The ideal RoR environment

rubyonrails-talk
1

I'm currently taking some time to re-examine my environment and wanted to get the input from the minds that read these forums in an attempt to piece together an ideal RoR environment. I'm trying to break out each necessary aspect of a project and get thoughts and opinions on what people think is the best tool for the function. For the sake of discussion, I'd like to assume a Linux/UNIX based environment but I do have a couple of projects running in Windows if anybody has any thoughts in that department. If I've made any glaring omissions...please tell me.

My thanks for any feedback you can offer.

Testing - Cucumber, RSpec, and ApacheBenchmark. Using Cucumber for the high-level tests and RSpec for low-level ones and AB for load- testing. I was one of the people that learned Ruby and Rails and ignored the testing features for a long time but I'm on board now and believe this to be a good testing suite.

Fixture-Creation - I am unfamiliar with this outside of rails fixtures. I have been told that they aren't the best way to go and that I should look into Factory Girl or Machinist as a replacement.

Version-Control - I currently use SVN, but would be interested in hearing arguments for learning Git.

Deployment - Capistrano...is there anything better?

WebServer - I currently use Apache & Mongrel. I've heard/read good things about Passenger and Nginx though. Is it worth switching to a Passenger / Nginx solution or maybe some hybrid?

Document-Generation - I have used both pdf-writer and prawn for PDFs in the past. I have some windows projects that have generated word and excel docs via win32ole. I spent a minimal amount of time trying to generate OpenOffice docs but never had the proper time to devote to getting it working.

Graphing - I have used Gruff in the past. I also played around with ZiYa and found it to be an intriguing possibility.

Thanks again for any thoughts. Now that I've typed it all out I'm thinking maybe this should've been separated into separate posts for clarity but as a start I hope this will do.

Bob

2

Testing - Cucumber, RSpec, and ApacheBenchmark. Using Cucumber for the high-level tests and RSpec for low-level ones and AB for load- testing. I was one of the people that learned Ruby and Rails and ignored the testing features for a long time but I’m on board now and believe this to be a good testing suite.

I agree, we use RSpec for unit testing, Cucumber + Selenium for integration tests (because our app uses a huge amount of client side javascript)

Fixture-Creation - I am unfamiliar with this outside of rails fixtures. I have been told that they aren’t the best way to go and that I should look into Factory Girl or Machinist as a replacement.

Factory Girl is nice, that’s what we use. Machinist is certainly just as good. It all depends on what syntax you prefer. They’re certainly both better than Rails fixtures.

Version-Control - I currently use SVN, but would be interested in hearing arguments for learning Git.

Git has several advantages over SVN. Since it’s a distributed version control system, you can just commit locally (even when not connected to the net), then push the changes to a central server. Where Git excels over SVN imo is the branching. In SVN branching and merging was a living hell, git just makes it something you use all the time. It’s hard to explain why git is so much better, but once you’ve experienced a few of subversion’s annoying quirks and tried out git, you’ll never look back. Use either github or gitosis (on your own server) for repo management.

Deployment - Capistrano…is there anything better?

Capistrano has never failed us. There’s a few alternatives like Vlad The Deployer and Fabric, but I see no reason to switch from Capistrano.

WebServer - I currently use Apache & Mongrel. I’ve heard/read good things about Passenger and Nginx though. Is it worth switching to a Passenger / Nginx solution or maybe some hybrid?

You could also use Apache+Passenger. Nginx is less memory hungry than Apache. What’s really nice about Passenger is the smart spawning and you have one (or more if you’re using a mongrel cluster) process and possible point of failure to take into account.

Document-Generation - I have used both pdf-writer and prawn for PDFs in the past. I have some windows projects that have generated word and excel docs via win32ole. I spent a minimal amount of time trying to generate OpenOffice docs but never had the proper time to devote to getting it working.

We use PrinceXML for PDF generation, but it’s expensive. I know someone who has succesfully created Word documents with OpenOffice, but haven’t got any experience doing so myself.

