https://bz.apache.org/SpamAssassin/show_bug.cgi?id=7689

--- Comment #5 from RW <[email protected]> ---
(In reply to Olivier Coutu from comment #2)
> (In reply to RW from comment #1)
> > FWIW it was always linear. It's possible to create contrived rule sets that
> > are non-linear, e.g. where each new rule depends all the previous rules. For
> > real-world rules sets there's no reason to think that the mean amount of
> > work per rule grows unbounded as new rules are added.
> 
> I agree that for cases when rules are mostly independant, it is close to
> linear. We just happened to created rules that are highly dependant on one
> another, with hundreds of total dependencies on some. Performance at runtime
> is fine, but lint is getting unbearable.

That doesn't mean it's not linear. It depends on how the average amount of work
per rule scales with the number of rules - the actual value doesn't matter, you
can have 1000 dependencies per rule and it can still be o(n).

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to