for 6.5svn I would vote to add .zonal as another module if
you like, but do not replace r.univar.
well, actually the diff is not so huge to make code auditing that
much of a problem, but it would need to be well tested.
generally I would prefer if we slow down new development in 6.5
in favour of 7 though, and we do that by not backporting major
architecture changes from trunk that do not fix major flaws in the
6.x architecture which can not be otherwise somehow worked around.
...ie just be more conservative with changes to 6.x now. otherwise
the stable branch never stabalizes and grass7 never gets worked
on...
for 6.5svn I would vote to add .zonal as another module if
you like, but do not replace r.univar.
well, actually the diff is not so huge to make code auditing that
much of a problem, but it would need to be well tested.
You helped yourself with development and testing of r.univar.zonal,
and I assume that a number of users managed to get it up and running
by installing it without g.extension. Otto Dassau developed
v.rast.stats2 using r.univar.zonal, and v.rast.stats2 is more than n
categories times faster than the current v.rast.stats, e.g. for 1000
categories that could be 10 minutes vs. 7 days. IOW, it looks like
r.univar.zonal has been used and tested quite a bit in the last nearly
2 (!) years. About k.i.s.s., it's true that r.univar.zonal is a bit
more complex than r.univar, but I guess over 90% of the code are
identical, the crucial stuff is just indented one more level for the
zoning loop. IMHO k.i.s.s. also means that code duplication should be
avoided.
On Thu, Oct 7, 2010 at 10:31 AM, Markus Metz
<markus.metz.giswork@googlemail.com> wrote:
Hamish wrote:
Hamish wrote:
for 6.5svn I would vote to add .zonal as another module if
you like, but do not replace r.univar.
well, actually the diff is not so huge to make code auditing that
much of a problem, but it would need to be well tested.
You helped yourself with development and testing of r.univar.zonal,
and I assume that a number of users managed to get it up and running
by installing it without g.extension. Otto Dassau developed
v.rast.stats2 using r.univar.zonal, and v.rast.stats2 is more than n
categories times faster than the current v.rast.stats, e.g. for 1000
categories that could be 10 minutes vs. 7 days.
For a "normal" user that even means working version versus killed job.
Please propagate the new code in from Addons rather than forcing
too many users to spend time on installing Addons while our Addon
installer is yet suboptimal.
IOW, it looks like
r.univar.zonal has been used and tested quite a bit in the last nearly
2 (!) years. About k.i.s.s., it's true that r.univar.zonal is a bit
more complex than r.univar, but I guess over 90% of the code are
identical, the crucial stuff is just indented one more level for the
zoning loop. IMHO k.i.s.s. also means that code duplication should be
avoided.
Yes, please let us avoid cloning code. Some years ago we spent months
on this, so let's not introduce it again.