If I'm developing a proof-of-concept application, does it make sense to invest time in writing automated tests? This is for a personal project where I am the sole developer.
I see the only benefit of automated testing at this point as:
If the concept catches, the tests already exist.
Some of the cons related to writing automated tests for this type of project could be:
It takes valuable time to write tests for an idea that might not be worthwhile to people.
At this level, time is better spent building a demonstration of your idea.
Can anyone provide pros and cons of investing time in writing automated tests for an application in its early stages?
This whole talk from the Google Testing Automation Conference is about your question:
http://www.youtube.com/watch?v=X1jWe5rOu3g
Basically, the conclusion is that it is more important to know you are building the right thing than to build something right (build the right "it", rather than build "it" right). The most important thing is to get a proof-of-concept through and make sure that it works and is liked. If people like your thing, then they will tolerate bugs; but if they don't like your thing, it can have no bugs and they still won't like it.
TDD is not really about testing, it's about designing. Doing TDD for your application will make it have a better design (probably) than just doing it on your feeling.
Your problem is : Do you need a good design ? Design is helpful for maintainance and most devs doing TDD consider themselves in maintainance mode just after having added their 1st feature.
On a more pragmatic perspective : if you're the only dev, have very accurate specs and work on this code to do it and never return to it (nor send someone else return to it), I would say that making it work is enough.
But then don't try to get anything back from it if your POC works, and just redo it.
You can save time by doing an ugly POC and come to the conclusion that your idea is not doable.
You can save time by doing an ugly POC and understanding much better the domain you're trying to model
You cannot save time by trying to get some lines of code out of an horrible codebase.
My best advice for estimating how much effort you should put in design (because overdesigning can be a big problem, too) is : try to estimate how long will that code live
Reference : I would suggest you to make some research on the motto "Make it work, make it right, make it fast" . The question you ask is about the 2 first points but you will sooner or later ask yourself the same question about optimization (the third point)
There's no "right" answer. TDD can make your concept stronger, more resilient, easier to bang on, and help drive API development. It also takes time, and radical changes mean test changes.
It's rare you get to completely throw away "prototype" code in real life, though.
The answer depends entirely on what happens if you prove your concept. True Proof-of-Concept applications are thrown away regardless of the outcome, and the real application is written afterward if the PoC proved out. Those PoCs obviously don't need tests. But there are way too many "productized PoCs" out there. Those applications probably should have tests written right up front. The other answers you've received give you solid support for both positions, you just need to decide which type of PoC you're building.
Related
I am looking to start developing a relatively simple web application that will pull data from various sources and normalizing it. A user can also enter the data directly into the site. I anticipate hitting scale, if successful. Is it worth putting in the time now to use scalable or distributed technologies or just start with a LAMP stack? Framework or not? Any thoughts, suggestions, or comments would help.
Disregard my vague description of the idea, I'd love to share once I get further along.
Later. I can't remember who said it (might have been SO's Jeff Atwood) but it rings true: your first problem is getting other people to care about your work. Worry about scale when they do.
Definitely go with a well structured framework for your own sanity though. Even if it doesn't end up with thousands of users, you'll want to add features as time goes on. Maintaining an expanding codebase without good structure quickly becomes fairly horrible (been there, done that, lost the client).
btw, if you're tempted to write your own framework, be aware that it is a lot of work. My company has an in-house one we're quite proud of, but it's taken 3-4 years to mature.
Is it worth putting in the time now to use scalable or distributed technologies or just start with a LAMP stack?
A LAMP stack is scalable. Apache provides many, many alternatives.
Framework or not?
Always use the highest-powered framework you can find. Write as little code as possible. Get something in front of people as soon as you can.
Focus on what's important: Get something to work.
If you don't have something that works, scalability doesn't matter, does it?
Then read up on optimization. http://c2.com/cgi/wiki?RulesOfOptimization is very helpful.
Rule 1. Don't.
Rule 2. Don't yet.
Rule 3. Profile before Optimizing.
Until you have a working application, you don't know what -- specific -- thing limits your scalability.
Don't assume. Measure.
That means build something that people actually use. Scale comes later.
Absolutely do it later. Scaling pains is a good problem to have, it means people like your project enough to stress the hardware it's running on.
The last company I worked at started fairly small with PHP and the very very first versions of CakePHP that came out (when it was still in beta). Some of the code was dirty, the admin tool was a mess (code-wise), and sure it could have been done better from the start. But do you know what? They got it out the door before their competitors did, and became extremely successful.
