read_excel: Wrong dtypes for numeric text fields

[suokunlong]

In the following example, df column a are exported as text (string) in xlsx file, but imported back to df2 as int.

In [161]: df = pd.DataFrame({'a':['01','02','03','04'], 'b':[5,6,7,8]})

In [162]: df
Out[162]:
a b
0 01 5
1 02 6
2 03 7
3 04 8

In [163]: df.dtypes
Out[163]:
a object
b int64
dtype: object

In [164]: df.to_excel('tmp.xlsx')

In [165]: df2 = pd.read_excel('tmp.xlsx')

In [166]: df2
Out[166]:
a b
0 1 5
1 2 6
2 3 7
3 4 8

In [167]: df2.dtypes
Out[167]:
a int64
b int64
dtype: object

Users usually set columns (which contain numeric data) as "text" in excel for special purposes, pandas should keep these dtypes as str, please do not try to convert it back to numberic. At least, pandas should provide an option in pd.read_excel() to switch this dtype conversion.

[jreback]

this is a dupe of #8212