Computed (but not stored) fields about datajoint-matlab HOT 4 CLOSED

Yes, many things can and should be computed on the fly rather than being stored. Currently this can be done in a number of ways and I am not sure if binding this into fetch would be the cleanest option. For example, how would users know that this extra field is available? Would it be included in the table declaration? Or should these fields somehow be added to the table heading?

Would you like to flesh this out a bit?

from datajoint-matlab.

I mentioned fetch because it was the first thing that comes to mind. I liked the idea of updating the table declaration (e.g. annotated with a new 'volatile' data type) so that in a given table/class, all fields and their descriptions appear in the same place. To compute these fields would require some extra callbacks and indirection, so you're right that it's probably not the cleanest option.

A cleaner option, perhaps, would be to simply describe extra fields as comments in the declaration, then make my own get_xyz(self, key) function for each field xyz.

from datajoint-matlab.

The self.get_xyz(key) solution is what people do now. You must be careful though since sequential fetch calls are not guaranteed to return the same things in the same order. So if you need other attributes from the same table along with xyz, you better get them in one fetch and then compute xyz for each item in the sequence.

from datajoint-matlab.

Thanks!

from datajoint-matlab.