When I came on board they were starting to hit the limits of their current potential scalability, and that is when they decided to start looking at CDN's, lighttpd caching techniques, and other ways to clean up the code and make things run smoother when under heavy load. I don't work for them anymore but it was a good experience in growing an architecture beyond what it was originally scoped at.
I can tell you right now if they had tried to do the scalability and optimizations before selling content and getting a website live - they would never have grown to the size they are now. The company is www.beatport.com if you're interested in who I'm talking about (To re-iterate, I'm not trying to advertise them as I am no longer affiliated with them, but it stands as a good case study and it's easier for people to understand what I'm talking about when they see their website).
Personally, after working with Ruby and Rails (and understanding the separation!) for a couple of years, and having experience with PHP at Beatport - I can confidently say that I never want to work with PHP code again =p
Funny to ask "scale now or later?" and label it "ruby on rails".
Actually, Ruby on Rails was created by David Heinemeier Hansson, who has a whole chapter in his book labeled "Scale later" :))
http://gettingreal.37signals.com/ch04_Scale_Later.php
I agree with the earlier respondents -- make it useful, make it work and get people motivated to use it first. I also agree that you should pick off-the shelf components (of which there are many) rather than roll your own, as much as possible. At the same time, make sure that you choose components for your infrastructure that you know to be scalable so that you can go there when you need to, without having to re-write major chunks of your application.
As the Product Manager for Berkeley DB, I've seen countess cases of developers who decided "Oh, we'll just write that to a flat file" or "I can write my own simple B-tree function" or "Database XYZ is 'good enough', I don't have to worry about concurrency or scalability until later". The problem with that approach is that a) you're re-inventing the wheel (and forgoing what others have learned the hard way already) and b) you're ignoring the fact that you'll have to deal with scalability at some point and going with a 'good enough' solution.
Good luck in your implementation.
In the light of BDD, would it be a good idea to use Cucumber features and scenarios to write up the functional requirements document at the start of a new Rails project?
Probably not. This would be a case of BDUF (http://en.wikipedia.org/wiki/Big_Design_Up_Front).
It's unlikely you'll be able to think of all the scenarios up front. You should split the high-level requirements into features to help estimate and prioritise them, but leave the detailed scenario writing to just before you're ready to begin implementing each feature.
If you are not the one making all the decisions and someone thinks you need these,
then it might appear to be a better choice than MS word.
I actually joined a project with a million un-implemented features,
so we had loads of integrations test in theory, just none
of them were actually implemented.
Its months later and we still have some,
its really difficult working in an environent where
everything is failing at once.
Its better to have 1 failing step at a time.
I also think feautures should be written by application developers who
understand user flow realities.
I like the business or client to explain stuff to me in a conversational syle,
a bit at a time, not their entire vision for world domination in 10 000 words,
I will keep that for bed time.
When writing code for a new system I don't want to introduce unnecessary complexity in the design that I might never have any need for. So I'm following YAGNI here, and rather refactoring as I see the need for more flexibility or as responsibilities becomes more clear. This allows me to move faster.
But there is a problem here with junior devs, in that they will not recognize when to refactor or where build out the design. They just stuff more code into the existing design.
So, what are the best ways to tackle this? Should I more often build a more future-proof design so when adding to it they have a good example to follow, even if we might never have to add anything? Or should I just go ahead with more code reviews, education, etc? Or both?
Have any of you had any experience with this type of problem? How did you solve it?
I would recommend code reviews or pair programming. It gives you the chance to educate your fellow developers and increase the overall quality.
Perhaps you begin by recognizing explicitly that part of your job is to help develop the junior devs. If you're not the boss, management should sign off on this. Management needs to recognize that your choices are to develop them now or clean up after them later, and you need management's backing for the time this will take.
Code reviews and pair programming are fine ideas. They are especially good because they are not "just for junior people"–I do both with one of my close colleagues; together we are nearly 100 years old and have more than 70 years of programming experience :-)
But there's a larger problem here: the programming methodology that enables you to be most effective (YAGNI + refactor) is not effective for your junior partners. My experience is that it takes people years to learn the benefits of YAGNI, so if you expect them just to learn your way of doing things, you are setting yourself up for disappointment.
