CODEWITHSUNDEEP
pythonpandas
Jeroen Vermunt
Jeroen Vermunt

Reputation: 880

pandas slightly changes assigned value

I'm assigning a value to a specific pandas location. However, the value seems to slightly change for some reason. Does anyone know why this happens? I posted a snippet of my code below. I added the prints statements to see in more detail what happens, but it makes no sense to me.

if entry != None:

orders.iloc[index, orders.columns.get_loc('start_time_corrected')] = entry['t'].astype(np.int64)

print(entry['t'].astype(np.int64))                                
print(int(orders.iloc[index]['start_time_corrected']))

print(np.datetime64(int(orders.iloc[index]['start_time_corrected']),'ns'))
print(entry['t'])

which gives the following output:

1615606416410318537
1615606416410318592

2021-03-13T03:33:36.410318592
2021-03-13T03:33:36.410318537

as you can see the value in the output slightly changes, even though it should be the same value.

Upvotes: 0

Views: 78

Answers (1)

user2357448
user2357448

Reputation:

There is a lost of casting from int to numpy.int64 going on here and that is why you have a difference in values.

Because you are casting orders.iloc[index]['start_time_corrected'] to int before you print it, you are going to lose information as python allocates 32-bits for for a normal integer type while numpy's np.int64 datatype is allocated 64 bits.

Upvotes: 1

Related Questions