Re: Re: [PEPr] +1 for Web Services::Services_Delicious

Stephan Schmidt wrote:

Hi, Christian Weiske wrote:

So leave the fields for username and password blank. If anyone wants to use the package and it doesn't work, he has the *chance* to test it - if he uses it is a different question.

There are times, when the service is responding with in internal server error due to a very high load. How should unit tests cope with that.

Mock objects can be used to easily simulate any input that the thing can provide - as long as it is designed to allow plugging them in. For examples, see the unit tests for PEAR - they literally simulate PEARweb as well as downloading files, so that error conditions can be tested as well. But having said this...

While I agree, that unit tests are a good thing, they are not suited for every package. We have rules, that they should be included in stable releases, but how do you test against a moving target. The tests are supposed to test the package, but it would be more like testing whether the service still responds the same way it did yesterday.

... I agree - as long as a package remains alpha or less stable, unit tests don't make much sense. When things calm down, they become feasible - it's an incredible amount of work to do tests, but much more to *re-do* tests.

There are plans to provide different versions of the service using different URLs. If this is available, it will make sense. Otherwise, it would be like testing wether 1+1 equals 2 while somebody is changing the law of mathematics....

1+1 = 6 votes in Florida, but sometimes 1+1 = -6 votes. I'd love to see regression tests for that system. :) Greg