I would encourage you to identify some methodology that you think is going to be useful with your junior partners. The particular methodology probably doesn't matter (heresy!); I've had success with composite/structured design, object-based design, algebraic specification (!), and extreme programming. But
Do pick something that has a name and some literature devoted to it, that your juniors can take pride in learning, and that is a skill they can carry to future projects.
In order to show that it is tasty, you may need to eat the dog food yourself. Pick something you can live with and be productive in.
Observe your juniors carefully and teach them a decision procedure they can use to identify when they should ask you for guidance.
Good luck!
There is a reason they are junior and you are senior.
The ability to realise when a change in design is needed is one of them.
I would carry on as you are but encourage them to come to you when things are getting difficult. You can then work with them to alter the design if needed, this will be easier for you than refactoring it and will help you pass on knowledge to your junior developers.
A very good way to show how far to build out a design is to be specify about what the design will do when built out, then write tests to cover the new functionality. When the tests pass, development is done.
You might realize along the way that you forgot to test for something. That's fine, and is useful feedback to help you specify better next time. Write the missing test(s), and only enough code to make them pass.
Then refactor. Knowing what to look for when refactoring takes a bit of practice. Start with
Is there duplication in the code we've just written that we can eliminate?
Is there duplication between what we've just written and pre-existing code?
Does the code we've just written concern itself with too many things? (I.e., should we break out collaborators?)
Repeat this a few dozen times, and it'll get easier.
Another way of looking at YAGNI is that any changes to code need to be justified.
Might it be helpful to require any commit needs an associated unit test (or BDD user story, choose your poison)? It helps to communicate your intent (you want people to think about why they are adding this feature) and you get regression tests for free.
Also gets the noobs to start thinking about modularity (usually needed to make your code testable) and will help a lot if you do need to refactor later on.
I'm all for code reviews and teaching, but I think futureproof design is also important. Maybe you could think of it in terms of you designing an API and the junior developers using the API. In this way you are the one who does the hard work that they would screw up (identifying duplicated code and eliminating it) while they do all the grunt work that isn't a productive use of your time.
Of course this has to be balanced with a need to develop your junior developers skills. Neither side of the equation can be neglected.
It may help to map out what work they will do and then verify it to help build their sense of judgement which is really what you are asking, to my mind. Pairing is one option but if you can't spare that much time then having a sort of "check point" to see how they are doing and preventing them from going down the wrong path.
I have a Rails project which I neglected to build tests for (for shame!) and the code base has gotten pretty large. A friend of mine said that RSpec was a pain to use unless you use it from the beginning. Is this true? What would make him say that?
So, considering the available tests suites and the fact that the code base is already there, what would be my best course of action for getting this thing testable? Is it really that much different than doing it from the beginning?
This question came up recently on the RSpec mailing list, and the advice we generally gave was:
Don't bother trying to retro-fit specs to existing, working, code unless you're going to change it - it's exhausting and, unless the code needs to be changed, rather pointless.
Start writing specs for any changes you make from now on. Bug fixes are an especially good opportunity for this.
Try to train yourself into the discipline that before you touch the code, first of all write a failing example (=spec) to drive out the change.
You may find that the design of code which wasn't driven out by code examples or unit tests makes it awkward to write tests or specs for. This is perhaps what your friend was alluding to. You will almost certainly need to learn a few key refactoring techniques to break up dependencies so that you can exercise each class in isolation from your specs. Michael Feathers' excellent book, Working Effectively With Legacy Code has some great material to help you learn this delicate skill.
I'd also encourage you to use the built-in spec:rcov rake task to generate code coverage stats. It's extremely rewarding to watch these numbers go up as you start to get your codebase under test.
Maybe start with the models? They should be testable in isolation, which ought to make them the lowest-hanging fruit.
Then pick a model and start writing tests that say what it does. As you go along, think about other ways to test the code - are there edge cases that maybe you're not sure about? Write the tests and see how the model behaves. As you develop the tests, you may see areas in the code that aren't as clean and de-duplicated (DRY) as they might be. Now you have tests, you can refactor the code, since you know that you're not affecting behaviour. Try not to start improving design until you have tests in place - that way lies madness.
Once you have the models pinned down, move up.
That's one way. Alternatives might be starting with views or controllers, but you may find it easier to start with end-to-end transaction tests andwork your way into smaller and smaller pieces as you go along.
The accepted answer is good advice - although not practical in some instances. I recently was faced with this problem on a few apps of mine because I NEEDED tests for existing code. There simply was no other way around it.
