Question

Does the VelocityEngine instance have a small enough footprint to allow for construction of a new instance every time I need to process a template, or is it best to cache the VelocityEngine and reuse the instance?

Answer 1

I think it usually won’t make a big difference, for a project I implemented I made it configurable, but in the end we did not cache the object, but just re-generate it each time without any performance problems.

Note that in my case the velocity processing is heavy, i.e. the rendering itself takes considerable time. For smaller payloads this might be different. But as Velocity works with reflection internally, it will always cause some overhead for it’s rendering and thus I would expect that the creating of the initial objects should not take much time compared to this.

I would therefore not spend too much time implementing something here unless performance analysis indicates that it causes trouble. General rule of thumb, measure what uses up most of the time, then try to fix it!

Additionally, as far as I remember (could not find the reference just now), the VelocityEngine should be thread safe per definition. But there have been bug-reports in the past about multi-threading problems (all fixed now!) and so creating a new Engine avoids possible future threading problems.