query properties of numpy dtypes

Question

I have a program which uses at runtime some numpy datatypes e.g. uint8, int8, uint16, int16, uint32, int32.

Is there a way at run time to query these datatype objects and determine the following properties?

• their minimum value
• their maximum value
• whether they are signed
• the size in bytes

Answer 1

For ints you could use iinfo to get part of the way to what you want:

d = np.dtype('int8')  # for example

min_value = np.iinfo(d).min
max_value = np.iinfo(d).max
signed = min_value!=0   # can also use "dtype.kind", see Jaime's comment
size = d.itemsize       # see comment