BUG: multi-index excel header fails if all numeric

[Dr-Irv]

If the multi-line headers come from Excel, and the header was not a string, the line all(['Unnamed' in c[n] for c in columns]): will fail because c[n] is an int and not iterable. So either force the headers to be strings (the proposed change) or come up with some other test when looking for 'Unnamed'

[jreback]

you need a test where this is actually failing

[Dr-Irv]

For some reason, I can't submit the simple XLSX file that illustrates the problem, or a CSV representation as a file. Here is the CSV version:
Id,A,A,B,C
is,1,2,1,3
1,13,23,33,43
2,24,44,64,84
3,35,65,95,125
4,46,86,126,166
5,57,107,157,207
6,68,128,188,248

Read this into excel, and save as a file "testmulti.xlsx"

Then this fails:
import pandas as pd
df=pd.read_excel('testmulti.xlsx',header=[0,1], index_col=0, parse_cols=4)

[jreback]

so add that as an example in test_excel.py. It should fail w/o your fix and pass afterwards

[Dr-Irv]

I had to create a new pull request with all the files that illustrate the bug and that the bug is fixed. pydata#11328

[jreback]

replaced by #11328