1 2 3 4 5 6 7 8 | |
I am confused. Why would you do this? What possible benefit does this have over just using an if/else/elsif block? So… I’m going to just change it to:
1 2 3 4 5 6 7 | |
Because that seems, you know, sensible…
]]>As mentioned in a prior post, I am doing some messing around with SOAP. As it happens, I am moving from my development area to try this in a server machine. No problem.
Note: I am going to include every hiccup I ran into, even those that aren’t the fault of the Ruby ecosystem, just to be thorough. I am just tracing the steps that a complete newb would.
This is a fresh, minimal install of Ubuntu, so I go to the RVM site, copy the install command and paste it into my terminal:
1 2 3 | |
Hitch #1: Curl isn’t installed on this machine.
Solution: sudo apt-get install curl
Next try:
1 2 3 4 5 | |
Hitch #2: Git isn’t installed on this machine.
Solution: sudo apt-get install git-core
Note: I did a quick check of the RVM docs, and it doesn’t seem to mention anywhere that git/curl may not be available and what to do about it if they aren’t.
1 2 3 4 5 6 7 8 9 | |
Success! this point RVM installs cleanly, and after logging in/out (or source ~/.bashrc), RVM is ready to go. However, it is of note that the WALL OF TEXT that RVM spits out is exceptionally daunting. It would be nice if the last thing shown was generally relevant (like, hey, you need to restart your shell), not here’s what you need for IronRuby. I know, I know, RTFM, but… it’s a thing.
I need ruby 1.9.2, so let’s grab that.
1 2 3 4 5 6 7 | |
Great! Mission accomplished, everything built cleanly.
Gemsets are good practice, so let’s make one:
1 2 3 4 | |
Awesome. Again, everything going smoothly (yay we <3 ruby!) It’s worth mentioning here that at this point I am getting warm and fuzzy feelings about Ruby+RVM. This is how it should be.
Last step before pulling out the editor, get the gems:
1 2 3 4 5 | |
Wait, WTF? There haven’t been any errors or anything, what is this?
Hitch #3: Mysterious “no such file to load – zlib” ERROR.
Solution: At this point the user will need to google. If they have good google-fu, they will find the stack-overflow solution (there are other ways to solve this, but the rvm one is the cleanest).
Ok, so, a minor hitch, just add a lib, change the rvm ruby install command a bit, and get caught up again.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 | |
Rockin. Minor misstep, let’s keep on rollin’.
Second time’s the charm! You’d think…
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 | |
Ok, missing libxml2. Again, would have been nice to know this earlier… but this is more like the git/curl situation, it’s an OS thing.
Hitch #4: Nokogiri gem has a compilation error.
Solution: After googling around, the nokogiri docs say that on ubuntu, you need a couple more libs: sudo apt-get install libxslt-dev libxml2-dev
After installing the libs, the gems finally install cleanly:
1 2 3 4 5 6 7 | |
Great! Gems are installed, everything is ready to rock n’ roll.
So, the code is super trivial (gist if you want to see), let’s run it:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 | |
Can I get another WTF? Everything installed clean! Ruby, RVM, Gems, everything… why is this breaking?
Hitch #5: Running the code throws a “no such file to load – openssl” error.
Solution: Google again to the rescue; more specifically stack-overflow to the rescue. This solution is similar to the first: use rvm to add the missing libs to ruby. Awesome, we get to re-compile again.
We’ve had some practice, so let’s do it again, though the stack-overflow said to install iconv in addition to openssl. Since we won’t get any information in advance as to whether or not iconv is required, let’s just do it now. Also, don’t forget the original zlib compilation that we did!
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 | |
And finally, it all works.
The upshot is that while this is annoying for an experienced user, it could represent a brick wall for a new user. This should be a straightforward and easy process. And I get the feeling that if I had removed the RVM variable, it would have made the problem even harder to solve because recompiling is harder when you have installed from apt.
I love ruby, but this is kind of embarrassing. What can we do to make this better?
]]>Unfortunately, it is a SOAP API, something which I know nothing about. But, hey, how hard can it be?
My plan is: learn and prototype with ruby, and then fool around a little with PHP (to make sure it works), then make a bid to do the work (and include my already spent time :P ).
