Table/categorical arrays and the <undefined> category

Hi

I create a table with different data. Some of the data I set as categorical arrays as such:

table.field = categorical(table.field);

This works fine. When there are NaNs or empty strings within table.field matlab automatically assigns '<undefined>' to those. Also not a problem. However, when I do a unique on them, each '<undefined>' is seen as unique instead of the same. The only way I see around this is to do something like:

table.field(isundefined(table.field)) = 'None';

That seems like a hack. Surely is must be possible to specify what the undefined "label" is in the syntax call?

Hope I'm missing something, Regards, Phil

[dpb]

Don't have table in the release here, but reading doc makes me wonder if standardizemissing would help or not. Wonder if it would fixup the whatever-it-is that makes the present situation look different.

http://www.mathworks.com/help/matlab/ref/standardizemissing.html

If not, it would appear your hack or the similar one using the id parameter to set a "non-standard" indicator would be it.

Seems like worthy of a bug/clarification report to TMW--doesn't seem like that's a good implementation even if WAD at present.

[Phillip]

Hi

Thanks for pointing that out. I did miss that function. Although it seems that it does the opposite to what I want. It takes non-standard inputs and changes it into the standard missing variables which is undefined etc. Still wouldn't solve my uniqueness problem.

[dpb]

I may have misinterpreted--my thought was perhaps one could use the function to force the missing value and that internally that might just cause the symptom to go away.

I still recommend this be submitted to "official" support for resolution of whether it is intended behavior or not--it seems wrong-headed if so on the face of it. Perhaps TMW has a reason we're not seeing, but I'd guess it's a fignewton of an implementation detail and isn't really thought out as of yet.