I have an calculated table chart that is very cpu-intensive, so for some users that doesn't need all the information in the chart I would like to lessen the waiting time.
I saw that columns can be hidden conditionally - if they are hidden are they still calculated or not?
I dont know the answer to your question but a better solution might be (if you use section access) to group your users and then have multiple charts that are conditioned to only calculate/show based on the user group.
I also don't KNOW the answer, but I'm confident that the hidden columns are not calculated unless they are used by columns that are not hidden.
First, consider that calculating the columns would be of no use. It would be a very poor design from a performance standpoint to calculate hidden, unused data. So after 9 versions and countless patches, I highly doubt that QlikView is doing so.
Second, in one of my applications, I have a chart with about 100 expression columns. I dynamically hide and unhide these columns, and typically only a few are shown at a time. If I unhide all 100 columns, it takes a VERY long time to calculate. If I just have, say, one simple column showing, it calculates very quickly.
That said, if you have a simpler situation than mine, such as wanting to see columns A, B and C for this group of users, and A, B, D and E for this other group, I probably would implement it with two different charts as Gordon suggests. There's a little dual maintenance of chart expressions, but it's easier than maintaining macro code. Not sure what actions are capable of, though, so hiding and unhiding columns might be more straightforward in version 9.
So. The tests were like this: - one chart with 7 dimensions and 8 expression columns; - the expr. cols. where sum of measurement units (pieces,kilograms,etc); - 8 variables for each expression column (each var with true/false value selectable from drop-down list in multibox); - one fact table (with quantities) and one dimension table (items with measurement units) (practicaly, the formulas were sum(quantity*measure_unit) ); - after each test i've closed and reopend the QV process.
1. I've opened the chart with all 8 expr. cols. visible. Calc. time = 2'47''. Then I started to hide expr. cols., one by one (changing to false the coresponding variable in the multibox). Calc. times were: - hiding the first expr. col. = 2'56'', - hiding the second expr. col. = 3'9'' - hiding the third expr. col. = 3'16'', - hiding the forth expr. col. = 3'24'', - hiding the fifth expr. col. = 3'42''. I've stopped testing.
2. I've opened the chart with 2 expr. cols. visibile and 6 hidden, all 8 enabled. Calc. time = 2'41''.
3. I've opened the chart with 2 expr. cols. enabled and visible and 6 disabled. Calc. time = 1'1''.
So, i might conclude that : - hidding an expression column doesn't imply a calculation is not performed; - disabling an expression column does imply a calculation is not performed.
I'm using ver.9SR1. I hope the tests were done properly.
What each of you have described above, it's also logical in my opinion. But it doesn't work that way. Yet! 🙂
Maybe if you get the same conclusion as i, we could do a request to them.
I couldn't come to a conclusive result regarding my own application when clocking the different alternatives (probably because of dynamic input to the calculations) but I'll go with dual charts this time.
I agree with how you think it should be, maybe for later versions they can stop calculating hidden and unused columns?
OK, looking at my application, I don't hide the columns; I disable them. So my earlier conclusions were not supported by the evidence.
Enabling all columns and changing the macro to merely hide and unhide, I can't seem to get any conclusive results. The time taken varies between a couple seconds and 30 seconds, and seems unrelated to the number of columns displayed, or even how much data I'm summarizing. I'm taking efforts to not hit the buffer, yet it often feels like a buffer vs. no buffer difference. I just can't tell what's going on.
Switching back to disabling columns, everything makes sense again. More columns and more data seems to predictably take more time.
So I'm not sure how much we can conclude from my testing. It would be consistent to conclude that hidden columns still calculate, but that's not the only possible conclusion. In any case, I will certainly continue disabling columns rather than hiding them.
Paul, in answer to your question about how I do it, it's the same dynamic reporting template that I've posted in the past. See attached.
