On 10/12/11 00:57, Marcello Gorini wrote:
Glynn:
Note that using the same map as both input and output results in
undefined behaviour. Even if it happens to work, there's no guarantee
that it will continue to work in future versions.
If you need to implement an iterative algorithm, you should generate a
new map, then replace the original with g.remove+g.rename afterwards.
Now, you scared me, because my ongoing work is pretty much based on this
kind of code. I use r.mapcalc to create maps that serve as counters such
as the command i++ in C. I initialize them as zero maps and then based
on some logic I keep incrementing them, pretty much like the following
pseudo-code to implement a sum:
sum = 0
for i = 1 to 10
sum = sum + i
end
But I do more that that.
For instance, if certain conditions are met in other maps at a given
iteration and the initialized map is zero, then update the initialized
map with the value of "i". If not, don't change it. In this way, my map
is updated every iteration until no more cells meet the condition or
some threshold is met and processing stops. I even use d.rast to display
the map as it is being updated.
How should I proceed to avoid the "undefined behaviour" and still get
what I want? (I hope I was able to make myself understood :) )
As Glynn suggests:
for i = 1 to 10
sum_new = sum + i
g.remove rast=sum
g.rename rast=sum_new,sum
or
for i = 1 to 10
sum_new = sum + i
g.rename rast=sum_new,sum --overwrite
Glynn, is there any reason not to do use g.rename with --overwrite ?
Moritz
_______________________________________________
grass-user mailing list
[email protected]
http://lists.osgeo.org/mailman/listinfo/grass-user