I started off doing all unit tests, then moved onto functionals.
Get in the habit of writing failing tests for any new code, or whenever you're going to change a part of the system. I've found this has helped me gain more knowledge of testing as I go.
Use rcov to measure your progress.
Good luck!
Writing tests for existing code may reveal bugs in your code. These tests will force you to look at the existing code so you can see what test you need to write in order to get it to pass and you may see some code that could possibly be written better, or is now useless.
Another tip is to write a test when you encounter a bug so it should never re-occur, this is called regressional testing.
Retrofitting specs is not inevitably a bad idea. You go from working code to working code with known properties which allows you to understand whether any future change breaks anything. At the moment if you need to make a change how can you know what it will affect?
What people mean when they say that it is hard to add tests/specs to exisitng code is that code which is hard to test is often highly coupled which makes it hard to write low-level isolated tests.
One idea would be to start with full-stack tests using something like the RSpec story runner. You can then work from the 'outside in' isolating what you can in low-level isolated tests and gradually untangle the harder code bit by bit.
You can start writing "characterization tests". With this,you might what to try out the pretentious gem here:
It is still a work in progress though.
I'm confused about what the various testing appliances in Ruby on Rails are for. I have been using the framework for about 6 months but I've never understood the testing part of it. The only testing I've used is JUnit3 in Java and that only briefly.
Everything I've read about it just shows testing validations. Shouldn't the validations in rails just work? It seems more like testing the framework than testing the your code. Why would you need to test validations?
Furthermore, the tests seem super fragile to any change in your code. So if you change anything in your models, you have to change your tests and fixtures to match. Doesn't this violate the DRY principle?
Third, writing test code seems to take alot of time. Is that normal? Wouldn't it just be faster to refresh my browser and see if it worked? I already have to play with my application just to see if it flows correctly and make sure my CSS hasn't exploded. Why wouldn't manual testing be enough?
I've asked these questions before and I haven't gotten more than "automated testing is automated". I am smart enough to figure out the advantages of automating a task. My problem is that costs of writing tests seem absurdly high compared to the benefits. That said, any detailed response is welcome because I probably missed a benefit or two.
Shouldn't
the validations in rails just work? It
seems more like testing the framework
than testing the your code. Why would
you need to test validations?
The validations in Rails do work -- in fact, there are unit tests in the Rails codebase to ensure it. When you test a model's validation, you're testing the specifics of the validation: the length, the accepted values, etc. You're making sure the code was written as intended. Some validations are simple helpers and you may opt not to test them on the notion that "no one can mess up a validates_numericality_of call." Is that true? Does every developer always remember to write it in the first place? Does every developer never accidentally delete a line on a bad copy paste? In my personal opinion, you don't need to test every last combination of values for a Rails' validation helper, but you need a line to test that it's there with the right values passed, just in case some punk changes it in the future without proper forethought.
Further, other validations are more complex, requiring lots of custom code -- they may warrant more thorough testing.
Furthermore, the tests seem super
fragile to any change in your code. So
if you change anything in your models,
you have to change your tests and
fixtures to match. Doesn't this
violate the DRY principle?
I don't believe it violates DRY. They're communicating (that's what programming is, communication) two very different things. The test says the code should do something. The code says what it actually does. Testing is extremely important when there is a disconnect between those things.
Test code and application code are intimately linked, obviously. I think of them as two sides of a coin. You wouldn't want a front without a back, or a back without a front. Good test code reinforces good application code, and vice versa. The two together are used to understand the whole problem that you're trying to solve. And well written test code is documentation -- it shows how the application code should be used.
Third, writing test code seems to take
alot of time. Is that normal? Wouldn't
it just be faster to refresh my
browser and see if it worked? I
already have to play with my
application just to see if it flows
correctly and make sure my CSS hasn't
exploded. Why wouldn't manual testing
be enough?
You've only worked on very small projects, for which that testing is arguably sufficient. However, when you work on a project with several developers, thousands or tens of thousands of lines of code, integration points with web services, third party libraries, multiple databases, months of development and requirements changes, etc, there are a lot of other factors in play. Manual testing is simply not enough. In a project of any real complexity, changes in one place can often have unforeseen results in others. Proper architecture helps mitigate this problem, but automated testing helps as well (and helps identify points where the architecture can be improved) by identifying when a change in one place breaks another.
