4 Replies Latest reply on Sep 13, 2013 12:36 PM by Jay Chang

    Extracts vs Data Sources

    Jay Chang

      Does anyone have a good explanation of how Tableau extracts work when they are used as saved data sources?  Here is my situation.


      I have an Access table that is an amalgamation of several support tables linked to a primary data table through ODBC.  Because my primary table is not incredibly user friendly, I create a master table that simplifies and enhances the data contained in the primary.  I then connect this primary table to Tableau.


      When I bring this table into Tableau, I want to do a number of Tableau-specific customizations to it (adding hierarchies, aliases, standard calculated fields, formatting, default sorting, default aggregations, etc).  Some of these I could do outside of Tableau (calc fields being the notable one) but others are possible only in Tableau.  A short while ago, I came up with what I thought was a brilliant idea - I'd do all the customizations once and then save this extract as a data source that could be used as the source of other vizes.


      Based off a question by Jonathan Drummey, it seemed that I needed to open my extract, manipulate it, then save it as a data source.  Importantly, Jonathan indicated that you could not manipulate the extract while other Tableau files based off that extract were open.  So what I've been doing is opening the extract, manipulating it, saving it, and then saving it again as a data source.  Then I close the file, open up another file, and replacing its connection to the data source with a new connection to the updated (same) data source.


      For the most part, this seems to work.  However, there are maddening instances where I've made changes in my source extract, saved it as a data source (with the same name) and used that modified data source in a new or existing viz.  When I check the changes, however, I do not see the changes made to the source.  Sometimes they're there, other times they're not.  This seems to especially be true when it comes to custom groups, aliases, and default sort orders.


      Does anyone have any good references to how I can create a standardized data extract/data source that can be used across multiple vizes?  Is what I'm trying to do even possible/reasonable?

        • 1. Re: Extracts vs Data Sources
          Matt Lutton

          It has been my experience that published data sources are great for data modeling, and giving access to data sources that include your manipulations with other desktop users.  However, the ability to control multiple workbooks from a published data source is limiting in my case, because when I connect to it, it is a live connection and performance issues result.  So, I end up duplicating extracts for similar workbooks, and publish them embedded with the visualization.  Another issue is that I have created many workbooks, and if I try to rebuild them using a published data source, I have to rebuild my filters, etc.  These issues may or may not exist in your case, I'm just sharing my experience thus far. 

          This is just my (albeit, limited) experience, but I hope they find a way to improve this in future releases.  It would be great to point multiple workbooks to an extract that performs like an extract.

          • 2. Re: Extracts vs Data Sources
            Jay Chang

            Could you elaborate on what you meant by the comment "it is a live connection"?


            In terms of rebuilding filters etc., that does not seem to happen as much.  Occasionally I'll get weird errors about fields existing already which cause me to have to refresh my workbook's extract.  But I've not noticed that existing filters having to be re-created.  If that were the case, I'd totally not pursue this line of experimentation.

            • 3. Re: Extracts vs Data Sources
              Matt Lutton

              Sure, when you publish an extract and commect a new workbook to it, you connect live. This makes sense, but unless I create a local copy of the extract, with my workbook, performance is very slow. However, this may not be an issue for you.


              I am not sure if I was clear about rebuilding filters. This only occurs when I've created a workbook using a different data source and replace that source with a published data source on Server. Even if the two extracts are identical, replacing the data source causes me to have to rebuild my filters. They still show up, but I have to choose the style and list type to rebuild them to the exact specs I want.

              • 4. Re: Extracts vs Data Sources
                Jay Chang

                Oh right ok that makes sense.  My performance hit is less because I'm connecting locally.  Not sure how it's going to suffer when I publish my workbook to the server.


                I'm wondering if I'm running into less of the filter issues because I'm doing it all locally and my data sources are also local.  I could understand Tableau getting confused if I use a local extract then switch to a server extract, even if they are identical.