Rounding error across all column values in Python

Question

Good afternoon,

I have a useful piece of code copied in below that appears to work on my dummy data, alas it doesn't work across all values as it goes down the column data.

Code is:

import pandas as pd
from decimal import *

Data['Value'] = Data['Value'].map(lambda x: Decimal(x).quantize(Decimal('0.001'), rounding=ROUND_HALF_UP))

Dummy data:

Data = {'Code' : ['1', '2', '3','4','5','6'],
    'Group' : ['a', 'b','c','d','e','f'],
    'Value' : [20.2535, 21.874, 7.6475,101.84556, 209.8625913, 99.7655037]}
       

Data = pd.DataFrame(Data, columns = ['Code', 'Group', 'Value'])

The output works for the first value however, anymore than that rounds down, not up. Can this be resolved because as he code states, I want to round halfs up?

  output:              Expected: 

0     20.253         0     20.254 << This one is incorrect
1     21.874         1     21.874
2      7.647         2      7.648 << This one is incorrect
3    101.846         3    101.846
4    209.863         4    209.863
5     99.766         5     99.766

Any help would be much appreciated. I'm looking for rounding to 3.dp where all values are rounded half up as per excel.

Thank you.

Why do you expect 20.253499999999998948396751075051724910736083984375 to get rounded to 20.254 instead of 20.253? It's clearly closer to the latter. — superb rain, Aug 17 '20 at 13:13
@Divyessh Wait, why were you sorry for that? It's not your question. Or are you behind both accounts? — superb rain, Aug 17 '20 at 14:00
@superbrain I changed it as if it will have more than 4 decimal places then my code will shift to regular rounding. — Divyessh, Aug 18 '20 at 04:09

aerijman · Accepted Answer · 2020-08-17T14:10:13.633

-1

Rounded DOWN (to the 3rd decimal):

Data.loc[:,'Value'] = np.round(Data['Value'] - 5e-4, 3)

Rounded UP:

Data.loc[:,'Value'] = np.round(Data['Value'], 3)

I expect that with all versions of python 5e-4 = 0.0005. If I am wrong, use 0.0005 or 5*10**-4

Data:

  Code Group       Value
0    1     a   20.253500
1    2     b   21.874000
2    3     c    7.647500
3    4     d  101.845560
4    5     e  209.862591
5    6     f   99.765504

DOWN output:

  Code Group    Value
0    1     a   20.253
1    2     b   21.874
2    3     c    7.647
3    4     d  101.845
4    5     e  209.862
5    6     f   99.765

UP output

  Code Group    Value
0    1     a   20.254
1    2     b   21.874
2    3     c    7.648
3    4     d  101.846
4    5     e  209.863
5    6     f   99.766

edited Aug 17 '20 at 14:10

answered Aug 17 '20 at 13:23

aerijman

2,522
1
22
32

are you not doing the same mistake that I earlier did – Divyessh Aug 17 '20 at 13:29
@Divyessh, `round(20.253500,3) = 20.253`. `np.round(20.253500,3) = 20.254`. So, **NO**, I am not doing the same mistake. – aerijman Aug 17 '20 at 14:32
@AndrewK, please, check this answer, since it has been _wrongly_ voted down. Thank you! – aerijman Aug 17 '20 at 14:33
Sorry, been away testing this but gave it a tick now. One thing. np.round in your solution doesn't round 0.0005 correctly to 3dp. Rounds it down rather than up. Other than that it works fine. Any ideas why? and how to correct this? – AndrewK Aug 26 '20 at 11:02

Rounding error across all column values in Python

1 Answers1