My first stop for a library search is, of course, the excellent Ruby Toolbox. They list three options:
The searching also indicated that SOAP4R was the original ruby SOAP library, but that it also is slow, old, and nobody likes it. I decided to try first with Savon because it was first.
Ruby:
PHP:
Base64.encode64() method appends an \n to the encoded string so you have to encode64(string).strip or Mr. SOAP API will throw a 400. Anyway, our access is auth’ed via IP, so this doesn’t matter and I never did get to see what a successful user/pass auth looked like.I hate to say it but to be honest, there has never been a better time to not be in the process of getting a computer science degree. I work in academia, so this is a really tough thing to admit, but I think it’s true. It’s not just the spiraling cost of education, and the cost/benefit analysis of tuition. The whole situation is changing.
The biggest game-changer, I think, is that there is a ridiculous amount of opportunity out there for free. It used to be that you had to go to university for the opportunity to get practice doing the work. This isn’t the case anymore. Employers are still mostly interested in examples of your work (basically, a portfolio of code), but with open source being what it is, the world is overflowing with public projects to work on. If you spent four years working food service and spending your off time coding on open source projects (with the same amount of time that you would have spent schooling), you would end up better off. Probably wildly better off. Many CS grads enter the job market with nothing but credentials and examples of canned homework projects. Sure, they know the basics of algorithms, but that is just a small part of what software engineering is.
Self-taught coders are often preferred over graduates because they have demonstrated at least one critical skill: the ability to independently find solutions to get things done. If they have worked on an active open source project, then they have another critical coveted skill: the ability to collaboratively work on software projects with other developers.
And to add to that, MIT open courseware is freely available, a curious student has access to world-class educational tools if they need it. More and more institutions are making their materials available, and Khan Academy just released its first set of computer science training classes for free. The entire landscape is changing.
During the recent EDUCAUSE conference, I heard this small-but-growing group of self-taught students “edupunks”. The presentation was posted online and I think it’s worth a look if you are interested.
I think the future of education is changing. I still think academia is important, but the focus needs to change. I don’t think that it will make sense to go to university for classes, but for hands-on training and research opportunities. University is where mentors and students gather and work. Live lectures are a waste of time, they should be just video. In the classroom you should be having real-world work experiences with teachers and peers. This transition will be painful and ugly, but I think it’s inevitable; the current system isn’t sustainable. Learning is about practice and mentorship. That’s what tuition should buy you.
]]>So if you want to use mysql, you do:
1
| |
There are other dbd’s for different databases.
Also: If you are getting this error:
1 2 3 4 5 6 7 | |
Then you need to do this:
1
| |
The last argument may need to be modified to match the “Referenced from” line from your error.
Test the fix by opening up irb and doing a require ‘mysql’
]]>The recipe is going to be used to create a puppet script, but it could also be boiled down to a ruby or shell script to easily create app instances for anyone’s use.
If apache2 is not yet configured with mod_rails, do the following:
/etc/apache2/mods-available/mod_rails.conf and put the following in it:1 2 3 4 | |
ln -s /etc/apache2/mods-available/mod_rails.conf /etc/apache2/mods-enabled/mod_rails.confOptionally make Ubuntu’s username policy more sensible (allow underscores)
1
| |
1
| |
1
| |
1
| |
1
| |
1
| |
1
| |
1
| |
1 2 | |
1
| |
home/<username>/rails directoryAt this point the application will be available running at http://hostname/<username>
With one of the deploy.rb files, a savvy rails user can deploy their application to the server with capistrano
A less savvy rails user can connect via ssh and checkout their their rails app to someplace in their home, then symlink it to the ~/rails/app directory
On windows, my password manager of choice is pwsafe, but it only offers command-line functionality for mac (and presumably linux, though I’m not sure). On the Mac, I quite like Keychain, it is tightly integrated with the OS, but the passwords are locked away somewhere. Supposedly you can leverage the @security@ executable and gain access to it, but then importing/exporting would be a total pain in the ass.
My searches had turned up an app called “Password Gorilla”, but it kind of sucked, there appeared to be nothing for me!
Then I ran into KeePassX, a password management package with binaries for all of the major OS’s. I installed it onto the mac: perfect. It’s exactly what I wanted. There is some wonky UI issues (it forgets which password db you are using when you lock it), but it is totally workable.
I plan on using FUSE to create an ssh directory to one of my servers and store the binaries and password database on them. So far so good!
