Error on reading files where <c> element misses the "r" attribute

[fabiospiga]

I need to read a file where the element <row> and <c> miss the "r" attribute, that is apparently optional in the OpenXML structure.

Here's an example:
https://www.atih.sante.fr/sites/default/files/public/content/3968/fichier_complementaire_ccam_descriptive_a_usage_pmsi_2021_v2.xlsx

The raised exception is:
Cannot invoke "java.lang.Integer.intValue()" because the return value of "org.dhatim.fastexcel.reader.SimpleXmlReader.getIntAttribute(String)" is null

because int rowIndex cannot be unboxed at org.dhatim.fastexcel.reader.RowSpliterator#next

Could you please provide support for this use case?

Thanks and best regards,
Fabio

[RamsesGomez]

yes I have the exact same problem

[ochedru]

You are right, the r attribute is optional. We should check how this is handled in Apache POI. For example, what happens if some elements miss the r attribute, but not all?

[ursjoss]

We have such a case as well. Interestingly, opening the seemingly broken file with Excel or libreoffice calc and saving it again results in fastexcel successfully processing the file. Looks like those programs will add the optional r attribute.

[ursjoss]

Opening the sample file provided by @fabiospiga and saving it also seems to fix the file in a way that let's fastexcel process it successfully (same as with our file). The FastExcelReaderTest that compares certain attributes of the parsed file when read with fastexcel with the same file being read with apache poi.

PR #514 fully fixes the issue we have experienced with our file. Adding it to the FastExcelReaderTest passes (after increasing byte array max override enough). However, the file provided by Fabio does not pass that test, even with the changes in PR #514. But also there, the initial exception is gone:

Cannot invoke "java.lang.Integer.intValue()" because the return value of
"org.dhatim.fastexcel.reader.SimpleXmlReader.getIntAttribute(String)" is null

I investigated to some extent to apply further fixes to let fastexcel process that file with the same outcome as with apache poi but I stopped.

I wanted to add a file to the FastExcelReaderTest but therefore failed to do so, as our file contains confidential customer data, and Fabios file still does not pass the test.

Maybe somebody else could contribute a file that can be added to the PR as a test case.