I am trying to get the elements in an ndarray that are strings. That is, exclude the elements that are integers and floats.
Lets say I have this array:
x = np.array([1,'hello',2,'world'])
I want it to return:
array(['hello','world'],dtype = object)
I've tried doing np.where(x == np.str_) to get the indices where that condition is true, but it's not working.
Any help is much appreciated.
You can make a function to do it, and loop over the array:
def getridofnumbers(num):
try:
x = int(num)
except:
return True
return False
output = np.array([i for i in x if getridofnumbers(i)])
if we want to keep all the numpy goodness (broadcasting etc), we can convert that into a ufunc using vectorize (or np.frompyfunc):
import numpy as np
#vectorize the fucntion, with a boolean return type
getrid = np.vectorize(getridofnumbers, otypes=[bool])
x[getrid(x)]
array(['hello', 'world'], dtype='<U11')
#or ufunc, which will require casting:
getrid = np.frompyfunc(getridofnumbers, 1, 1)
x[getrid(x).astype(bool)]
When you run x = np.array([1,'hello',2,'world']), numpy converts everything to string type.
If it is one dimensional array, you can use:
y = np.array([i for i in x if not i.replace(".","",1).replace("e+","").replace("e-","").replace("-","").isnumeric()])
to get all non-numeric values.
It can identify all floats with negative sign and and e+/e- )
like, for input: x = np.array([1,'hello',+2e-50,'world', 2e+50,-2, 3/4, 6.5 , "!"])
output will be : array(['hello', 'world', '!'], dtype='<U5')
xafter your first line, you'll notice the entire array hasdtype='<U11'therefore there is nothing to filtertype(x[0])and you will see that the integers are actually being stored asnumpy.str.