]]>PRIVVMPAGES="150000:300000"vzctl enter <the new machine>1 2 3 4 | |
1
| |
passenger-install-apache2-module and follow the instructionsafter starting apache2, i was able to see my site up and running. yay!
]]>1 2 3 4 5 6 7 8 9 10 11 | |
First, a recap. I want a system that can authenticate a user from multiple sources. The sources will be organized into a priority list. So, for example, if User1 asks for a page that is restricted, first I check to see if they have a pubcookie or known CAS/SSI(Central Authentication Server/Single Sign On) cookie. If so, I check to see if they have a user in the system. If so, then I log them in. If not, then I create one (noting the authn source) and log them in. If they don’t have a cookie, then I send them to a login page. With their username and password credentials, I try to log them into the list of sources. If there is only one username that matches that user in my system, I try to log them into that first, otherwise I progress through the list. If a remote authentication server verifies their username/email/password as a valid user, i check for a local user. If it exists, I log them in, if not i create a local user, note their authn source, and log them in.
The ramifications of this is that there are allowed to be users with the same username. However, the tuple of username, password and authn source is unique. there will only be one Foo with the password Blarg which is authenticated with my company LDAP server. There might be a local user with the username Foo, but I’m assuming that everyone will have different passwords. If someone did use the same username and password as another person, and they used a higher priority authn service, they would always be the person that would log in as that user.
This is not ideal, I will think on it further.
]]>Ok, so I have been bitten several times when I do something like:
1
| |
Why is my username blank? This is a simple example, but I have run into this a number of times. It’s like a missing semicolon problem, but ERB instead of PHP or C. Strangely enough, even if you specify a puts method, you still get no output:
1
| |
(This also will give you nothing)
You need a <%= opening in order to print out anything. So, it would be nice if Textmate colored them differently. I went on a quest to make this happen… and failed.
Ultimately I ended up in the #textmate channel on freenode. They were very helpful and pointed me to this replacement grammar for textmate, which for the sake of being complete, I’ve included Right here:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 | |
So, here’s how to do it:
And that’s it! Yay for solutions!
]]>By shortcuts, I mean things like this:
1 2 3 4 5 6 7 8 9 | |
As you can see, it went ahead and created the model, test, fixture and migration. The column definitions were stuffed auto-magically into the migration.
This is really cool, but it has some limitations. If you are doing anything more complicated than it won’t work for you. For example, if we wanted to do a many-to-many relationship with a toppings table, you would have to set that up manually (as far as I know).
The confusion that I referred to earlier has to do with generating a migration. Let’s say that I already created my topping model manually by creating an app/model/topping.rb file. Then it occurs to me that I also need a migration, so I run the following:
1
| |
Unfortunately, this will create the migration file, but the column declarations are ignored. This is confusing, given the following behavior:
Let’s say, instead, I created the toppings model with the following command:
1
| |
This would have generated an empty ###_create_toppings.rb migration. Then I decide to add the columns to it like so:
1
| |
Rails does some magic and infers from the migration name that we are adding columns to the topping table. But why didn’t it add columns in the earlier example? My only explanation is that migration generation is for updating already existing models/tables, it’s not intended for model/table creation. Arguably, what good is a created table without an associated model? For simple cases, this holds true. For a many-to-many situation, where you need to define a join table, this kind of shortcutting will only take you so far.
]]>This is posted in other places, but I want to make sure that it doesn’t get lost to the wilds of the interwebs.
The default black background themes in Terminal suck if you are doing ANSI colors in the terminal. I like colors in the term, but the dark-blue is aneurysm inducing. So, after futzing with the defaults a bunch, I decided to see if the internet had any solutions. By his noodlyness, they did.
So, here’s what needs doing:
~/Library/Application Support/SIMBL/Plugins directoryI am going to try to make a terminal theme that matches my favorite TextMate theme, Succulent (screenshot, download)
]]>When it has you create your database, it has you run the following sql code:
1 2 3 4 5 6 7 8 | |
What I don’t get is: how is that any better/easier than giving a quick run through the migration process? The above code could just as easily been replaced with:
After creating your database, and setting your database.yml connection variables, we will be building a table to hod our To Do list categories. We will need a model definition for a Category, and using a generator, we get a place to define both the database table for the Categories and the model for a Category.
