BUG: read_excel doesn't respect string data

[chris-b1]

From SO

In [16]: df = pd.DataFrame({'a': ['001','002']})

In [17]: df.to_excel('temp.xlsx')

In [18]: pd.read_excel('temp.xlsx')
Out[18]: 
   a
0  1
1  2

I think it would probably make sense for read_excel to not try and convert strings to numeric, or at least have another keyword argument.

[jreback]

xref #8212

this should be done via the dtype kw (and not s converter) as it's more consistent with how the other parsers work

and it should be coercing normally as that is the expected behavior

[chris-b1]

I guess my point is that (unlike csv, etc) Excel numeric data will already be de-serialized as a python numeric type, so the only data that will be strings are those explicitly stored as strings in Excel. So a different default could make sense?

[jreback]

zipcodes are usually stored as integers with a format
not sure u can detect this (but maybe it's not being taken into account)

[chris-b1]

Right, if that were the case I think the data should still be read as integers. In this case the zipcodes were stored as Excel text (i.e. if you typed '00500)

In [25]: ws = xlrd.open_workbook('test.xlsx').sheet_by_index(0)

In [26]: ws.cell_value(1, 0)
Out[26]: 55.0

In [27]: ws.cell_value(2, 0)
Out[27]: u'00500'

[jreback]

ok then might be a bug then

[stevemaughan]

Zipcodes are usually stored as strings. You also have zip+4 which also need to be stored as strings e.g. 32771-5407

[jreback]

dupe of #8212