Hi John --
I just tried this out and can see the issue, too.
I'd open a case with support.
It appears to be happening because of some extra optimization work we're doing to make those filters of yours run & display more quickly - If you export your data source as a TDS and optimize from inside the new TDS, you get "one time" growth (expected if your calculated fields are not yet materialized), but you don't see the TDE growing over and over again.
If you remove ALL the filters in your current workbook and optimize, no "over and over again" file growth
If I was in a Holiday Inn Express commercial pretending to be a developer or someone from support, I'd tell you that it seems we're not cleaning up that old filter optimization information when you execute multiple "optimize" actions. But I'm not in a Holiday Inn Express commercial so I won't.
Very cool viz, btw.
1 of 1 people found this helpful
This looks to be closely related to an issue I kept reporting during the v8 beta, but which I'm pretty sure didn't get sorted. The time taken to optimise a large extract is substantially longer than in v7 (because of the new filter optimisation logic, I presume) - but what is much worse is that you incur all of that overhead every time you optimise, rather than just a one-time hit. I often want to materialise a new calculation I've just added. On a 100 million row extract that I do a lot of work on that used to take a matter of seconds (depending a bit on the calculation). Certainly the time taken to optimise was often much less than the time taken to refresh a view using the unoptimised calculation.
But with v8 the optimisation might take 10 minutes or more - and as I say you get that every time. At one point during the beta it even took the same time if you just optimised the workbook repeatedly without making any changes to filters (or anything at all in between). I think I recall that that no longer happens, but I'm (fairly) sure you still take the hit if you have changed anything in the use of filters. Maybe I'd better go back and re-test that after making that assertion, it's a wee while since I've needed to do that on anything big enough that it really hurts...
Thanks for looking at this Russell! Is this also related to the file size difference of refreshing the ALL extracts vs refreshing the sole extract directly?
I really miss the summary results that use to be provided of the optimization like x number of calculations optimized, etc.
I published this Viz in Public with some more graphics.
Russell - Can you possibly be a lifeline on a similar issue?
Photo thumbnails look like garbage here: https://public.tableau.com/profile/joshpreston#!/vizhome/ResearchConferenceTEST/Dashboard1
Photo thumbnails look fine here: https://public.tableau.com/profile/joshpreston#!/vizhome/GTComputingintheNewsFY2015_v2/Dashboard1
The first viz gets bigger each time I add photos; it's now at 8.1 MB (The whole folder of pics is 1.4. MB (each pic averages 14KB))
Any way that the pic quality will improve if viz size is reduced? It's attached. thanks much! Josh