From: Thrun Robert IHMD (ThrunR_at_ih.navy.mil)
Date: Mon Feb 12 2001 - 17:45:34 CET
Received: (from mdom_at_localhost) by karto.ethz.ch (8.9.3/8.9.3/SuSE Linux 8.9.3-0.1) id RAA09122 for cavexml-outgoing; Mon, 12 Feb 2001 17:46:05 +0100 Received: from ih-router.ih.navy.mil (ih-router.ih.navy.mil [205.68.95.65]) by karto.ethz.ch (8.9.3/8.9.3/SuSE Linux 8.9.3-0.1) with ESMTP id RAA09118 for <cavexml_at_cartography.ch>; Mon, 12 Feb 2001 17:46:04 +0100 Received: from ih-router.ih.navy.mil (root_at_localhost) by ih-router.ih.navy.mil with ESMTP id LAA26374 for <cavexml_at_cartography.ch>; Mon, 12 Feb 2001 11:45:15 -0500 (EST) Received: from ihmdex04.ih.navy.mil (ihmdex04.ih.navy.mil [206.39.129.72]) by ih-router.ih.navy.mil with ESMTP id LAA26370 for <cavexml_at_cartography.ch>; Mon, 12 Feb 2001 11:45:15 -0500 (EST) Received: by ihmdex04.ih.navy.mil with Internet Mail Service (5.5.2650.21) id <1BC0STR5>; Mon, 12 Feb 2001 11:45:40 -0500 Message-ID: <9718D3B1ED18D31180F000A0C99DE22301F7E94F@ihmdex03.ih.navy.mil> From: Thrun Robert IHMD <ThrunR_at_ih.navy.mil> To: "'cavexml_at_cartography.ch'" <cavexml_at_cartography.ch> Subject: RE: Other areas that haven't been discussed. Date: Mon, 12 Feb 2001 11:45:34 -0500 X-Mailer: Internet Mail Service (5.5.2650.21) Content-Type: text/plain; charset="iso-8859-1" Sender: owner-cavexml_at_karto.baug.ethz.ch Precedence: bulk Reply-To: cavexml_at_cartography.ch
On Sunday, February 11, 2001 5:15 PM, John Halleck said:
> To: 'cavexml_at_cartography.ch'
>
> On Fri, 9 Feb 2001, Thrun Robert IHMD wrote:
> On Thursday, February 08, 2001 7:52 PM, John Halleck said:
> >
> > > Identification of a spanning tree.
> > > Most graph programs I've seen (and therefore I assume cave
> > > survey programs) find loops by working from a (minimal)
> > > spanning tree. This takes processing to find, but one only
> > > needs to note if they are in the tree or not...)
> >
> > Any program that uses a spanning tree will generate its own
> > using the survey data.
>
> As a statement of fact this is silly. (I.E. Your program
> starts from scratch every time, therefore you assume that
> everybody wants to do this.)
Maybe I should have said that a program that uses a spanning
tree CAN generate its own.
and is a second message John Halleck said:
> Spanning trees have many uses. For example, if a program
> finds loops having a spanning tree availiable makes it easy.
> Most importantly from my point of view, spanning trees
> can be updated when combining two files, with less effort
> than they can be computed from scratch. Just as with
> L.S. adjustments, you can update what you have with something
> new with relatively small overhead compared with the overhead
> of doing the entire thing from scratch.
>
> I am personally a fan of programs that do incremental updates
> of data instead of massive "do it all over" approaches. You
> may or may not me, I don't know. But I think that membership
> in the spanning tree (or not) is a useful thing to know about
> a shot. While I could recompute it, if I provide it, with
> some basic programs to use it, then I think it could save
> others work.
At first I thought I did not even generate a spanning tree.
Then I realized that I generate one for the connecting of
survey shots and the calculation of unclosed coordinates.
Relative to the mechanical read operations, the parsing,
and the trig functions, the generation of the spanning tree is
very fast. I can't think of any program that puts out and
then reads in a spanning tree for a later stage of processing.
In my program here is no clean split between generating a
tree and using it. I do not have one well-defined module
that generates a tree followed by another module that uses
a tree. There is no good place to put a tree read.
I thought of using several stages of processing where
intermediate data files are written out. After getting my
first working version I concluded that the only worthwhile
functional split is between the generation of coordinates
(adjusted or unadjusted) and plotting. There is a good
chance of having to recalculate everything after correcting
a mistake in the data.
If you can read in a separate spanning tree and update it
incrementally, you are the only one.
That being said, I now realize that I write out the spanning
tree as part of my unadjusted or adjusted coordinate data
along with ordinal station numbers.
Bob Thrun
This archive was generated by hypermail 2b30 : Thu Mar 01 2001 - 18:00:01 CET