1

Hello, In a Mendix webinar from 2014 , optimization of not()- and or-constraints is being discussed in XPath retrieves. It is mentioned that it is advisable to create separate DB-retrieves for each over-association OR-constraint, and use a Union-action to combine the lists. I found this very interesting and created a small experiment. I created a very simple domain model: As for data: I created 25.000 Main-objects, and 500 Associated1 and Associated2 objects (each of which is associated with 50 Main-objects). Next, I created 2 microflows that measure the duration of the different retrieves. The first microflow had 1 retrieve with the following XPath constraint: The second microflow had the same constraints, but each of them was put in a separate retrieve action.  A Union-List action was added as a third action in this microflow. Both of these flows were run 2.500 times, and the average duration for each option was calculated. The end result was that the combined retrieve took an average of 13ms, and the separate retrieves combined with the union action took an average of 3ms. This is a very large difference, which really surprises me. My question is whether I perhaps forgot other influences which might cause these differences in the way I performed my test, and second, if anyone else has any knowledge about splitting retrieves this way, or has experience using this method. Thanks in advance, and sorry for the long read. Greetings Martin

asked 2021-04-21

2 answers