Few days ago one of our customers reported the application being slow.
We analyzed the situation and found the bottleneck; in his application, one of summary columns calculated count of records based on another summary column that involved complex calculation. And it was not a single case, small fraction of other applications exhibit the same or similar pattern.
But why you are required to select the column to calculate count of records?
It’s not a secret, TeamDesk uses SQL Server as a database backend. In SQL [Server], the calculating the count over the column takes in account only the records with non-blank values in the selected column. So do we, yet these details were not documented.
We’ve analyzed all the settings related to Count calculation and found out that 0.5% use "count of formulas" or "count of totals" – it is clearly wrong usage pattern, as you can achieve same result by calculating Count of Id; yet latter is significantly faster.
You are required to select the value column for Count and, without clear understanding of the behavior users seem to select random one.
To avoid the misuse we are planning to remove column selection for Count function; this will also streamline setup process a little bit. The Count function will calculate count of related records; if you (still) need count of non-blanks in certain column, you can add "Column is not <blank>" to the filter criteria.
The update will be issued next week, if you have any questions, please contact support -at- teamdesk -dot- net
Ahh, that’s great. I didn’t know that about the Count function on summary columns. The update makes sense!
But why you are required to select the column to calculate count of records?