[Clam-devel] Re: test changes after refactoring of InstantTunningEstimator

David García Garzón dgarcia at iua.upf.edu
Tue Jul 10 08:28:32 PDT 2007


That series of patches were so nice :-)

Committed revision 10372.
 * Roman Goj patch: InstantTunningEstimatorTest
 ** Using doIt overload with vector<pair> inpunt instead the double* one
 ** Using binsPerSemitone=1.
    Step toward using a normalized [0,12) interval for tonal positions
    instead of [0,12*binsPerSemitone)
 ** Typo: centerAndStrenght -> centerAndStrength

Next,
* Locate and solve any remaining InstantTunningEstimator constructed with 
bpo!=1
* Compile, Test, Patch
* Turn the bpo default value to 1
* Compile, Test, Patch
* Use the default bpo instead of explicitly put 1
* Compile, Test, Patch
* Remove the bpo constructor parameter and turn it a class constant
* Compile, Test, Patch
* Inline/remove the constant and fade it out
* Compile, Test, Patch

At this point you'll have a InstantTunningEstimator class which is independent 
of the binPerOctave/Semitone parameter.

By the way, the svn clam-test-data files seems to be corrupt and hold the 
tests. Are you really passing the tests or just compiling them? Test farm 
will tell us. ;-) Meanwhile you can use the CLAM-TestData on the download 
section of the website.



On Tuesday 10 July 2007 15:22:15 Roman Goj wrote:
> Step 3:
>
> I thought this could be a seperate patch, but no - it still includes
> the two previous ones :( (I changed 1.5/3.0 I did in the last patch to
> 0.5, obviously should have done this from the beginning... didn't
> notice 1.5 isn't like 0.1 produces something nice when divided by 3 ;)
> )
>
> but this patch addresses also:
>
> the last two tests in InstantTunningEstimator have been changed to use
> the std::vector interface of InstantTunningEstimator's doIt and also
> to use binsPerSemitone=1 (I decided against doing it in two seperate
> patches ;) )
>
> Maybe I'll wait now for some feedback from David (maybe I should've
> sent these to your e-mail, David, rather then the list? Sorry if
> anyone feels offended by the e-mail flood in this topic!), before
> doing the next three steps.
>
> cheers,
> roman






More information about the clam-devel mailing list