Okay, I think I'm brain damaged -- nope, I figured it out

[Geoff Canyon]

On Thu, Sep 12, 2013 at 1:18 PM, Richard Gaskin
<ambassador at fourthworld.com>wrote:

> If the final output is going to disk, you may find that opening the output
> file for append yields what I believe is the fastest write mode (CouchDB
> and some others are append-only).  Perhaps buffering some of the output and
> appending in batches may be optimal
>

Absolutely true, and I made this change a few hours later.

Oddly, I found little-to-no difference between open for append vs. open for
write -- I expected append to be faster, as you obviously did.

I also found little difference between large and small batches in the write
command. Very puzzling.