Pragmatic Development Notes

After several successful updates of Ubuntu (starting from 8.10) last one was, more or less, complete mess. I was waiting impatiently for new Lucid Lynx to be released and started update as soon as I downloaded and burned DVD iso image. It seems everything goes smoothly till first reboot. Emacs was completely removed and new version was not installed. On my Compaq 8710w laptop with Quadro FX 1600M graphics card splash screen was displayed in low resolution and looked quite ugly. Moreover I was not able to install Emacs again.

I’ve decided to reinstall everything, so I’ve made backup of my home folder and start installation on a newly formatted partitions. Situation was a little bit better. I was able to install Emacs and all applications I need but problem with graphics card was still there as well as lousy splash screen. I’ve found few suggestions how to fix that by altering grub files and it really solved resolution problem but, at the same time, whole splash screen was shifted left and basically I was again on the beginning.

There is one more thing that annoyed me. I like Shere Khan Black Hand mouse pointers so I’ve installed that theme, but arrow pointer was still white, although some other pointers were changed (hand is small and black). Whatever I’ve tried arrow was keep staying white.

Then, after a few days, I saw black arrow and was really surprised. It turned out that it was unpleasant surprise because compiz stopped working and Nvidia driver was not enabled. Obviously it was time to try latest Nvidia drivers so I’ve downloaded and installed them. I must say that situation is now a little bit better – compiz is working, but I still have no luck with mouse pointers, nor with splash screen.

If Canonical’s goal is to increase number of Ubuntu users, they should really try to avoid such problems, since average user will certainly not be able to fix them and will switch to some other operating system.

Setting up Rails and Ruby to use MS SQL server was always painful task. Fortunately things have changed – a lot! With new Rails SQL Server 2000, 2005 and 2008 Adapter and Christian Werner’s ruby-odbc gem you can do it in a few minutes.

If you want to use these gems on Windows grab Ruby installation from RubyInstaller site and be sure to install DevKit prior to installing ruby-odbc.

Versions of ruby-odbc before 0.9999 do not work on mingw based (RubyInstaller) Ruby. Luckily author was very fast and made new version very quickly after I sent him a patch. Thanks Christian!

Both gems work well on Ruby 1.8.6 and 1.9.1 Ruby versions on Windows with old ActiveRecords, but I hope rails adapter will be ported to ActiveRecords 3 soon.

Friend of mine criticised me why I didn’t put some screenshots on project page. Since I couldn’t find a way to do it there I’ve decided to put some screenshots here, so here they are.

I like to listen various Internet radio stations while I work. My favourite application was streamtuner mostly because it had support for SHOUTcast. Unluckily new SHOUTcast design cannot be handled by streamtuner and future development on this application is stopped. In other words there is a little chance for streamtuner to be upgraded so I’ve decided to make my own streaming directories browser – RStreamTuner.

As usual developing in Ruby was real pleasure and only doubt was what to use for GUI part. I’ve started with Ruby/Tk and left it very quickly due to very poor documentation (and probably lack of Tcl knowledge). I had to find replacement and my primary goal was to find library that works equally well on Linux and Windows. WxWidgets with WxRuby gem was my final choice. Although RStreamTuner is still in early development phase it is quite usable. At the moment it supports SHOUTcast and Xiph directories, but I plan to add more as soon as I find good one. Good candidate could be Live365 but they’ve made their pages quite hard to parse so I’ll probably leave it for later.

Once I reach beta version, making support for new streaming directories will be quite easy and will require implementation of just a few methods. As a matter of act it is already so simple but I’m not satisfied with the code and will change it once I find a little bit more time. Code is quite rough (I had little time to make RStreamTuner and was eager to get working version as soon as possible). Besides I want to write detailed instructions how to add support for new directories. Anyway if you like to listen Internet radion stations and like Ruby clone RStreamTuner. Of course if you have suggestions do not hasitate to drop me a note and if you make support for new streaming directory just send me a pull request and I’ll try to grab it as soon as possible.

Unfortunately author of Rip has decided to completely leave MS Windows aside. For details you can see this thread

When I first found a post about Rip I was eager to try it. Since most of the time I work on MS Windows platform it was natural to try it there, but Rip concept is Linux oriented and it couldn’t be installed and used on Windows. Going back to the old, good Linux gave me a feeling what Rip is and how it works, and I’ve decided to try to fix it as soon as I find some spare time. Additional motivation was a fact that Ruby installer for windows RC1 is about to be released and that’s one more project that I participate in, from time to time.

So here it is. If you are willing to try Rip on Windows just download gem and install it. Rip is in development alpha state, therefore this Windows version can be marked as development pre-alpha.

Although Rip environments can be created, deleted and gems can be installed, there are features that do not work. Complete Windows specific code is on my fork on Github. A reason why I wanted to put this version available so early is mostly the same as you can see on the Rip homepage – need for suggestions, ideas, patches and tests.

