/*.rsa && \ ls /root/.abuild/*.rsa.pub && \ cat /root/.abuild/*.rsa.pub"

This will output the private and public key as well as the correct name for it (including the fingerprint suffix).

[igarewa]

A protip by kennu about python, vim, and json.

[igarewa]

Testing javascript that uses haml_coffee_assets (*.hamlc templates) with jasmine-headless-webkit

*.hamlc templates are great. jasmine-headless-webkit (JHW) is great. But getting the two working together is a pain.

There's a good chance you see something like window.HAML is undefined.

# spec/javascripts/helpers/spec_helper.rb require 'sprockets' require 'haml_coffee_assets' require 'haml_coffee_assets/tilt/template_handler' # JHW does not support ERB templates. However, haml_coffee_assets uses ERB # templates for its config. This pre-renders the ERB templates and passes # them to JHW. # Borrowed heavily from here: https://raw.github.com/dzello/headless_hamlc/master/spec/javascripts/helpers/spec_helper.rb Jasmine::Headless.register_engine '.hamlc', ::HamlCoffeeAssets::Tilt::TemplateHandler # this routine compiles and runs the hamlcoffee.js.coffee.erb and makes # the result available to JHW def add_haml_coffee_compiled_asset_path(asset_paths) # find the gem asset path that contains the hamlcoffee.js.coffee.erb haml_coffee_gem_asset_path = asset_paths.find { |path| path =~ /haml_coffee_assets/ } # compile the erb file into hamlcoffee.js.coffee compiled_haml_coffee_template = ERB.new( File.read(File.join(haml_coffee_gem_asset_path, "hamlcoffee.js.coffee.erb"))).result(binding) # create a tmp directory to put the compiled code in # this assumes you are running out of your project root! FileUtils.mkdir_p (tmp_haml_coffee_assets_path = File.expand_path("../../../../tmp/haml_coffee_assets/javascripts", __FILE__)) File.open(File.join(tmp_haml_coffee_assets_path, "hamlcoffee.js.coffee"), 'w') do |f| f.write compiled_haml_coffee_template end # add the new asset path containing the compiled file asset_paths << tmp_haml_coffee_assets_path # remove the original asset path from the gem, let you get the 'Skipping File' warning for the erb asset_paths.reject! { |path| path == haml_coffee_gem_asset_path } end # load any haml coffee configuration settings #require File.expand_path("../../../../config/initializers/haml_coffee_assets.rb", __FILE__) # now update the files list class to use these paths instead of the original updated_asset_paths = add_haml_coffee_compiled_asset_path(Sprockets.find_gem_vendor_paths(:for => 'javascripts')) Jasmine::Headless::FilesList.instance_variable_set(:"@asset_paths", updated_asset_paths)

Make sure you require hamlcoffee in your javascript/coffeescript.

My gem versions

jasmine-headless-webkit (0.9.0.rc.2)

haml_coffee_assets (1.14.0)

[igarewa]

Random 'Pr' inserted into HTTP request POST data

Have you ever seen an HTTP request come into your application with the last two characters of the HTTP POST data unexpectedly changed to 'Pr'?

Seems like a pretty weird edge case. But I was seeing this on a project recently, and it was baffling.

Since 'Pr' is short and also a common acronym, this was very hard to research. However, I eventually found some others with the same problem:

http://stackoverflow.com/questions/4903088/problem-ajaxing-json-object-on-mac-firefox-version-3-6-12/13387096#13387096

http://stackoverflow.com/questions/6585660/jquery-ajax-error-in-firefox-on-mac-with-managed-users

https://bugzilla.mozilla.org/show_bug.cgi?id=489575

It is caused by a conflict between Firefox and Mac OS 10.5 parental controls. Newer versions of Mac OS (i.e. 10.6 and later) are not affected. Safari on 10.5 is also not affected.

Client-side workarounds are easy: switch browsers; disable parental control; or upgrade the operating system (10.5 is pretty old these days).

Server-side workarounds are a bit harder. The main options are:

Use SSL

Add a dummy field (only works if you are dealing with an HTML form)

Pad the post data with client-side javascript then stripping out the padding server side.

Try to detect missing data with some sort of server side middleware and guess what the last two characters should be.

I opted to ignore the problem. It is rare and getting rarer. There have been 3 big OS releases since 10.5. As of writing I've seen estimates than 10.5 is only 4% of Mac users. And I'm guessing that very few of those enable parental controls.

[igarewa]

Enabling input of pinyin tones in Fedora 16

The m17n input engine for entering pinyin tones (aka m17n:zh:pinyin, aka zh-pinyin.mim) is disable in Fedora 16 (and possible some earlier versions). There seems to be no GUI way to re-enable it. In fact, there seems to be no way to re-enable it without messing with the config files directly.

