8 Replies Latest reply on Jan 31, 2017 4:01 AM by Dovi Lilling

    Union two identical extracts

    Al Reynolds

      Hi Guys

       

      I have read a few forums similar to my situation but cant seem to get a definitive answer out of it.

       

      Due to performance constraints outside of Tableau i have been forced to create multiple (identically formatted) data extracts (which are actually from the same MYSQL database) The reason for this is long winded, but in summary, the volume of data, the number of clients, frequency of updating and performance related issues extracting the same data over and over again when not required.

       

      So i have a Tableau workbook now with multiple connections to an identically formatted dataset. (for different clients obviously)

       

      All i want to do now is UNION ALL of these extracts the tableau side and use in my analysis so i can compare data across all clients.

       

      Is this possible? I see its possible in Excel worksheets.

       

      Al

        • 1. Re: Union two identical extracts
          Hrvoje Gabelica

          Why don't you write Custom SQL for those two sources and make an extract ?

          • 2. Re: Union two identical extracts
            Al Reynolds

            Hi Hrvoje, thanks for the response

             

            So i now have 4 identical extracts in my workbook. I want to combine them with 'UNION ALL'. I have limitations doing this on my database server so cant do a custom SQL to my database. Is there a way to create a new extract on the 4 i already have?

            • 3. Re: Union two identical extracts
              Thomas McCullough

              You can append extracts that have the same schemas. Works similarly to UNION, but is a big pain.

               

              Tom

              • 4. Re: Union two identical extracts
                Jonathan Drummey

                Sometimes you can get the effect of a UNION by using a "scaffold" source that has the right dimensions (and nothing else) that becomes the primary source and then blend in the additional sources. If you're doing multiple extracts because the DB can't handle the load then your volumes are likely to be too large for a Tableau data blend, those top out at around 100K linking values in my own use (beyond that and the view refresh times start getting too slow). So the only option I can think of at this time is what Thomas wrote.

                 

                At the Tableau Conference last month they demoed UNION'ing multiple data sources, so I'd expect to see that as a new feature sometime in 2016.

                 

                Jonathan

                • 5. Re: Union two identical extracts
                  Alexander Mou

                  Having tons of data for tableau to visualize is always a bad idea. Human

                  eyes can only see so much at a time. It is further aggravated in the era of

                  ADD/ADHD with short attention span.

                   

                  In case of large volume of data, it's preferred to pre-aggregate them in

                  database and create a view with only the necessary data for visualization.

                   

                  On Tue, Nov 10, 2015 at 6:35 AM, Jonathan Drummey <

                  • 6. Re: Union two identical extracts
                    Jonathan Drummey

                    Hi Alexander,

                     

                    I think you and I are talking about different things. The number of marks in the view (from your post) and the number of linking values aka "cardinality of the blend" aka number of blended values (from mine). I'm with you that 100K+ marks in a single view is likely to be non-useful, though I have seen views from Allan Walker (maps) and Richard Leeke (network traffic) that were drawing that many marks and had useful information being displayed.

                     

                    What I was referencing was that in order to pull off a "scaffold + data blend for union" the number of distinct values of the linking dimension(s) may be so large that the time Tableau takes to do the blend as part of refreshing the view (no matter how many marks are actually displayed in the view) is too long for what is desired.

                     

                    Jonathan

                    • 7. Re: Union two identical extracts
                      Alexander Mou

                      Jonathan,

                      When replying via email, the forum always makes it look like addressing the

                      last reply.

                       

                      What I was trying to say is a general comment to the problem in this

                      thread.

                       

                      Regarding the vizzies you mentioned, I wonder what are the minimum number

                      of marks to get the same insights? Can we subsample the dataset to get the

                      same result?

                       

                       

                      On Wednesday, November 11, 2015, Jonathan Drummey <

                      • 8. Re: Union two identical extracts
                        Dovi Lilling

                        You may also want to upvote my idea regarding this: https://community.tableau.com/ideas/7076