1 2 3 4 5 6 7 8 9 | |
Edit the migration so that it looks like this:
1 2 3 4 5 6 7 8 9 10 11 12 13 | |
Then run rake db:migrate and you have your table and you are now ready to edit the model.
I don’t see this as being worse than the SQL. Why isn’t it presented in this more “railsy” way?
]]>There are a bunch of tutorials online (like this one) about how to publish your mercurial repos. But if you are mod_rewriting all over the place, you need a couple extra steps.
I had to add this directive at the beginning of my mod_rewrite declarations to ignore the hg directory:
1
| |
My apache conf blok looks like this (not dissimilar to the examples):
1 2 3 4 5 6 7 8 9 10 11 12 13 | |
So, it’s working now, and that’s lovely. :) Now, i just wish I could tie the htaccess to the Redmine user tables and use that to Auth against… Another day, another project…
]]>Here’s what I did:
I had to install two packages:
1
| |
Then I was able to install the mysql gem with the config parameter:
1
| |
You do have to be root (or use sudo) to run those commands.
Then I restarted the mongrel cluster, and it’s been running well so far.
Hopefully the problem is fixed!
]]>.hgignore file), it worked just fine.
I am, however, having trouble getting apache2+mercurial to work happily. For some reason, the hgwebdir.cgi script isn’t being run, it’s just spitting out the text.
To add to the problems, I tried running the script from the command-line (it’s just a python script) and it spewed a bunch of errors (I think the problem is a lack of wsgi libraries, but I’m not sure). If I do get it working, I’ll mention what I did here.
About the Mephisto/mongrel cluster problem, it’s still hanging every few hours. I am at a loss as to why, and I don’t want to just install things until it works. I like my VMs to be lean, that’s the whole point. I tried poking the rails IRC channel about it, but didn’t get any takers.
UPDATE: I finish this up here.
]]>Nginx seems to be motoring along just fine with no problems.
I am going to implement a change recommended by this google code post. The fix is to re-install the mysql gem with this option:
1
| |
However, I don’t have mysql installed on my virtual machine, since I have a centralized mysql server that I use for all of my virtual machines. As a result, I don’t have a mysql config to point to. So, I installed just the mysql client with the following command:
1
| |
This added 23M of extra stuff to the VM, but it also threw a my.cnf into the /etc/mysql dir.
The next step was installing the mysql gem. The hitch here is that I don’t have a mysql server or dev stuff installed, so the gem doesn’t want to compile. I’m also curious how active record can work at all without the mysql gem installed… I’m going to ask around for some more information and post it here when I’ve figure out where to proceed.
]]>The ideal system would have the following features:
There’s already an acl_system2 plugin out there that does some of this stuff. It does authentication against the local database, and controller-method level permissions. But it doesn’t have a concept of “owner”. So you couldn’t easily set an object to be “editable by the owner or an admin”. It doesn’t let you set permissions on the model level. And it doesn’t let you use any kind of GUI to change the permissions, it’s all hard-coded.
Next up is restful_authentication, it does just authn, and it doesn’t auth against any external sources. That doesn’t mean it couldn’t, but it would be nice if it did. Perhaps it makes sense to use it as just the local authn mechanism, and roll other remote authn mechanisms separately and then put them all under an authn system. Using the role_requirement plugin, you can also get roles. But it is still statically defined in the code, there’s no UI for updating the role permissions during program runtime.
Also, Redmine has a pretty good authn/authz system that will authorize against an LDAP directory. The way that it’s organized, it looks like adding more Authz sources is possible. However it only allows one Role per Person per Project. That’s not good enough for what I want. I want a many to many user/role relationship. It also doesn’t have groups or an importer. That is a problem for us, as we constantly want to add people to projects when they haven’t yet logged in, and so we have to tell them: log in, then tell us, then we can add you. An importer from the LDAP server would be ideal.
Finally, there’s a cool writeup on implementing restful authentication with all the bells and whistles. It’s a great tutorial for getting authn/authz (using OpenID instead of LDAP) running. However, it’s not very modular and still has the the problem of grabbing users form the Authn source (that wouldn’t have been possible with OpenID anyway, but could have been possible with LDAP).
I’m going to make what I want, and then I’ll stick it somewhere for the world to grab it if they want.
]]>