[GRASS5] What to release as 5.1.0?

[Markus Neteler]

On Tue, Feb 25, 2003 at 01:07:06PM +0100, Bernhard Reiter wrote:
> On Tue, Feb 25, 2003 at 12:39:34PM +0100, Radim Blazek wrote:
> > On Monday 24 February 2003 05:09 pm, Bernhard wrote:
> > > However we try to get 5.1.x out and stable as fast
> > > as possible which you probably should count in quarter of years.
> > 
> > The question is what may be released as 5.1.0, I think that it could 
> > be released when it contains most of vector+sites functionality
> > from 5.0. Say 90% of modules?
> 
> Sounds fine to me, though what about raster?

Let's say 90% of modules. This could be ready within some days
of generating the relevant Makefiles and the entries in the
50-to-51-links tool. Only hybrid modules with raster *and* vector
support such as r.flow need an update (which may be rather easy to
do).
 
[...]
> > Unfortunately the system how categories are stored and then attached
> > to lines and points IS bad in principle. It is not reliable and correct 
> > data imported to 5.0 may lose it's attributes. That was why 5.1 was started.
> 
> When do we loose vector attributes?
When you import vector maps. GRASS 5.0 supports only a single attribute.
Real world vector maps often come with a database table of attributes
(such as SHAPE/DBF, E00, MapInfo etc).
It is a real pain to work with such data in 5.0.

> If the vector support as limited as it is in 5.0. has serious flaws,
> then we could consider to limit ourself to only fix this flaw for 5.2
> and delay other things for 5.4.

That means to reinvent the wheel? Bernhard, the idea of 5.1 is to
overcome the serious flaws of 5.0 :-) BTW, it's functional...
I would not waste a minute to change the 5.0/5.2 design in terms of
the vector engine.
 
[...]
> > Yes, and it will remain as it is if everybody says "it is not stable, I'll
> > work on stable". If you think that some hidden team of developers exists,
> > which will suddenly release stable 5.2, then that's not true.
> 
> I agree with Radim that we need to people to progress 5.1.
> Which does not mean that we should stop supporting 5.0.x.

Right. But new vector development should be done in 5.1. Fixes may
go into 5.0 as well. Note that the 5.1 vector engine is much (!)
cleaner than that of 5.0.

[...]
> > I think that everybody who wants to use v.digit 
> > more than one day, should consider
> > v.digit/5.1
> > v.out.ascii/5.1
> > v.in.ascii/5.0

This is a good suggestion for the time of transition.

Another example: I have imported the world vector map of UNEP
http://www.grid.unep.ch/data/grid/downloads/boundaries.php
in E00 format (gnv19.tar.Z). With m.in.e00 (5.0) you get doubled
centroids and tons of warnings in v.support. 
So I did:

gnv19.tar.Z -> m.in.e00 (5.0) -> v.convert (5.1) -> v.clean (5.1)
-> v.out.ascii (5.1) -> v.in.ascii (5.0)

Note that v.clean (5.0) will fail here...

When OGR will support ASCII E00 one day (already a request in bugzilla
of OGR), it will be much easier: 

v.in.ogr (5.1) -> v.clean (5.1) -> v.out.ascii (5.1) -> v.in.ascii (5.0)

The v.in.ogr is already there and supports many formats.

To make GRASS a faster dinosaur, we have to focus on 5.1. Otherwise
it were dead (say, without users) one day.

Markus

PS: There is a 5.1 tutorial: 
    http://grass.itc.it/grass51/tutorial/
    and the documented API
    http://grass.itc.it/grass51/manuals/