Hey hi guys, I'm working on a tricky 2 way selection solution, a 2 list box drill down selection if you will.
The concept is really simple, but is not very Qlikview friendly...
I want to make multiple selections in the SAME field, using sub setting, or filter grouping.
More specifically, I have survey questions and answers, and I want to filter respondents by different answers people entered.
So STEP 1 is selecting the question, then select the answer to filter (the actual selection). STEP 2, is selecting the question #2, then select the answers to filter further (the 2nd actual selection).
Now, I've develop a little prototype with listbox expressions (here is attached), and the display work well, but the behavior doesn't. I guess since I'm using expressions, each time I make the second selection (even if I hold either CTRL or SHIFT), it clears the first one. Whereas if I make the SAMEselections using the "WANTED BEHAVIOR" listbox (regular listbox), it's working.
Quick note: Tree view is out of the question (that's what my customer use right now, heh).
I don't know if this is a good approach for your situation, but it's something I've used before. First, assign a unique ID to every record. Then create "add to memory" and "clear memory" buttons. The add to memory button should select and lock possible IDs. The clear memory button should unlock and clear the ID field.
Now what you do is you select your first question and your first answer, and hit add to memory. Select your second question and your second answer. You are now looking at all the records with the first answer to the first question AND the second answer to the second question. This can be extended indefinitely. Unfortunately, it's non-intuitive, slow to use, slow to execute, and plays poorly with bookmarks unless your record IDs are consistent (and not reassigned on every load by recno() for instance).
Another thing to consider would be transforming each question into a field, sort of an inverse crosstable. Then you can select answers to all questions at the same time, which would be much faster, but might well play havok with the rest of your data model and charts.
We have found a simpler approach but it hits the perfo a bit... but it works better with the users so to speak. This is simply loading the table 3 times (for 3 different filter) and linking them thru a unique respondent identifier.
The only thing that worries me is if there is a "performance threshold" that can be hit in matters of "linking power"... for example, if I have 8 fact tables with 150M records each that link to EACH OTHER via one single field... is there a limit to this... or maybe at one point... it just doesn't make sense anymore (performance wise)... hmm.
And thanks for the suggestion John, but as you said, not very user friendly.
.png "fernandotoledo"){kind=avatar}
