[GRASS-user] meaning of "geometry"

[Markus Metz]

Dear Dave,

On Tue, Oct 25, 2016 at 9:25 PM, Dave Roberts <droberts at montana.edu> wrote:
> Dear Markus and Anna,
>
> I'd have to dig a little for man pages, but one interesting example is a
> series of GRASS-user postings on "grass vector model, cats and layers
> concept" from 6/13 which you were central to Markus.  At the time, the
> question was "why does adding a new layer to a map require creating a new
> map?"  And your answer (over two postings) was "Because you need to modify
> vector geometries in order to add a new layer. Categories and layers are
> first and foremost stored together with the geometries... [Adding a new
> layer] ... changes the geometry directly."
>
> Nikos Alexandris (and I) were baffled by this because it doesn't change any
> vertices, nodes, or centroids.
>
> Now, having studied the structure of the coor file I see that adding a new
> layer to a vector object increases the length of that record in the coor
> file, so that all records downstream of that record also have to be
> re-written.  It's simpler and safer to write a new vector object than to
> modify the original.  It would be possible to simulate the expected behavior
> by writing the new map to "tmp", deleting the original map, and renaming the
> tmp map to the original name, but obviously the programmers elected not to
> do that.  It could, however, be easily implemented in a shell or python
> script.

Adding a new layer with e.g. v.category does change the geometries
because the combination of (category, layer) is stored in the coor
file  together with the geometry can be of type point, centroid, line,
boundary.

>
> So now when I see "geometry" in a manual page or GRASS-user posting I
> interpret that to mean "vertices, nodes, centroids, and associated
> categories ordered by the (not necessarily consecutive) layer numbers" as
> that's the essential nature of a record in the coor file that manages all
> this.

"Geometry" refers first of all to points, lines, and areas. Centroids
and boundaries are special cases. Nodes are not a geometry feature,
but an internally used structure that connects lines and boundaries.
>From a user perspective, nodes are important for vector network
analysis, otherwise nodes are used internally to construct areas.

>
> Obviously that's more operational than conceptual so I was wondering where
> the concept of geometry is actually defined.

Can you provide an example of a GRASS manual where the usage of the
term "geometry" is not clear?

Markus M

>
> Thanks, Dave
>
>
> On 10/24/16 14:39, Markus Metz wrote:
>>
>> On Mon, Oct 24, 2016 at 4:07 PM, Anna Petrášová <kratochanna at gmail.com>
>> wrote:
>>>
>>> On Mon, Oct 24, 2016 at 9:04 AM, Dave Roberts <droberts at montana.edu>
>>> wrote:
>>>>
>>>> Frequently in GRASS help files or emails in this list the term
>>>> "geometry" is
>>>> used.  Is geometry synonymous with a record in the coor file, or does it
>>>> have a broader meaning?
>>>
>>>
>>> I would say geometry of features, such as points, lines, boundaries,
>>> areas.
>>
>>
>> and centroids. In most cases, the vector features of interest are only
>> points, lines, and areas.
>>
>>> Depends on the particular context I guess, could be related to
>>> topology too.
>>
>>
>> An example of the context, i.e. a particular manual, would be helpful
>> for clarification.
>>
>> Markus M
>>
>>> I am actually not sure what is in the coor file, but you
>>> are not supposed to access it directly anyway.
>>>
>>> Anna
>>>
>>>>
>>>> Thanks, Dave
>>>> --
>
>
>
> --