If you don't mind messing with the config files, then just edit the file /usr/share/ibus-m17n/setup/default.xml. Search for an entry with the name m17n:zh:pinyin, and set he rank to an integer greater than zero.

If you'd rather not mess with the files set up by the package manager, I don't think you can actually enable this input engine. (If you know how, please add a comment!) However, you can create a new input engine that is essentially a copy. Create a file called ~/.m17n/zh-mypinyin.mim with the following contents:

;; zh-mypinyin.mim -- Input method for Chinese Pinyin (input-method zh mypinyin) (description "Input method for Chinese Pinyin characters.Note that it's not for inputting Han characters.") (title "MyPinYin") (include (zh pinyin) map vowel) (include (zh pinyin) map tone) (state (init (vowel (shift select))) (select (tone (shift init))))

[igarewa]

Ensuring models get reloaded by Spork when using FactoryGirl

I use guard and spork to speed up my testing cycle. I recently had an issue where a particular model was not being reloaded between test runs. This was really slowing down the test cycle because either I had to manually force spork to reload (by, for example, saving spec_helper.rb) or update my Guardfile to reload spork every time that model changed.

The reason this particular model was not reload was that it was being required too early. Spork forks after a certain point so most files you are likely to change get reloaded between each test run. However, if you require (explicitly or implicitly) a file from one of the files the loads before the fork, then that require file will not get reloaded.

In my case the problem was a FactoryGirl factory. I had this:

factory :started_event, :class => Event do state 'started' name end

By referencing the Event model, Rails automatically found and required apps/models/event.rb. The solution was to change that to a string instead of the actual constant. Then file was loaded after spork forked. Here's what I ended up with:

factory :started_event, :class => 'Event' do state 'started' name end

[igarewa]

Verifying SSL Certs with Net::HTTP#ca_path

If you want to have fine grained control over which CAs you trust (instead of just trusting everything in /etc/ssl/certs, for example), or if you want to explicitly trust self-signed certs (common in test/development environments), then the Net::HTTP#ca_path is your friend. But it is not a very well documented friend.

First, here is an example of an SSL request.

require 'cgi' require 'net/http' require 'net/https' def post(url, body, username, password) uri = URI.parse(url) http = Net::HTTP.new(uri.host, uri.port) http.use_ssl = true http.verify_mode = OpenSSL::SSL::VERIFY_PEER http.ca_path = File.expand_path(File.join(File.dirname(__FILE__), 'certs')) req = Net::HTTP::Post.new(uri.path, {"Content-Type" => "application/xml"}) req.basic_auth username, password req.body = body http.request(req) end

This is a fairly common example of making an HTTPS request with ruby. The first interesting bit is http.verify_mode = OpenSSL::SSL::VERIFY_PEER. This makes sure that the SSL cert of the server is actually verified. You will see many examples on the web that say to set this to http.verify_mode = OpenSSL::SSL::VERIFY_NONE. Setting it to NONE will disable SSL cert verification. This causes you to lose many of the benefits of using SSL in the first place. In my opinion, then only time this is OK is for occasional testing or debugging.

The second interesting bit is http.ca_path = File.expand_path(File.join(File.dirname(__FILE__), '..', '..', 'certs')). This tells Ruby where to find the CAs that you trust. This particular line says to look in the directory of the current file for a subdirectory called certs.

The point that is most poorly documented is what the certs in that directory should look like. The answer is that they should be in PEM format, and they should be named according hash value of the cert. You can generate the hash value with this command:

openssl x509 -noout -in secure.example.com.pem -hash

In order to keep things readable, I suggest you name the files according to the domain name, then create a symlink for the hashed version. Here is an example of creating the symlink:

ln -s secure.example.com.pem `openssl x509 -noout -in secure.example.com.pem -hash`.0

The extra ".0" is needed. I'm not sure why.

The above Ruby code was used with Ruby 1.9.2. I know 1.9 made a few changes to SSL handling, so if you are running this on 1.8.x YMMV.

Update (2011-05-02): You can automatically create the hash symlinks with the tool c_rehash, which should be automatically installed with openssl. Just run c_rehash . in the directory with the certs.

[igarewa]

Quickly dump ActiveRecord to CSV (Ruby 1.9)

def dump_model(klass) File.open("#{klass.to_s.underscore.pluralize}.csv", 'w+') do |f| columns = klass.columns.collect(&:name) output = CSV.generate do |csv| csv << columns.map { |column| klass.human_attribute_name(column) } klass.all.each do |item| csv << columns.collect { |column| item.send(column) } end end f.puts output end end dump_model(Posts) # => look in RAILS_ROOT/posts.csv

[igarewa]

awk is one of those amazingly useful UNIX tools that is insanely hard to learn. This link helps you get some useful things done without actually needing to learn a ton about awk programming.

