display np.int64 numbers as integers, do not apply 'floatfmt'

[jonsel087]

This is more a result of how pandas data frames are stored internally, but the issue comes up when you tabulate any dataframe that contains float type columns. When the df is converted to a numpy array, numpy will upcast all int values to floats.

from tabulate import tabulate
import pandas as pd
df = pd.DataFrame({'A':[1,2,3,4],'B':[.20942, .092384, .019823, .01923]})
print(tabulate(df, floatfmt='.2f', headers=df.columns, tablefmt='pipe'))

|    |    A |    B |
|---:|-----:|-----:|
|  0 | 1.00 | 0.21 |
|  1 | 2.00 | 0.09 |
|  2 | 3.00 | 0.02 |
|  3 | 4.00 | 0.02 |

This can further be shown that it's caused by internal storage of np.int64 dtype:

print(tabulate([['int','float'],[np.int64(1), 3.14159]], headers='firstrow',tablefmt='pipe', floatfmt='.2f'))

|   int |   float |
|------:|--------:|
|  1.00 |    3.14 |

Is there a safe way to convert int64 types to vanilla int when parsing dataframes? Maybe there is a workaround i'm not thinking of.

PS - Note that np.int64(1) == int(1) evaluates to True

[MoritzMaxeiner]

Hey @astanin, I appreciate the package, but just wanted to let you know that pandas DataFrames np.int64 numbers are still evaluated as float when there are other float columns. This is because this call to values causes numpy to upcast all integer columns to float columns. Ideally instead of just using the values, each cell would be considered individually.

I'm using a quick and dirty hack to get around this by reapplying the appropriate dtype to each cell afterwards here:

rows = [list([tabular_data.dtypes[tabular_data.keys()[i]].type(cell) for i, cell in enumerate(row)]) for row in vals]

This is probably unsafe in ways I don't know, but it does allow me to print the issue's primary example correctly (without this hack it still prints the output the original reported posted):

from tabulate import tabulate
import pandas as pd
df = pd.DataFrame({'A':[1,2,3,4],'B':[.20942, .092384, .019823, .01923]})
print(tabulate(df, floatfmt='.2f', headers=df.columns, tablefmt='pipe'))

|    |   A |    B |
|---:|----:|-----:|
|  0 |   1 | 0.21 |
|  1 |   2 | 0.09 |
|  2 |   3 | 0.02 |
|  3 |   4 | 0.02 |