Re: ert.el --- Emacs Lisp Regression Testing

[Christian Ohler <ohler+emacs@fastmail.net>]

Phil Hagelberg, 2008-01-03:

> I'm the author of elunit. I'm obviously very interested in unit testing
> in Emacs Lisp (preferably test-driven development, but any tests are
> good.) elunit has mostly been driven by my own needs, so it's no
> surprise that it doesn't seem like a great fit for everyone, but I would
> love to see unit testing get more traction in the Emacs community,
> whether it's through elunit or ERT.

I'm familiar with (an earlier version of) elunit; I've used it for 
writing a test suite for xmtn (a DVC backend for the distributed version 
control system monotone).  It served its purpose, but I wanted something 
that offers an even better workflow for test-driven development.  None 
of the existing tools did that.


> ERT looks several orders of magnitude more complicated than elunit, but
> obviously it has a lot more functionality.

There are some parts in ERT that can be simplified, but the parts that 
are really tricky (`ert-run-test-debugger' and friends and `should') are 
inherently tricky.  Other tools take simpler approaches but have less 
functionality and are less general.


> I've beaten my head against the problem of testing with a fresh
> instance, and it's always been rather awkward.

Hmm, what do you mean here?  What were your goals, what was your 
approach, and why were you unhappy with the results?


I skimmed your page on behave.el ( http://dev.technomancy.us/Behave.el 
).  What are its main features?  Adding tags and docstrings to ERT would 
not be difficult.  I think ERT's `should' is already more powerful than 
behave.el's `expect'.

I would like to look at your code, but the command from 
http://dev.technomancy.us/DotFiles ,

     git clone http://git.caboo.se/technomancy.git

says

     Cannot get remote repository information.
     Perhaps git-update-server-info needs to be run there?

Christian.