[igarewa]

Preparing a CentOS 5.5 VirtualBox guest for Chef

This will be the longest post yet. I had enough trouble doing this that I thought someone doing the same thing may find it useful.

I make this VM is for testing out chef scripts.

The OS is CentOS 5.5 32-bit. VirtualBox doesn't include an option for CentOS, so just choose Redhat.

It was pretty much a basic/default install using the Netinstall ISO.

In order to allow SSH from my host OS to this VM, I followed this guide: http://muffinresearch.co.uk/archives/2010/02/08/howto-ssh-into-virtualbox-3-linux-guests/

When you restart the machine after adding the 2nd NIC, you should be prompted to configure it.

The IP static address I chose is 192.168.56.10.

Once this is done, you should be able to do ssh [email protected] from the host machine.

create /root/.ssh/authorized_keys

copy the host machine's public SSH key into there

I wouldn't do this on production (not for the root user anyway)

# network setup #

add this line to /etc/sysconfig/network

GATWAY=10.0.2.1

your IP maybe different, depending on your virtual box settings.

Run ifconfig to find out what it should be. Basically, look at the IP address and change the last digit to be "1"

remove this line from /etc/sysconfig/network-scripts/ifcfg-eth1

GATEWAY=192.168.56.10

shutdown -r now

we could do "service network restart" but rebooting ensures that it works after rebooting

ping yahoo.com to test your internet works

# Install dev tools #

(needed for compiling, etc.)

yum -y groupinstall "Development Tools"

# Installing Ruby #

zlib and openSSL isn't included by default

yum install zlib zlib-devel

yum install openssl openssl-devel

# this is also useful: Perl-compatible regular expressions

yum install pcre pcre-devel

reference

http://blog.jeffcosta.com/2010/09/10/7-steps-to-rails-on-centos/

http://www.railshostinginfo.com/blog/2007/07/13/how-to-get-ruby-openssl-to-work-on-centos/

http://library.linode.com/development/frameworks/ruby/ruby-on-rails/nginx-centos-5

confirm ruby isn't already installed

rpm -qa | grep ruby

mkdir /source

cd /source

wget ftp://ftp.ruby-lang.org//pub/ruby/1.9/ruby-1.9.2-p0.tar.gz

tar zxf ruby-1.9.2-p0.tar.gz

cd ruby-1.9.2-p0

./configure

make

make install

cd /source/ruby-1.9.2-p0/ext/openssl

ruby extconf.rb

make

make install

# Installing chef #

gem install chef

# Other Gotchas #

If doing anything with users, you need a special/hacked version of ruby-shadow because the standard version doesn't work with ruby 1.9.

I have my hacked version here: http://github.com/thoughtless/ruby-shadow-cookbook

Special thanks to Mathias Biilmann (http://workingwithrails.com/person/5686-mathias-biilmann http://mathias-biilmann.net/ http://github.com/biilmann) who did all the hard work in making ruby-shadow work with 1.9.

[igarewa]

Free fonts have met their match. We know how hard it is to find quality freeware that is licensed for commercial work. We've done the hard work, hand-selecting these typefaces and presenting them in an easy-to-use format

[igarewa]

Copy from vim to clipboard

"+y

Thanks to: http://stackoverflow.com/questions/677986/vim-copy-selection-to-os-x-clipboard

+ is apparently a special register. It works in reverse to. To paste from clipboard:

"+p

[igarewa]

Quickly dump ActiveRecord to CSV

require "fastercsv" def dump_model(klass) File.open("#{klass.to_s.underscore.pluralize}.csv", 'w+') do |f| columns = klass.columns.collect(&:name) output = FasterCSV.generate do |csv| csv << columns.map { |column| klass.human_attribute_name(column) } klass.all.each do |item| csv << columns.collect { |column| item.send(column) } end end f.puts output end end dump_model(Posts) # => look in RAILS_ROOT/posts.csv

[igarewa]

Testing Ruby on Rails routes in the console

app.my_route_path # => "/my/route/" app.my_route_url # => "http://www.example.com/my/route/"

app is already defined for you when you run script/console

[igarewa]

Stop whatever is running on a certain port

lsof -i :2998 -Fp | awk '/^p([0-9]+)$/ { sub(/p/,"",$0); print $0 }' | xargs kill -2

2999 is the port

2 (at the very end) is the signal (in this case, SIGINT--the same thing as CRTL+C)

I think I will assign this to an alias and use it to make sure I don't have anything else running before I start a local web server for development.

[igarewa]

Search word under cursor in Vim

* to search forwards

# to search backwards

Thanks to: http://www.unix.com/unix-dummies-questions-answers/42968-how-can-i-search-word-under-cursor-vi.html