deepak deepak - 1 year ago 55
Python Question

How to search for rows using an expression in numpy rec array

Consider i have a numpy recarray as below:

r = np.rec.array([(1,2,'Hello'),(2,3,"World")], \
dtype=[('foo', 'i4'),('bar', 'i4'), ('baz', 'S10')])

Now i want to get all the object that returns 'True' for the following expression stored in dict,

expression = {'foo':1,'bar':3.,'baz':'Hello'}

I am looking for a method like**expression) to select me all the objects that match those conditions passed to the select.

expression is dynamic, changes very frequently.

Answer Source

You can combine array comparisons with boolean operators. For example (with a poor-mans implementations of a float comparison):

EPS = 1e-7
mask = np.ones(len(r), dtype=np.bool)
mask &= r['foo'] == 1
mask &= (r['bar'] > 3-EPS) & (r['bar'] < 3+EPS)
mask &= r['baz'] == 'hello'

Or you can use or, | instead of and, &, if you prefer that. (Or a combination, since the two r['bar'] will have to be &-ed anyway.)