Best regards

Peter De Berdt

3

SVN is a nightmare when working in teams, remotely, or trying to be agile. You can *make* it work, but only because you change your behaviour to work the path of least resistance in SVN - not ideal AFAIC.

Git is better, Mercurial may be better still (a little more flexability than Git, but not as wide support); or if you want to be really cutting edge, look at Fossil (http://fossil-scm.org). Check out www.hginit.com for a very good "from SVN to Mercurial" blog/tutorial - but the principles are very similar for Git.

4

Bob wrote:

I'm currently taking some time to re-examine my environment and wanted to get the input from the minds that read these forums in an attempt to piece together an ideal RoR environment. I'm trying to break out each necessary aspect of a project and get thoughts and opinions on what people think is the best tool for the function. For the sake of discussion, I'd like to assume a Linux/UNIX based environment but I do have a couple of projects running in Windows if anybody has any thoughts in that department.

My thoughts: don't use Windows for *anything*. It's a bad server OS. It's a bad desktop OS. Avoid.

My OS preferences: Mac OS X for development, Ubuntu Linux for server. They are the most pleasant and capable in each case IMHO.

If I've made any glaring omissions...please tell me.

My thanks for any feedback you can offer.

Testing - Cucumber, RSpec, and ApacheBenchmark. Using Cucumber for the high-level tests and RSpec for low-level ones and AB for load- testing. I was one of the people that learned Ruby and Rails and ignored the testing features for a long time but I'm on board now and believe this to be a good testing suite.

I've never used ab. I highly, highly recommend Cucumber and RSpec. And of course, it goes without saying that all development is done story-first: write the Cucumber story, watch it fail, write the RSpec specs of what you need to implement the story, watch them fail, make them pass.

Fixture-Creation - I am unfamiliar with this outside of rails fixtures. I have been told that they aren't the best way to go and that I should look into Factory Girl or Machinist as a replacement.

Fixtures are broken. They should be removed from Rails. Under no circumstances should any Rails developer ever attempt to use them. Machinist is the way to go here, with Factory Girl a reasonable second choice (it's a bit more cumbersome).

Version-Control - I currently use SVN, but would be interested in hearing arguments for learning Git.

Do yourself a favor and get rid of Subversion today. Subversion is about the best centralized version control system that I'm aware of, but centralized version control is a paradigm with many problems. No project in 2010 should be without distributed version control.

Deployment - Capistrano...is there anything better?

Not that I've heard of.

WebServer - I currently use Apache & Mongrel. I've heard/read good things about Passenger and Nginx though. Is it worth switching to a Passenger / Nginx solution or maybe some hybrid?

Why are you torturing yourself with Apache and Mongrel? Passenger (and Ruby EE) is absolutely the right thing to use. It makes Rails deployment as easy as PHP deployment. Nginx is also probably a good choice.

Document-Generation - I have used both pdf-writer and prawn for PDFs in the past. I have some windows projects that have generated word and excel docs via win32ole. I spent a minimal amount of time trying to generate OpenOffice docs but never had the proper time to devote to getting it working.

Rails is a Web framework. Word and Excel documents have no place on the Web -- they do not work well as interchange formats or play nicely with Web browsers. Do not design your Web applications in such a way that Word and Excel documents are generated -- that's just irresponsible.

I use Prawn for PDF generation. At some point I plan to look at Flying Saucer and wkhtmltopdf.

Graphing - I have used Gruff in the past. I also played around with ZiYa and found it to be an intriguing possibility.

No experience here. I'd probably go with Google Charts/Visualizations up to the point where it stopped suiting my needs.

Thanks again for any thoughts. Now that I've typed it all out I'm thinking maybe this should've been separated into separate posts for clarity but as a start I hope this will do.

Bob

Best,

5

Bob wrote:

I'm currently taking some time to re-examine my environment and wanted to get the input from the minds that read these forums in an attempt to piece together an ideal RoR environment. I'm trying to break out each necessary aspect of a project and get thoughts and opinions on what people think is the best tool for the function. For the sake of discussion, I'd like to assume a Linux/UNIX based environment but I do have a couple of projects running in Windows if anybody has any thoughts in that department. If I've made any glaring omissions...please tell me.

