BUG:

[koopa31]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

from skimage.io import imread
import pandas as pd

path = "path/to/image"
# Image is uint8
image = imread(path)

list = []

list.append(image.max())

df = pd.DataFrame(zip(*list))
df.to_excel(os.path.join(path, "quantification.xlsx"), index=False, engine="xlsxwriter")

Issue Description

Hi,
I faced a conversion problem trying to store images max in a pandas dataframe and writing it to xlsx file. Indeed, when I read image.max() value, it is a positive integer value equal to 129. When I write it in the excel table, it is equal to -127; which seems to be a conversion to int8 I did nit ask for.
Thank you
Clément

Expected Behavior

To correct that I force the value to be int(max), but is there an actual solution not to make this implicit conversion from uint8 to int8 ?

Installed Versions

version is 1.3.4

[rhshadrach]

Thanks for the report; can you provide a reproducible example? It would be helpful to programatically define df (e.g. df = pd.DataFrame(...)) with maybe just a few rows that reproduces the issue.

[attack68]

This looks likely to be a fundamental error in the construction and skimage package rather than pandas export to excel.

[koopa31]

Actually it is not, as the output of skimage function does not contain those values. It just creates an uint8 image which is not a supported type for the to_excel function.

[attack68]

So the following should fail?

import pandas as pd
df = pd.DataFrame([[1,2], [3,4]], dtype="uint8")
df.to_excel("file.xlsx")

It works correctly on main.

[koopa31]

No because conversion to int8 affects numbers greater than 127. It shifts numbers from 0/255 to -127/+127

[rhshadrach]

I'm still seeing the correct xlsx file using df = pd.DataFrame([[168,255], [3,4]], dtype="uint8") and specifying engine="xlsxwriter"

[koopa31]

Alright probably because I did not specify dtype="uint8" at the end. If you do not specify anything, then uint8 data are not converted properly. Hence, I did int(data).
Thanks for the solution. Maybe, this behavior is a bit dangerous as a user won't know that if he does not specify the type in df, it will change the values.