In general, it is fairly slow but most people will agree it's worth the wait. for example, once you learn to set up your data frames correctly, you'll eliminate a ton of other fussing. Personally, ggplot saves me tons of time over other graphics packages I've used because of the grammar and the high quality design. Just composing the call is fast for complex data, way faster than most other syntaxes.
I dont really know what the plans for speed improvement are... if any right now. There are certainly people on here who can comment. If I understand correctly, there may be some future code shift from proto to S3, but that is more for profiling than speed purposes??
Anyway, there are a few particular instances when it slows down and which can be avoided, in my experience:
- when there is a factor variable with many levels: move to a continuous variable.
- when computing statistics in the call to ggplot, e.g. a stat function: pre-compute your stats.
Of course, a specific example might be worth offering for any specific guidance...
James