Oh, one thing that you didn't mention: database. If you are using MySQL, I strongly recommend getting rid of it and switching to PostgreSQL, which is a much more capable database system. And whatever DBMS you are using, please respect the database's intelligence by giving it the tools to help you out. That means using foreign key constraints (with the foreigner gem), making proper use of aggregate functions, indices, and bulk queries.

Best,

6

Tsch!

Rails is a Ruby framework - and it can be used to make web applications, or non-web intranet applications, or a mix of whatever a client asks. If a client asks to produce the functionality to export data into the office-application format of their choice, then it's not irresponsible to do exactly that. (at a preference, I output PDFs and CSVs, but if I *had* to make a Word doc - and I can imagine situations it would be asked for - then so be it)

7

Michael Pavling wrote:

Version-Control - I currently use SVN, but would be interested in hearing arguments for learning Git.

SVN is a nightmare when working in teams, remotely, or trying to be agile. You can *make* it work, but only because you change your behaviour to work the path of least resistance in SVN - not ideal AFAIC.

Git is better, Mercurial may be better still (a little more flexability than Git, but not as wide support);

Really? I thought Mercurial's alleged simplicity (such as it is -- I couldn't understand it when I tried it) was achieved by its being *less* flexible than Git and so having fewer options to deal with.

or if you want to be really cutting edge, look at Fossil (http://fossil-scm.org).

How is Fossil cutting-edge? It just looks like "DVCS for dummies", or like a project-management system (with wiki and tracker) that decided to implement its own VCS. I tend to think any DVCS whose stated goal is to avoid excessive branching is missing the point.

Check out www.hginit.com for a very good "from SVN to Mercurial" blog/tutorial - but the principles are very similar for Git.

Best,

8

Michael Pavling wrote:

Rails is a Web framework. �Word and Excel documents have no place on the Web -- they do not work well as interchange formats or play nicely with Web browsers. �Do not design your Web applications in such a way that Word and Excel documents are generated -- that's just irresponsible.

Tsch!

Rails is a Ruby framework - and it can be used to make web applications, or non-web intranet applications,

An Intranet app is still a Web application by my definition. It works over HTTP and is viewed through a client's Web browser. It works the same way as a public Web application. Whether it can be seen by people outside the company or not is irrelevant in this regard.

or a mix of whatever a client asks. If a client asks to produce the functionality to export data into the office-application format of their choice, then it's not irresponsible to do exactly that.

Yes it is. If a client asks me that, then it is my responsibility as an ethical, responsible developer to tell them that IMHO they are making a suboptimal choice, and to explore other alternatives with them.

The client gets to dictate business needs, not technical implementation.

(at a preference, I output PDFs and CSVs, but if I *had* to make a Word doc - and I can imagine situations it would be asked for - then so be it)

For me, no way, unless there were some *external* requirement (say, interfacing with a third-party system that the client had no control over, and that only accepted Word files). I won't do something that I know to be wrong.

Best,

9

Testing - Cucumber, RSpec, and ApacheBenchmark. Using Cucumber for

the high-level tests and RSpec for low-level ones and AB for load- testing. I was one of the people that learned Ruby and Rails and ignored the testing features for a long time but I’m on board now and believe this to be a good testing suite.

I agree, we use RSpec for unit testing, Cucumber + Selenium for integration tests (because our app uses a huge amount of client side javascript)

I agree.

Fixture-Creation - I am unfamiliar with this outside of rails

fixtures. I have been told that they aren’t the best way to go and that I should look into Factory Girl or Machinist as a replacement.

Factory Girl is nice, that’s what we use. Machinist is certainly just as good. It all depends on what syntax you prefer. They’re certainly both better than Rails fixtures.

I agree.

Version-Control - I currently use SVN, but would be interested in

hearing arguments for learning Git.

