> From: "Bill Stoddard" <[EMAIL PROTECTED]>
> Sent: Thursday, August 16, 2001 3:33 PM
>
>
> > > > My issue is that table merges are a o*n problem, while hash merges are an o*n2
problem :(
> > >
> > > Hash merge is O(mn) in the worst case (the case when your hash function
>completely
sucks
> > > :-). The merge will tend to O(mlogn).
>
> That varies based on the number of directory merges, as well, and how large the base
table is.
>
> > Okay, I've thought about this some. Your veto of my patch is unreasonable.
>
> Hmmm? Perhaps, but I committed effectively the same code an hour ago, because I'm
>not
going
> to defend my veto, nor contribute any further to this clueless discussion.
Sorry, I discovered this after I sent my note. I -still- have not received the CVS
notification of the change you made a couple of hours ago :-(. I don't catorgorize the
discussion as 'clueless', but whatever...
<snip>
> Either which way, I don't have any
> more energy for this, it's sapped a good week of my life, as it stands, on an issue
> (table -> hash) that I had 0 interest in :(
>
Ditto. I thought Brians patch was pretty good. I missed the dir_merge problem during
review. My bad and I've learned a good lesson. I appreciate your effort to fix Brian's
patch but I spent a -lot- of time debugging the damn seg fault that resulted. I don't
blame you or anyone else. Software development is about learning from mistakes. You
are a
good developer if you don't make the same mistakes more than once :-)
Cheers,
Bill