Here comes short explanation of Windows specific part. Rip on Linux uses symbolic link that points to the active environment. Unfortunately such a feature is not available on Windows. Although on Vista and Windows 7 mklink can be used, this command requires administrative privileges. It can be disabled through Local Security Policy settings, but that would be too much work just to get Rip working on Windows. Therefore I’ve decided to alter environment variables when Rip environment is created or changed.

This approach has its own drawbacks. If user has more command prompts opened at the same time and Rip environment is changed in one of them, this change will not be visible in others. Cause of this is in the way command prompt is implemented. Namely it doesn’t handle WM_SETTINGCHANGE broadcast message. I really do not understand why MS developers didn’t implement this.

Main consequence of such command prompt implementation on Windows is that Rip has to change environment variables in two stages. First stage changes environment for command prompt where Rip command is executed. This change is done by calling batch file where variables are set. The second stage changes values in the registry. Variables are set in a user part of registry (HKEY_CURRENT_USER\Environment) and broadcast message is sent so other applications (such as Explorer) can update their environment which causes newly opened command prompts to pick up new environment.

At the end what are problems you could expect if you install Rip. I haven’t tested all Rip commands mostly because I was focused on basic functionality – Rip installation, creating and switching between Rip environments and installing packages in active one. Most likely some of commands will fail. In addition gems that have to build shared libraries will fail even if you have RubyInstaller DevKit installed. Moreover even if gem has binaries built for mingw32, Rip will invoke build procedure and will report that gem is not correctly installed and you will not be able to uninstall it from Rip environment later. Actually such gem is installed and can be used. One such example is Luis Lavena’s win32console gem. Even if build fails and Rip reports that gem is not correctly installed all necessary files are actually present in Rip environment and you will be able to run tests.

I’m still trying to find out how to determine whether gem needs to be built or it already has binaries. As soon as I find out this I’ll fix this part too. That’s all for now and happy Ripping.

At the and of May 2009 Savannah disks crashed. You can read more about it here. This is something no one can predict. You never know when your hardware will let you down, and there is no cure for that. Disks will be replaced, system reinstalled, backups restored and that’s it. Or is it?

Not completely if you are using CVS or SVN and there were some commits after the last backup. Emacs developers spent few days in the discussion to determine what commits they lost.

Luckily they had Git mirror of CVS repository but what surprises me is they didn’t move to Git yet. If they were using Git they would have been able to restore complete repository in a few minutes (maybe longer for large repositories). Since they already have Git mirror for Emacs code I really do not understand why do they not switch to Git completely. Having full history at each developer’s computer is a huge benefit. Not to mention that Git gives you possibilities you can only dream of in CVS and SVN.

Recently I had to process arrays of data with more than thousand of members but requirement was not to process them one by one. Instead I had to split array into chunks and to process each of sub-arrays separately. I’ve found one definition of chunk method within comments on this post and I’m including it here. This method splits array in given number of sub-arrays and I needed a version that will split array in the sub-arrays that have predefined size (ok, I know I could simply calculate a number of chunks by dividing total number of elements by required number of elements, but I just wanted one method that’ll do all that for me). So I wrote chunk_max_num method. If maximum number of elements per array is zero or negative it will return original array and for all other cases it will return array which has arrays of given size as members.

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

require 'enumerator'

class Array
  def chunk_max_num(max_num)
    return slice(0...length) unless max_num > 0
    quot, mod = length.divmod(max_num)
    quot += 1 if mod > 0
    (0..quot).map {|i| i*max_num}.enum_cons(2).map {|a,b| slice(a...b)}
  end
 
  def chunk(pieces)
    q,r = length.divmod(pieces)
    (0..pieces).map {|i| i * q + [r,i].min}.enum_cons(2).map {|a,b| slice(a...b)}
  end
end

First step

In my previous articles I wrote a lot about Emacs, its customization and adjusting it for Ruby and Ruby on Rails development. Although these articles give detailed instructions how to set everything up, following all steps are somewhat cumbersome and require quite a lot of work.

Fortunately there is a solution. Just clone Emacs starter kit from Github and you will be able to develop Ruby and RoR applications in Emacs in just a few simple steps. Emacs starter kit uses Elpa packaging and is very easy to configure. But let’s start from the beginning.

Clone Emacs starter kit with Git, to your empty .emacs.d directory. Be sure that init.el file is in .emacs.d folder. If you do not use Git (do you realize what you are missing?) you can get archived sources if you press download button on Emacs starter kit page on Github. Unpack archive to the .emacs.d and be sure that init.el is in .emacs.d folder.

Emacs starter kit has some packages that are still not in Elpa but you’ll have to install additional packages required for Ruby and RoR. Let’s do it now:

1

M-x package-list-packages

New buffer with list of available packages will open and you can select those you want to install. If you plan to use Emacs for Ruby and RoR you’ll definitely need Rinari. It will install all dependent packages.