Git has several advantages over SVN. Since it’s a distributed version control system, you can just commit locally (even when not connected to the net), then push the changes to a central server. Where Git excels over SVN imo is the branching. In SVN branching and merging was a living hell, git just makes it something you use all the time. It’s hard to explain why git is so much better, but once you’ve experienced a few of subversion’s annoying quirks and tried out git, you’ll never look back. Use either github or gitosis (on your own server) for repo management.

I agree on Git. I for one use Github for repository management. I find it great when doing team work.

Deployment - Capistrano…is there anything better?

Capistrano has never failed us. There’s a few alternatives like Vlad The Deployer and Fabric, but I see no reason to switch from Capistrano.

Vivre Capistrano!!

WebServer - I currently use Apache & Mongrel. I’ve heard/read good

things about Passenger and Nginx though. Is it worth switching to a Passenger / Nginx solution or maybe some hybrid?

You could also use Apache+Passenger. Nginx is less memory hungry than Apache. What’s really nice about Passenger is the smart spawning and you have one (or more if you’re using a mongrel cluster) process and possible point of failure to take into account.

I use Nginx myself.

Document-Generation - I have used both pdf-writer and prawn for PDFs

in the past. I have some windows projects that have generated word and excel docs via win32ole. I spent a minimal amount of time trying to generate OpenOffice docs but never had the proper time to devote to getting it working.

We use PrinceXML for PDF generation, but it’s expensive. I know someone who has succesfully created Word documents with OpenOffice, but haven’t got any experience doing so myself.

For Prawn and Prawnto is still a superb combination!! I have ever tried OpenOffice, and it is also cool (but not as cool as Prawn/Prawnto).

10

If a client asks to produce the functionality to export data into the office-application format of their choice, then it's not irresponsible to do exactly that.

Yes it is. If a client asks me that, then it is my responsibility as an ethical, responsible developer to tell them that IMHO they are making a suboptimal choice, and to explore other alternatives with them.

For instance: They want a application (web-technology-based or otherwise) to manage their foos and bars - it's going to be internal only. They use MS Office desktop applications. They want their application to export mailshots as Word documents so they can edit them before printing.

There is no alternative to explore.

Certainly, if you think they are making bad choices, then it's right to tell them. But the final decision is theirs (and I'd recommend their decision should be to let me decide :wink:

The client gets to dictate business needs, not technical implementation.

But they're paying, so if they have a technical implementation requirement, which they stick to despite my advice to go with another choice, then I would judge it to be unethical to not do what they want. If my objection is big enough, then I can quit the job. [1]

[1] I'm currently working on some PHP development, which is running on IIS and MSSQL, because some middle-manager bought an expensive server, which has to be seen to "work". It's frustrating, but that's the situation. I work with it, or I don't. They *won't* install *nix and MySQL, etc. I really don't like it, but my mortgage has to be paid.

11

In so far as it is a moderately new development by people who have used other DVCS applications and said to themselves "we'd do it slightly differently". I don't mean to say it's necessarily any good... it may or may not be (fundamentally), and it may or may not be (for some specific situation).

I have seen some reports from people whom I respect the opinion of, which say they like it, and it is good from their PoV.

HTH

12

Michael Pavling wrote:

If a client asks to produce the functionality to export data into the office-application format of their choice, then it's not irresponsible to do exactly that.

Yes it is. �If a client asks me that, then it is my responsibility as an ethical, responsible developer to tell them that IMHO they are making a suboptimal choice, and to explore other alternatives with them.

For instance: They want a application (web-technology-based or otherwise) to manage their foos and bars - it's going to be internal only. They use MS Office desktop applications. They want their application to export mailshots as Word documents so they can edit them before printing.

There is no alternative to explore.

Certainly, if you think they are making bad choices, then it's right to tell them. But the final decision is theirs (and I'd recommend their decision should be to let me decide :wink:

Yes, the final decision is theirs. I'm perfectly happy to forgo a job that would mean putting myself in an untenable position.

