[Dcmlib] Never trust the programer :)

[Mathieu Malaterre]

God ! I wish we never switch to speak english, it is making so much 
confusion.

Benoit:
No your code is *slightly* different from the previous one, because it 
is now asynchronous. You are testing only once if the number of value 
count is actually 3. Whereas we could *imagine* a case where you could 
deduct a value from: first an array of 3 and if if fails deduct the same 
value from an array of 1. Therefore you have to test first that 
GetValueCount() == 3 and *then* test GetValueCount() == 1. But I agree 
that in your case it should work since RET and non-RET tags seems to 
always keep the same VM (when properly written in the dict :-P).

JP:
So far the only mistake*s* (plural form) I corrected were on *our* side. 
So blaming manufacturer all the time, is kind of lame. First we should 
admit that our ACR-NEMA dict is not perfect, then correct it. And then 
after blame the constructor :). Seriously none of the VM for Orientation 
and Position were right... I mean com'on !


But I do agree that my assert is too strong in particular for the super 
duper 'special' case where Spacing is written with a VM of 3. I'll have 
to think about it. But right now I am just going over the ACR-NEMA dict 
which is IMHO not up to date at all.

My 2 cents, sorry if I hurt your feelings :(
Mathieu

Jean-Pierre Roux wrote:
> Mathieu Malaterre wrote:
> 
>> Benoit,
>>
>>     I just added an assert for kick and it is seg faulting. For instance:
> 
> 
> 
> Mathieu.
> 
> Be carefull with the VM : I added them manualy in our dicomV3.dic.
> Nobody used them till now ...
> 
> And think of the bozos that produce supposed to be Dicom images.
> We coded a lot of heuristics to allow gdcm Reader to go on when a 
> duscrepancy is found (dicomV3.dic vs current Dicom Image -being read-)
> ex : Pixel Spacing has a VM = 2;
> We have images where a single value is found (assume the pixel is square)
> We have images where 3 values are found (We know the middle one is 
> always 0 ?!?, just ignore it)
> If you check too much, you'll have more 'gdcm breaker images'.
> 
> VM should be usefull to enforce consistency, no allowing user to *write* 
> illegal stuff.
> Or inside a checker, that would warn user the image is not kosker.
> But *not* hanging gdcm.
> IMO...
> 
> JPRx
> 
>>
>>   DataEntry *entry = GetDataEntry(0x0020,0x0032);
>>    if( !entry )
>>    {
>>       gdcmWarningMacro( "Unfound Image Position Patient (0020,0032)");
>>       entry = GetDataEntry(0x0020,0x0030);
>>       if( !entry )
>>       {
>>          gdcmWarningMacro( "Unfound Image Position (RET) (0020,0030)");
>>          return 0.0f;
>>       }
>>    }
>>
>>    if( entry->GetValueCount() == 3 )
>>    {
>>       gdcmAssertMacro( entry->IsValueCountValid() );
>>       return (float)entry->GetValue(0);
>>    }
>>    return 0.0f;
>>
>> so if 0x0020,0x0032 is not found, you check 0x0020,0x0030, which is 
>> then found. But if it is found THEN you compare entry->GetValueCount() 
>> against 3, but how do you know that VM is the same for 0020,0032 and 
>> 0020,0030 ?
>>
>> Anyway this could have been safe, unfortunately 0020,0030 was entered 
>> as VM=1 in the dict.
>>
>> Question: Is this safe to let the programmer use GetValueCount ? Or 
>> should we enforce IsValueCountValid ? Maybe an alternative would be to 
>> run IsValueCountValid on the whole gdcmData just to see how respectful 
>> the DICOM data are to 'our' dict...
>>
>> Mathieu
>> _______________________________________________
>> Dcmlib mailing list
>> Dcmlib at creatis.insa-lyon.fr
>> http://www.creatis.insa-lyon.fr/mailman/listinfo/dcmlib
>>
> 
>