My problem is that
costs of writing tests seem absurdly
high compared to the benefits. That
said, any detailed response is welcome
because I probably missed a benefit or
two.
I'll list a few more benefits.
If you test first (Test Driven Development) your code will probably be better. I haven't met a programmer who gave it a solid shot for whom this wasn't the case. Testing first forces you to think about the problem and actually design your solution, versus hacking it out. Further, it forces you to understand the problem domain well enough to where if you do have to hack it out, you know your code works within the limitations you've defined.
If you have full test coverage, you can refactor with NO RISK. If a software problem is very complicated (again, real world projects that last for months tend to be complicated) then you may wish to simplify code that has previously been written. So, you can write new code to replace the old code, and if it passes all of your tests, you're done. It does exactly what the old code did with respect to the tests. For a project that plans to use an agile development method, refactoring is absolutely essential. Changes will always need to be made.
To sum up, automated testing, especially test driven development, is basically a method of managing the complexity of software development. If your project isn't very complex, the cost may outweigh the benefits (although I doubt it). However, real world projects tend to be very complex, and the results of testing and TDD speak for themselves: they work.
(If you're curious, I find Dan North's article on Behavior Driven Development to be very helpful in understanding a lot of the value in testing: http://dannorth.net/introducing-bdd)
I haven't really used Rails much, but I would think that these automated tests would be useful as smoke tests to be sure that the thing you just did doesn't break something that you did last week. This will become increasingly important as your project grows.
Also, writing the tests before you write the code (using the Test-Driven-Development model) will help you write the code better and faster, since the tests force you to fully think the problem through. It will also help you to know where to break up complex methods into smaller methods that you can test individually.
You are right, writing and maintaining tests takes a lot of time. Sometimes more time than the code itself. However, it can save you time in bug fixing and refactoring for the reasons above.
Tests should validate your application logic. Personally, I think my most important tests are the ones I run in Selenium. They check that what shows up in the browser is actually what I expect to see. However, if that's all I had, then I would find it hard to debug - it helps to have lower level tests as well and integration, functional, and unit tests are all useful tools. Unit tests let you check that the model behaves the way you expect it to (and that means every method, not just validatins). Validatins will certainly Just Work, but only if you get them right. If you get them wrong, they will Just Work, but not the way you expected. Writing a couple of lines of test is quicker than debugging later on.
A simple example like the one at http://wiseheartdesign.com/2006/01/16/testing-rails-validations just checks validations in a unit test. The O'Reilly article at http://www.oreillynet.com/pub/a/ruby/2007/06/07/rails-testing-not-just-for-the-paranoid.html?page=1 is a bit more complete (though still fairly basic).
Automated testing is particularly useful in regression testing where you change something and run a suite of tests to check that you didn't break anything else.
Tests are a form of repetition, but they don't violate DRY because they express things in a different way. A test says "I did X so Y should happen". Code says "X happened, so now I need to do Z, which happens to cause Y to happen". i.e. a test stimulates a cause and checks an effect, while code responds to a cause, and effects something.
A lot of the testing tutorials and the sample tests created by the Rails generators are pretty lame and IMHO that can give the mistaken impression that you're supposed to test stupid stuff like the built in Rails methods, etc.
Since Rails has it's own test suite, there's no point in you writing or running tests that only test built in Rails functionality. Your tests should exercise the code you're writing! :-)
As for the relative merit of running tests vs just refreshing in your browser.. The larger your app gets, the more of a pain in the ass it is to have to manually run through numerous scenarios and edge cases to make sure nothing in your application has broken. Eventually, you'll stop testing your entire application after each change and just start "spot testing" the areas you think should have been affected. Inevitably, you'll find something that used to work months ago that is now completely broken, and you have no certainty when it broke or which changes broke it. After that happens enough times... you'll come to value automated testing.... :-)
For example:
I work on a 25000+ lines project (yes, in rails 1.2) and last monday I was told if I could make Users dissapear from every list except admin ones if they had "leave_date" attribute set to the past.
You can rewrite every list action (50+) to put a
#users.reject!{|u| Date.today > u.leave_date}
Or you can override the "find" method (DRY;-), but only if you have tests (on everything that finds users!) you will know you didn't break anything by overriding User#find !!
Everything I've read about it just shows testing validations. Shouldn't the validations in rails just work? It seems more like testing the framework than testing the your code. Why would you need to test validations?
There's a good Railscast showing one way to test controllers.