[Clam-devel] Re: adding the documents to the scheme of SilenceDetector

[JunJun]

Hi, here I send a patch to populate a pool, and populate segmentation with random data.

I think sending patches might be better than documentation of the exploration. Could you drop some more patch-specific targets on the Todo list.

The patch should get the full path within the repository. Please take a look at the diffs. The memory of Windows_mingw native compile starts to fade. Is it my task to  update the wiki?   

June

----- Original Message ----- 
From: "David García Garzón" <dgarcia at iua.upf.edu>
To: "JunJun" <wangjun at dsp.ac.cn>
Cc: <clam-devel at llistes.projectes.lafarga.org>
Sent: Friday, May 30, 2008 4:22 PM
Subject: Re: adding the documents to the scheme of SilenceDetector


> Be careful when merging. You almost reverted the changes i did: changing the 
> uri before adding any attribute and using literals to set the documentation 
> instead intermediate vars.
> 
> El Friday 30 May 2008 04:17:52 JunJun va escriure:
>> Hi!
>>
>> >BTW, you may end up editing files in CLAM/src so it's better if you
>> > generate the patches from an upper level to comprise all the repository.
>>
>> no, I edited files in mingw/local/include/CLAM, which have no svn function,
>> what to do?  The changes there, however, are all for the native debug and
>> are meaningless to patch them.
> 
> Anyway do the patches from the bottom of the clam repository. Still you can 
> restrict the changes included in the path specifying the files or directories 
> on the command line, for example:
> 
> svn di Annotator/src CLAM/src/data/Description/Pool/Schema.hxx > my.patch
> 
> The important bit is that you get the full path within the repository on the 
> patch.
> 
>> I 'd like to add a TODO to the session list. Please check it out-
> 
> Although i am adding thing  to the TODO's, it is you tool so feel free to use 
> it.
> 
>> TODO: fix one of the pool's inconsistent states- Given a schema, what
>> should pool do with no added scope instances? * Given a description scheme,
>> and given a user-defined descriptors file suffix (.detector), an empty pool
>> should be dumped. e.g. given the schema of SilenceDetector:
>>            <?xml version="1.0" encoding="UTF-8"?>
>>            <Schema>
>>             
>> <Uri>descriptionScheme:www.iua.upf.edu:clam:JunJunSilenceDetector</Uri>
>> <Attributes>
>>                <Attribute name="Silences" scope="Song" type="Segmentation">
>>                      <ChildScope></ChildScope>
>>                     <SegmentationPolicy>Discontinuous</SegmentationPolicy>
>>               </Attribute>
>>             </Attributes>
>>           </Schema>
>>
>> the empty pool named "x.mp3.detector" would be like:
>> <?xml version="1.0" encoding="UTF-8"?>
>> <DescriptorsPool>
>>   <ScopePool name="Song" size="1">
>>     <AttributePool name="Silence" size="0"/>
>>   </ScopePool>
>> </DescriptorsPool>
>>
>> One problem is, how to initiate the size of the TBD AttributePool  ,
>> size="0"? size="1"?
> 
> I guess that the size attribute of the AttributePool refers the size of the 
> one and only segmentation. You can add segmentations and it will increase. 
> But you spotted another 'bug', congratulations! If you had two songs, as the 
> Attribute pool can have just a single size attribute, just the last one would 
> survive. So we need an inner XML element for each segmentation. 'Gladly' we 
> have used segmentations just on the song level and normally Song is a 
> singleton scope, so we can keep it like this by now but write down the bug on 
> the TODO's.
> 
> I added some steps to spot even more irregularities.
> 
> David.
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PopulatePoolAndSegmentationWithRandomData.patch
Type: application/octet-stream
Size: 5939 bytes
Desc: not available
URL: <http://lists.clam-project.org/pipermail/clam-devel-clam-project.org/attachments/20080606/035123a8/attachment-0004.obj>