I recall having a similar problem with version 8.1.6 and it was resolved by upgrading to 8.1.10
Thanks for your reply.
Currently I do have Tableau 8.1.6 version. Let me try to change/upgrade to Tableau 8.1.10 and check it.
The workbook you posted is not the workbook you are having problems with, so it's more difficult for me to help, especially since the problem is not duplicated. Besides the difference in data sources, the data in the posted workbook is over-simplified and missing the dimensionality that is in the screenshots, which may be the cause of the error message.
First of all, try the upgrade, there have been a bug fixed that's not covered in the release notes. However, I'm doubtful that will fix the problem based what is shown in the screenshots (please let me know what the outcome of the upgrade is). There are three secondary dimension filters on PO Cancel Flag, PO Delete Flag, and Supplier Name. If you remove those filters, I'm guessing the COUNTD() will work. This is because of the following criteria:
1) Secondary dimension filters trigger DB2 (for Data Blending 2) blending. This is the new data blending introduced in Tableau v8 that lets have secondary dimension filters and blend on dimensions that aren't in the view.
2) DB2 blending doesn't support non-additive aggregates like COUNTD() from secondary sources.
Assuming that's what's going on, there's the question of how to work around this. There are a few possibilities:
1) Flip the blend so the primary is secondary and vice versa.
2) Use data source filters instead of the Filters Shelf. Data source filters are processed before blending, so don't trigger the DB1/DB2 distinction.
3) Use aggregate measures instead of filters, for example by using ATTR(Dimension). If you're getting * for the dimensions, then you'd need to increase the granularity of the view with additional dimension(s) on the Level of Detail Shelf and then use table calculations to re-aggregate at the appropriate level of granularity. This will be the most complex to set up.
If your users don't need to be setting any filters, or you can get away with a single-select parameter for filters, then I'd suggest #2, it'll be the fastest to set up. Otherwise #1 would be my second choice, with #3 the third due to the added complexity involved.
Upgraded to Tableau 8.1.10, but still have the issue.
Ok, try removing the three secondary dimension filters on PO Cancel Flag, PO Delete Flag, and Supplier Name. When you do that, does the COUNTD() measure work?
I have removed all the secondary dimension filters, even then it is throwing same error.
There have a been a couple of times where once the error is triggered it doesn't go away that easily. Try building the view from scratch without any secondary dimension filters at all and see what happens.
If that doesn't work, then you're going to need to expand all the groups of Dimensions in your secondary source and see if there is more than one linking field there. Based on your screenshots, there shouldn't be, but it's worth checking.
If neither of those work, then I don't think I'd be able to help you further here, and I'd suggest contacting Tableau tech support.
As mentioned by you, I have created a view from the scratch without any secondary dimension filters on the view then it works fine and I am not seeing the error any more.
Thanks for you help and suggestion.
Excellent! Is there any more help you need on that particular view?
On Tue, Sep 30, 2014 at 2:33 PM, Mahesh Chinthalapalli <
I need to plot a combination chart with PR and PO ($ Amt) on Primary Y-Axis and Counts on Secondary Y-Axis. However Tableau does not support Combination charts. Can you provide me any suggestion on this.
I'm not sure what you mean by "However Tableau does not support Combination charts." You create a dual axis chart with different mark types for each axis, here's a quick primer:
In cases where we want multiple measures on a single axis, we can use the Measure Names/Values "unpivot":
However, your case is trickier because the goal is to use two measures for one axis and two for the other, and we can't use two instances of Measure Names/Values in the same view. The workarounds are to either restructure the data so it is "taller" (i.e. instead of having each measure be a separate column, use one column for the measure value and another column to name the measure) or to use a scaffold data source & a data blend. In http://drawingwithnumbers.artisart.org/bars-and-lines/ I went through some ideas on how to restructure the data, in the attached workbook I created a simple scaffold source that has a record for each measure & quarter, then used data blending to build that:
One likely issue with the scaffold source is due to using COUNTD() as a measure. To avoid DB2 blending and it's limitations (no COUNTD() on secondary sources) then you'll have to make sure the primary source has all the dimensionality that you'd need for filtering (to avoid using dimension filters on secondary sources that would trigger DB2 blending), and then to preserve DB1 blending you'd need to have all those dimensions in the view, and to deal with that increased dimensionality then you'd need to use table calcs to aggregate over that to get back to the correct results. It's possible, but not easy, so I'd first try the restructuring of the data.
Thanks for your quick reply and example workbook provided. Unfortunately my data model do not have a dimension or column which tells the name of the measure. I am trying to come up with the suggestion provided by you.
Again thanks for your suggestion/help.
I understand that your data as it comes out of whatever system doesn't have
a dimension or column, you'd have to effectively alter the model for
Tableau, for example by using a custom query to unpivot your data.
On Wed, Oct 1, 2014 at 3:04 PM, Mahesh Chinthalapalli <