The new excel reader nodes brought up two issues that have more or less quite some impact on our productive daily workflows in my department. Hence we stay with the old reader nodes for the time being, but here are the issues:
The reader brings up the message “Execute failed: The following row could not be converted to the specified KNIME types: …” and then lists every row entry with its type. This issue can be solved when unchecking the “Limit data row scanned”:
But of course that’s is not too pleasant. With the generic selected 50 rows, it just doesn’t work.
We are using SAP output from a SAP BW, and that is unfortunately an SAP problem, but in SAP you can select to show a key and a description for a column. What SAP does is, that the key column still has the original column header and the description column comes with no header at all.
In the old Excel reader world, this empty column would be named column 2, or whatever the position of the empty header column is. Now it seems that it is not the numeric position anymore, but the actual alphabetical colum from the excel universe. See screenshots below:
Old Excel Reader:
New Excel Reader:
The new Excel Reader, hence, will lead to several adjustments in column renames, or partly in our logic that we built around this SAP bug and followed the renaming convention of the old reader nodes. Is there a setting, that would allow the old reader node logic again?
Thanks for your support and keep up the good work!
thanks for this feedback and sorry for the trouble.
Regarding your first point, that is a bit tricky as in many cases scanning 50 rows is actually sufficient and would then save a lot of time as we do not need to scan the whole file. However, I understand that in the other cases this can get annoying. Just a question to get more feedback: would it be less annoying having this option in the (first) Settings tab?
Regarding your second point, I think we could add further options here. I think a setting for the prefix (which is currently “empty_” and was “Col” before) and one to decide between alphabetical and numerical numeration would work for you, correct?
Which problem do you mean? I think the node outputs a path variable that should work with the reader nodes.
Thanks for the quick reply Simon!
Yes, I think to have the row scan on the first tab would help, as it would not be too intuitive to go to tab3 here. I also figured out, that after approximately 500 rows to scan, the datatypes are read in correctly and the node runs through. But I guess this strongly depends on the dataset…
About the numerical position rename, I guess an additional option would help here. I assume that our SAP BW is not the only one that produces this bug, which a lot of users had to adapt to in KNIME.
The path variable issue emerged as follows:
I configured “local file browser configuration node”:
The output now is however not the “unique” variable “input-file” as specified, but a bunch of variables:
This lead to the problem that all excel reader nodes, the variable had to be reinserted, and the configuration had to be changed a bit, as now only the “file-input (URL)” was accepted. So I had to configure the Excel Readers to “read from custom URL”:
Thanks again for the quick feedback and cheers,
There is unfortunately not a way around this if you want to replace all deprecated nodes with the new one. The new path variable contains additional information, such as the file system. Before, you would have needed several flow variables to control a node to be able to use paths of different file systems. However, once you are making use of the new variables, they should hopefully be more handy to use.
Thanks again for all the feedback. We’ll work on improving the things you mentioned and let you know if there are news!