Re: [PATCH] random_normal function - Mailing list pgsql-hackers

Any particular justification for placing stddev before mean?  A brief survey seems to indicate other libraries, as well as (at least for me) learned convention, has the mean be supplied first, then the standard deviation.  The implementation/commentary seems to use that convention as well.

Some suggestions:

/* Apply optional user parameters */ - that isn't important or even what is happening though, and the body of the function shouldn't care about the source of the values for the variables it uses.

Instead:

/* Transform the normal standard variable (z) using the target normal distribution parameters */

Personally I'd probably make that even more explicit:

+ float8    z

...

* z = pg_prng_double_normal(&drandom_seed)

+ /* ... */

* result = (stddev * z) + mean

And a possible micro-optimization...

+ bool   rescale = true

+ if (PG_NARGS() = 0)

+    rescale = false

...

+ if (rescale)

    ... result = (stddev * z) + mean

+ else

+      result = z

David J.