How can this happen? TableViewer shows linkage where there's none?
Hi,
I just had something very strange, something that QlikView had never done before:
- I had an app with a master_calendar and one fact_table;
- The master_calendar has a date_field labeled %Datum;
- I loaded one field twice in the other table to have the same name ?=> QlikView linked them just like it should.
Now I added a third table which does have a date_field, but I rename all the fields in the script and it's then labeled "SPORD_%Datum". I will rework that now. But here's what happened:
The table_viewer showed me those three tables were linked through a field called SPORD_%Datum - all three of them.
And I wondered for a while why none of my charts would show any data - until I realized that that field actually exists in only one table, so the tables COULD NOT all be linked.
Here you go. That is what the table_viewer is showing me.
In the master_calender, I don't have that field, however. There is a field %Datum, but not SPORD_%Datum.
This even seems to break the link that did exist between the calendar and the one fact_table - the chart I had build on that data does not show any data anymore.
I'll repair that now, but it sure would be interesting to know how this could happen.
P.S.: Must have something to do with this strange name SPORD_%Datum - when I remove that line from the corresp. LOAD statement, this doesn't happen anymore.
And you are sure that not your script creates those tables/fields like it should - maybe with a forgotten hidden script part or include-variables which are the right one or a must_include after an exit script; statement (it will be always executed unless is it uncommented) or any rename statement - then this would mean that there is an error with the meta-data handling which seems to me rather unlikely. What showed the the log-file what the script does?
And you are sure that not your script creates those tables/fields like it should - maybe with a forgotten hidden script part or include-variables which are the right one or a must_include after an exit script; statement (it will be always executed unless is it uncommented) or any rename statement - then this would mean that there is an error with the meta-data handling which seems to me rather unlikely. What showed the the log-file what the script does?
Well, I think that DataNibbler is in fact renaming fields, see his OP: "... but I rename all the fields in the script..."
It would be interesting to see the script, or even better a small sample app that demonstrates the issue.
yes, I'll try to assemble some fake data and do this if I find the time. Or maybe the log_file would also help you? I can attach that no problem - well, I'll just have to pick one or make one where this issue turns up as I have repaired it in the meantime - should be easy, though, just a few lines' difference.
