summarise bug when the number is too big

[antonio-yu]

data  = pd.DataFrame([['xiaomi',80000000000]
                      ,['xiaomi',90000000000]
                      ,['xiaomi',30000000000]
                      ,['huawei',20000000000]
                      ,['huawei',60000000000]
                      ,['huawei',70000000000]
                      ],columns=['brand','sale'])
data >> group_by(f.brand)>> summarize(sale= f.sale.um(),avg = f.sale.mean(),max = max_(f.sale)) 


output 
brand | sale | avg | max
<object> | <int64> | <float64> | <int64>
xiaomi | 200000000000 | 2.000000e+11 | 200000000000
huawei | 150000000000 | 1.500000e+11 | 150000000000

It returns the same results when the column 'sale' is too big . It seems when the number is bigger than 1e+08 , this bug shows.
On ther other hand, the type of avg is scientific notation, but others are normal.

[pwwang]

It's not a bug. The latter f.sale refers to the column created by summarize with the first item, instead of the original column in the original data frame.

See the same behavior in R:

r$> data                                                                                               
# A tibble: 6 × 2
         sale brand 
        <dbl> <chr> 
1 80000000000 xiaomi
2 90000000000 xiaomi
3 30000000000 xiaomi
4 20000000000 huawei
5 60000000000 huawei
6 70000000000 huawei

r$> data |> group_by(brand) |> summarise(sale = sum(sale), avg = mean(sale))                           
# A tibble: 2 × 3
  brand          sale          avg
  <chr>         <dbl>        <dbl>
1 huawei 150000000000 150000000000
2 xiaomi 200000000000 200000000000

To avoid this, either:

1. use a different aggregation name:

data >> group_by(f.brand) >> summarize(sale_sum = f.sale.sum(), avg = f.sale.mean(), max = max_(f.sale)) 
#                                      ^^^^^^^^

2. put the sale column at last:

data >> group_by(f.brand) >> summarize(avg = f.sale.mean(), max = max_(f.sale), sale = f.sale.sum()) 

[pwwang]

For the format of avg, it's because it is a float and others are integers.

To control the format of float:

pd.options.display.float_format = '{:.2f}'.format