1 of 1 people found this helpful
Good question - it's very easy to get into the sort of tangle you describe by having multiple workbooks (or even multiple connections in the same workbook) accessing the same extract.
I can't answer the questions about how server behaves - we only have desktop, but I know that version 7 has added a lot of new capabilities for sharing central definitions of connections with server, so that may help.
For your example 2, duplicating the connection actually makes a clone of the extract file itself these days - so changes in one connection will not impact the other one - it will be up to you to copy calculated fields across and you can then optimise / regenerate each extract independently. That may be the simplest, but the time and/or space taken to make a second copy of the extract may not be ideal.
The other option is to define the first connection and create the extract, then define a second connection which connects directly to the extract. Then just make all changes in your first connection and refresh or optimise the extract from that connection. If you now hit refresh (i.e. F5, not refresh extract) from a worksheet using the second connection Tableau will pick up the changes. The new calculated fields appear with auto-generated names like: "Calculation_9770307094358026". This is the version 7 behaviour, I can't remember seeing the names in that format in 6.1, so I'm not quite sure what happens there.
Thanks Richard. I was experimenting today with duplicating the connection and you're right, it just creates a clone, so doubles the size of the workbook. I will try the other option tomorrow, but it sounds like this will mean having to do all of the calc field creation in the first connection, which isn't the connection to the extract. Perhaps I'm wrong about that, will check it out in the morning.
I was at one of the Tableau 7 roadshows and from what was said it seems like 7 handles the extract and server side of things better. Shame my company don't use it yet :-)
It would be good to have some good server 'best practices' written by the people at Tableau as it's causing lots of pain where I work now and there's very little good information available online.