Sean Leyne wrote:

>So I've had a quick look at the definition of the AVG function, so
>instead of 20 source code references where down to about 10 (if memory
>serves). Most of the references are in syntax parsers (there is, of
>course, more than 1).

There are (some) good reasons for having more than one parser -
ISQL and DSQL have very different requirements for their output,
for example.

>What I found interesting is that I didn't actually see where the
>implementation of AVG is actually coded.

Look in evl.c (evaluate). The routine is eval_statistical.

Regards,

Ann