Dynamic Attribute Filter

When testing I read data from a CSV. I'd like to limit the samples to several categories which is dynamically generated from a training set.
The training set might only have 20 categories but the test set could have 200.  I only want to test on the 20.  
The rest of the samples will be filtered out.

I read in the training set and extract the category list.
I remove duplicates to now have a unique list of categories.
This is what I want to filter my test set on.

I save the list to a file for later lookup if needed.
Now i'd like to read in the test data, filter on that list of categories and press on with testing.

How would I do such a thing? 

Thanks.