Matlab floats need inf about datajoint-matlab HOT 7 CLOSED

Yes. Unfortunately, MySQL does not support the full IEEE 754 standard for floats with its special values for NaN, -Inf, and +Inf. We use NULL to stand for NaN but not sure what to do about the Infs.

from datajoint-matlab.

the problem is not specific to MySQL but to SQL in general. It's one of its major flaws. I am not sure what we can do other than document this deficiency. Valid floats can be stored in blobs but then they cannot be used in searches and sorting.

from datajoint-matlab.

we could solve this in a very complicated way by adding an additional field to flag +/- Inf for every float or double field.

from datajoint-matlab.

I propose to simply document this deficiency and mark as a wontfix.

from datajoint-matlab.

I support the wontfix idea - just document how one could get around this problem by having another field that can be used to maintain +/-Inf should that be necessary.

from datajoint-matlab.

Thanks for looking into it!

from datajoint-matlab.

another solution is to reserve special values for -Inf and +Inf The probability of hitting them by chance is 2^(-61) for doubles and 2^(-31) for single floats. But if feels wrong.

from datajoint-matlab.