[Gridflow-dev] test suite (fwd)

[Alexandre Castonguay]

Bonjour Mathieu,

I agree with automating the tests, what do ou have in mind for the GUI tests?  

Cheers,

Alexandre

Le Samedi 12 Novembre 2005 01:54, Mathieu Bouchard a écrit :
> any comments on the following?
>
> ---------- Forwarded message ----------
> Date: Wed, 13 Jul 2005 19:07:32 -0400 (EDT)
> From: Mathieu Bouchard <matju at artengine.ca>
> To: gridflow-dev at artengine.ca
> Subject: [Gridflow-dev] test suite
>
>
> GridFlow could afford a better test suite. base/test.rb should be replaced
> by something more appropriate i'm thinking something that would look like:
>
> [#expect, send 1 2 3, get 4 5 6] X [# + 3]
>
> so [#expect] would be a new control object destined to do quality
> assurance. actually there's already one sort of like that in
> base/test.rb, but it's not usable outside of Ruby.
>
> there could be a [, radius 42] option so that instead of using exact
> equality like [# ==] -> [#ravel] -> [#fold &] it would use euclidean
> distance [# sq-] -> [#ravel] -> [#fold +] -> [# sqrt] as is used in
> statistics and acoustics to evaluate the energy or the RMS amplitude.
> Alternatively it could use manhattan distance [# abs-] -> [#ravel] ->
> [#fold +] which is what used in chemistry labs to compute error
> propagation. Which one to use could be controlled by a message "norm 2"
> and "norm 1" respectively.
>
> This "radius" extension to the normal Extreme Programming testing is
> necessary to address two issues:
>
> 1. the usual equality testing can't be used successfully with floats
>    because slight deviations in computation can be obtained quite easily.
>    For example you just can't rely on (A+B)+C==A+(B+C), ever, although the
>    error radius could usually be about 16 million times smaller than the
>    involved numbers.
>
> 2. similar rounding issues can occur in integer math and that it can be
>    sometimes okay for one high-level effect to yield slightly different
>    numbers as long as it appears roughly the same on-screen.
>
> Other automated testing patterns should be developed to address issues
> such as testing GUI objects like [display] [#peephole] [#edit].
>
> Possibly that it's time to introduce semi-automated testing, in which the
> part that is fully automatable gets automated, but is interleaved with
> documentation taking the tester by the hand, so that the tester doesn't
> have to think too much and doesn't forget what it is s/he is supposed to
> test; this makes testing time shorter thus tests become more affordable.
>
> What do you think?
>
> ,-o--------o--------o--------o-. ,---. irc.freenode.net #dataflow |
>
> | The Diagram is the Program tm| | ,-o-------------o--------------o-.
>
> `-o------------o-------------o-' | | Mathieu Bouchard (Montréal QC) |
>
>   | téléphone:+1.514.383.3801`---' `-o-- http://artengine.ca/matju -'
>
> _______________________________________________
> Gridflow-dev mailing list
> Gridflow-dev at artengine.ca
> http://ns.artengine.ca/mailman/listinfo/gridflow-dev
>
> _______________________________________________
> Gridflow-dev mailing list
> Gridflow-dev at artengine.ca
> http://ns.artengine.ca/mailman/listinfo/gridflow-dev