Date: Sun, 15 Jul 2001 18:26:51 -0600
From: "Roger S. Miller" <[email protected]>
X-Mailer: Mozilla 4.76 [en] (X11; U; Linux 2.0.27 i586)
X-Accept-Language: en
To: Markus Neteler <[email protected]>
Subject: r.drain -- new version

Markus,

I worked on the original r.drain for a while and wasn't really able to
understand what it was doing, much less where the problem was.  I
figured that I might be able to spend the rest of the weekend figuring
it out and in the end I would still have a program with unusual behavior
in flat spots.  I decided instead to finish the version that I started
based on r.fill.dir.

My version of r.drain (which I called r.d2) is in the attached zip file.

r.d2 reproduces most of the behavior of r.drain.  It does not use the
"null_value" option and it won't read input from a sites file.  It uses
all of the flags defined for r.drain.  My first draft of the model
didn't correctly handle the ew,ns and diagonal resolutions (as
r.fill.dir does not).  I believe that I fixed that problem.  There are
also some other minor behaviors that �differ a little from r.drain.

r.d2 runs pretty quickly on a small, clean dem, but it can take quite a
while  to run on a large cell map - particularly one with very many
areas where the flow directions are ambiguous.  I didn't write anything
into the program to warn the user about the execution time.

I tested it on the fcell map that you sent me when this issue first came
up and on a lat-long cell map.  It worked well in both cases.  The
version of r.fill.dir that I started with was the version I had archived
here, which was not updated for all the changes needed to make
r.fill.dir run on all of our platforms.  I built in the changes that I
could remember, but there may be a few more that come up when you try to
compile it on the Cray.

I hope this meets your needs.


Roger