How do I make the following conversion in NumPy?
["1.1", "2.2", "3.2"] ⟶ [1.1, 2.2, 3.2]
Add a comment
|
6 Answers
Well, if you're reading the data in as a list, just do np.array(map(float, list_of_strings)) (or equivalently, use a list comprehension). (In Python 3, you'll need to call list on the map return value if you use map, since map returns an iterator now.)
However, if it's already a numpy array of strings, there's a better way. Use astype().
import numpy as np
x = np.array(['1.1', '2.2', '3.3'])
y = x.astype(np.float)
5 Comments
ePascoal
and if you have a array with an string that i want to maintain? like ['a','1.1','2.2','3.3'] -> ['a',1.1,2.2,3.3]
Joe Kington
@MrMartin - Then use a
list. Numpy arrays are deliberately homogenously typed. If you really want, you can use an object array (e.g. np.array(['apple', 1.2, 1, {'b'=None, 'c'=object()}], dtype=object)). However, object arrays don't have any significant advantages over using a list.
Subham
print(list(y)) gives [1.1, 2.2, 3.3] [Program finished] as requested by OP
germ
Does not work if any of the strings is an empty string ('').
Sarim Bin Waseem
It's now: ` y = x.astype(float) ` np.float is deprecated.
Another option might be numpy.asarray:
import numpy as np
a = ["1.1", "2.2", "3.2"]
b = np.asarray(a, dtype=float)
print(a, type(a), type(a[0]))
print(b, type(b), type(b[0]))
resulting in:
['1.1', '2.2', '3.2'] <class 'list'> <class 'str'>
[1.1 2.2 3.2] <class 'numpy.ndarray'> <class 'numpy.float64'>
answered Jun 12, 2018 at 15:19
Herpes Free Engineer
2,7343 gold badges29 silver badges37 bronze badges
3 Comments
jodag
I benchmarked all the answers here in python 2.7. Assuming I'm given a list of 512 strings which represent floating point numbers, this approach was the fastest (slightly faster than pradeep bisht's answer, about 1.5 times faster than Thomio's answer, and more than twice as fast as the accepted answer). Have an upvote!
DavidS
"order='C'" is not needed here since this is a 1-dim array. This also works (at least in Python 3.6.9): b=np.array(a, dtype=float)
Yu Da Chi
add
.tolist() at the end, like that b = np.asarray(d, dtype=np.float64).tolist() to get comma separated listIf you have (or create) a single string, you can use np.fromstring:
import numpy as np
x = ["1.1", "2.2", "3.2"]
x = ','.join(x)
x = np.fromstring( x, dtype=np.float, sep=',' )
Note, x = ','.join(x) transforms the x array to string '1.1, 2.2, 3.2'. If you read a line from a txt file, each line will be already a string.
Comments
You can use np.array() with dtype = float:
import numpy as np
x = ["1.1", "2.2", "3.2"]
y = np.array(x,dtype=float)
Output:
array([1.1, 2.2, 3.2])
Comments
You can use this as well
import numpy as np
x=np.array(['1.1', '2.2', '3.3'])
x=np.asfarray(x,float)
Comments
If you have an array that contains invalid values such as an empty string (''), then a straightforward casting would raise an error. If you just want to convert this "problematic" array into a numpy float array and handle the invalid values later, then pandas package has a function (pandas.to_numeric) that sets invalid values to NaN and converts the rest to float.
import pandas as pd
a = np.array(["1.1", "2.2"], float) # OK
lst = ["1.1", "2.2", "3.2.", ""]
a = np.array(lst, float) # ValueError: could not convert string to float: '3.2.'
a = pd.to_numeric(lst, errors='coerce') # array([1.1, 2.2, nan, nan])