Hi Markus, hi everybody,
last night I tried to track down v.trim bug that has been marked as
release critical... well, my opionion in short is that v.trim has
too many bugs and to make it useful it needs a major rewrite, not a
bug fix. Well, now a more in depth description to support my opinion.
v.trim is expected to remove small spurs and node in excess from a
vector
layer based on a trimming factor that is the minimum distance between
nodes
(beware, I said nodes, not vertices, I mean line endings). Here the list
of bugs I've seen in that small program (277 lines of code):
* trim factor is just a distance (a double), not a percentage, so why
limit
it between 0 and 100?
* a line is eliminated if the euclidean distance between its nodes
(line endings) is less than the trimming factor... no surprise that
closed loops gets eliminated, start node = end node, so euclidean
distance is 0 regardless of how big is the loop;
* two consequent lines are joined into one regardless of their
attributes
-> this is loss of information
* in fact attributes are not treated at all, the output file loses all
the label points
There are also other problems, but it seems to me that the ones cited
are enough (for me) to vote for v.trim removal from grass 5 stable:
fixing it is almost equivalent to building it from scratch,
it seems much more a new feature than a bug fix :-). What is your
opinion?
BTW, we still have to take a decision about zero handling in vector
files...
Regard
Andrea Aime
----------------------------------------
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo@geog.uni-hannover.de with
subject 'unsubscribe grass5'
On Wed, Mar 28, 2001 at 08:13:43AM +0200, Andrea Aime wrote:
Hi Markus, hi everybody,
last night I tried to track down v.trim bug that has been marked as
release critical... well, my opionion in short is that v.trim has
too many bugs and to make it useful it needs a major rewrite, not a
bug fix. Well, now a more in depth description to support my opinion.
v.trim is expected to remove small spurs and node in excess from a
vector
layer based on a trimming factor that is the minimum distance between
nodes
(beware, I said nodes, not vertices, I mean line endings). Here the list
of bugs I've seen in that small program (277 lines of code):
* trim factor is just a distance (a double), not a percentage, so why
limit it between 0 and 100?
* a line is eliminated if the euclidean distance between its nodes
(line endings) is less than the trimming factor... no surprise that
closed loops gets eliminated, start node = end node, so euclidean
distance is 0 regardless of how big is the loop;
* two consequent lines are joined into one regardless of their
attributes
-> this is loss of information
* in fact attributes are not treated at all, the output file loses all
the label points
There are also other problems, but it seems to me that the ones cited
are enough (for me) to vote for v.trim removal from grass 5 stable:
fixing it is almost equivalent to building it from scratch,
it seems much more a new feature than a bug fix :-). What is your
opinion?
Yes, I have comment it in the compile list.
BTW, we still have to take a decision about zero handling in vector
files...
If you have a preferred way for r.poly, please implement it. We have
to speed up for 5.0 now, otherwise noone will believe in it any more.
As vector section completely changes, we simply need a "rather" working
r.poly.
Thanks for analysing v.trim!
Markus
----------------------------------------
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo@geog.uni-hannover.de with
subject 'unsubscribe grass5'