There are few other that might come in handy like css-mode, javascript or yasnippet but you can always add them later. If you install yasnippet you should probably need yasnippets-rails, but more about that later. Installing packages is very easy. Just place cursor on the package line and press ‘i’. When you are finished with selection press ‘x’ and packages will be installed. At any time you can press ‘h’ to get quick help for Elpa packaging system. In order to uninstall package you have to press ‘d’ (and ‘x’ after that).

Further customization

When you are done with packages installation just restart Emacs and you are ready to continue work in your favorite language – Ruby and framework – Ruby on Rails :) But what if you want to adjust some settings or keep some of those you set according to my previous articles? Luckily that is easy to do, too.

First create sub-directory in .emacs.d directory with your user name on the system you are using. Put your .el files there and Emacs starter kit will load them automatically during Emacs start up. If you have changed color theme and you’ve used theme from color-theme library put color-theme-library.el in this folder and add, for example, line

1

(color-theme-deep-blue)

to any of .el files in this folder. I keep all my additional settings in customization.el file.

I already mentioned yasnippet and yasnippets-rails packages. Simplest way to install yasnippet library is from Elpa system following above mentioned procedure. Unfortunately yasnippets-rails cannot be installed that way so you should clone it from the Github:

1

git clone git://github.com/eschulte/yasnippets-rails.git

Directory yasnippets-rails should be in the directory with your user name under .emacs.d. Add following lines to your customization.el file:

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

(add-to-list 'load-path 
   (concat dotfiles-dir "/<user-name>/yasnippets-rails"))

(add-hook 'ruby-mode-hook ; or rails-minor-mode-hook ?
          '(lambda ()
             (make-variable-buffer-local 'yas/trigger-key)
             (setq yas/trigger-key [tab])))

(require 'yasnippet)
(add-to-list 'yas/extra-mode-hooks
             'ruby-mode-hook)

(yas/initialize)
(setq yas/window-system-popup-function 'yas/x-popup-menu-for-template)
(yas/load-directory (concat dotfiles-dir "/<user-name>/yasnippet/snippets"))

(yas/load-directory 
   (concat 
      dotfiles-dir "/<user-name>/yasnippets-rails/rails-snippets/"))

(make-variable-buffer-local 'yas/trigger-key)

and you will be ready for using new snippets during development.

Final word

My previous articles were based on Rails reloaded package for RoR development in Emacs. It is good library and if you like fancy GUI things you can still use it. There is a new version on Github.

Still, I would recommend trying Rinari that is a package from Elpa. It doesn’t have any GUI features but once your fingers “learn” all shortcuts your RoR development will be much easier and faster.

It has been a few months since my last post on the old blog powered by Wordpress. All this time I was searching for a blog engine made in Ruby on Rails. It was not easy to decide what to use. I’ve tested Mephisto, Typo, Radiant, Simplelog and several others I cannot remember now. All of them really have a lot of features, but for my needs they are way too complex. What I was searching for is some simple application that I can easily tweak and adjust to my needs. I was about to move to Mephisto or Simplelog when I found Enki.

As soon as I’ve looked into the code I knew that it is the one. Simple, lightweight blogging engine without fancy stuff that I’ll rarely use, yet with all features I need for my own blog. Moreover Enki’s source is on Github and it was easy to make a fork and start to work on my own version. So I did it. I must say it can be used just as it is, but I wanted to give it a new “look and feel”, so I’ve started with changes on a style sheet and didn’t stop there.

Enki handles comments in a very flexible way. In order to post a comment only name and text of a comment must be filled. On the other hand, if commenter wants to give a little bit more data he can use OpenID authentication. Nowadays with all those spammers around, this was not good enough for me. I didn’t want blog overloaded with bunch of stupid messages so I’ve decided to change comments handling a little bit.

There are lot of ways to protect a blog from spam. You can forbid posting comment without authorization. That way everyone must register before he is allowed to post a comment. That’s too much work for commenter and, frankly speaking, for me too. I wanted to make commenting as simple as possible, still under decent level of control. In order to accomplish it I could use reCAPTCHA or Akismet. Both of them are quite good, but I’ve decided to use Akismet. It leaves commenter same level of simplicity as in the original Enki source but every comment must pass spam filter. So I’ve dropped OpenID authentication, downloaded akismet.rb, adjusted comments controller and fixed tests. It was quite simple and easy.

And here it is up and running.

Finally few thoughts about Enki. Although it can be used out of the box, I think very few users will actually install it that way. But if you are RoR fan, and you like to work in Ruby and Ruby on Rails this is the right choice for you. Code is readable and easy to change. It is not over-engineered except in one part – gems. Enki requires cucumber which, on the other hand, requires rspec, rspec-rails, webrat and, optionally, term-ansicolor, nokogiri, builder and few others. For my mind it is a little overhead for such a simple project, but it is definitely neglectable comparing to other good sides of Enki. Lack of support for images uploading might be a limitation in Enki’s usability, but author wanted to keep it simple so this shouldn’t stop you from using it. At the end if you like Ruby on Rails it will be a fun implementing support for it, isn’t it? Go ahead, grab attachment_fu or paperclip and do it.