The client gets to dictate business needs, not technical implementation.

But they're paying, so if they have a technical implementation requirement, which they stick to despite my advice to go with another choice, then I would judge it to be unethical to not do what they want. If my objection is big enough, then I can quit the job. [1]

[1] I'm currently working on some PHP development, which is running on IIS and MSSQL, because some middle-manager bought an expensive server, which has to be seen to "work". It's frustrating, but that's the situation. I work with it, or I don't. They *won't* install *nix and MySQL, etc. I really don't like it, but my mortgage has to be paid.

So the client is making you support the server but won't take your advice on how to configure it? Very simply put, I wouldn't take a job like that for any amount of money. It's a setup for failure.

(My current Rails job is for a big company using MS SQL -- but doing it right. The Rails app itself is on a *nix box with Passenger, and we developers all have test and development instances of the DB on the MS SQL server, as well as Windows VMs that we can use if we'd rather run our DBs locally.)

Best,

13

Michael Pavling wrote:

or if you want to be really cutting edge, look at Fossil (http://fossil-scm.org).

How is Fossil cutting-edge?

In so far as it is a moderately new development by people who have used other DVCS applications and said to themselves "we'd do it slightly differently". I don't mean to say it's necessarily any good... it may or may not be (fundamentally), and it may or may not be (for some specific situation).

I have seen some reports from people whom I respect the opinion of, which say they like it, and it is good from their PoV.

That's good to know.

HTH

Best,

14

Michael Pavling wrote:

15

"HAL, do the mailmerge..." :slight_smile:

16

Michael Pavling wrote:

[1] I'm currently working on some PHP development, which is running on IIS and MSSQL, because some middle-manager bought an expensive server, which has to be seen to "work". It's frustrating, but that's the situation. I work with it, or I don't. They *won't* install *nix and MySQL, etc. I really don't like it, but my mortgage has to be paid.

So the client is making you support the server but won't take your advice on how to configure it?

Not quite... they won't let me anywhere near the server full stop... paranoid city. I just give them the code and DB scripts to manipulate the data as needed... ho hum.

Very simply put, I wouldn't take a job like that for any amount of money.

You've not got my mortgage! :slight_smile:

17

Michael Pavling wrote:

Michael Pavling wrote:

[1] I'm currently working on some PHP development, which is running on IIS and MSSQL, because some middle-manager bought an expensive server, which has to be seen to "work". It's frustrating, but that's the situation. I work with it, or I don't. They *won't* install *nix and MySQL, etc. I really don't like it, but my mortgage has to be paid.

So the client is making you support the server but won't take your advice on how to configure it?

Not quite... they won't let me anywhere near the server full stop... paranoid city. I just give them the code and DB scripts to manipulate the data as needed... ho hum.

If the server is set up properly, and if they've got a decent sysadmin, that sounds great. If not, then it's a recipe for disaster.

Very simply put, I wouldn't take a job like that for any amount of money.

You've not got my mortgage! :slight_smile:

AFAIK, you have no idea what my financial obligations are. But no job is worth my sanity. I also don't really like the idea of enabling clueless behavior in clients -- I'd rather they get it through their head that if they want a good developer, they have to provide the necessary support to that developer.

My current job pays very well, and it also provides an excellent environment for me and my fellow developers. If both were not the case, I'd be out of there in a second.

Best,

18

Wow...thanks for all of the great feedback so far. I had a feeling there were better options than SVN out there. In fact, it was when I started looking into Git that I had the idea of taking some extra time to see if other elements of my environment would also benefit from an upgrade (ie my Apache & Mongrel setup). Sounds like that's the best place to begin improving to see an immediate benefit.

Peter mentioned Selenium which made me realize that I also have Watir in my toolbox for both testing and for navigating to pages for the purpose of scraping. I never tried Selenium but that might be something worth looking into as well.

I didn't intend to start a debate of whether it was good practice to generate office documents as an output but I have projects that have required it in the past and I'm sure it will come up again in the future.

Bob