CODEWITHSUNDEEP
pythonnumpynull
sten
sten

Reputation: 7486

Is there NULL/NIL value in numpy? for np.uint16 ! beside -1?

I'm using np.isin() to calculate overlap between two values ... f.e.

np.isin(randint(0,10,3), randint(0,10,3)).sum()

the problem is I have a case where I need NULL value (all zero rows would be good candidate) :

  z = np.array([0, 0, 0], dtype=np.uint16)
  np.isin(z,array([0,2,3])).sum()
  : 3

but the overlap should be ZERO not 3, because ZERO is real data. Currently I use null-value of 65535 i.e. -1, which I dont like very much :

  z = np.array([0, 0, 0], dtype=np.uint16) + np.uint16(-1)
  np.isin(z, np.array([0,2,3], dtype=np.uint16)).sum()
  : 0

The problem as you see is that the NULL value can not be ZERO, because ZERO is value that is legitimate data..

Is there some standardized way of handling NIL/NULL data in numpy ?

I also should have mentioned the type should be np.uint16

In [137]: zz = np.zeros(5, dtype=np.uint16)

In [138]: zz
Out[138]: array([0, 0, 0, 0, 0], dtype=uint16)

In [139]: zz[:] = np.nan

In [140]: zz
Out[140]: array([0, 0, 0, 0, 0], dtype=uint16)

Upvotes: 1

Views: 2225

Answers (1)

BlackBear
BlackBear

Reputation: 22989

You can use np.nan:

>>> np.nan == np.nan
False
>>> z = np.array([np.nan, np.nan, np.nan])
>>> np.isin(z,np.array([np.nan,2,3])).sum()
0

Upvotes: 1

Related Questions