[GRASS-stats] Why does rgrass7::writeVECT() silently add a cat column if it doesn't exist?

[Roger Bivand]

On Thu, 21 Oct 2021, Tobias Pilz wrote:

> Hi Vero,
>
> as I understand it, vector files in GRASS need a key column if they have an 
> attribute table. But I don't see that this has to be a column named 'cat'. 
> However, GRASS is quite inconsistent here. Some functions ensure the 
> existence of a 'cat' column, such as v.db.dropcolumn (just gives an error if 
> you try to remove the 'cat' column), whereas other functions let one define 
> other columns as key column (e.g. manually in the GUI or v.db.addtable).
>
> The latter happened with my vector map, i.e. a different column was defined 
> as key column. Therefore it is confusing why rgrass7::writeVECT() just adds 
> the cat column and without any notification. This makes no sense to me and is 
> annoying as my subsequent workflow produced unexpected results and I had to 
> trace back the error which took a while.

In sf vector objects, there is no concept of key column, I think. In sp 
vector objects (used before, should be avoided from now on as sf is 
actively maintained), the geometry ID is keyed to the row.names of the 
data slot data.frame object, but is also not seen as a user-facing key 
column.

So you are adding (as a user) a concept not present  in the R-side 
objects.

In practice, cat can be useful where cleaning has split line or 
polygon objects imported into GRASS into multiple features, which can then 
be unioned back after exporting from GRASS to R, but this isn't a concept, 
it is a pragmatic fix.

Hope this clarifies,

Roger


>
> From my perspective as user it would be great if a more consistent solution 
> regarding the treatment of key columns could be found, i.e. either always 
> ensure the existence of a column named 'cat' or just accept if the user 
> specifies a different key column.
>
> Regards,
>
> Tobias
>
> Am 20.10.21 um 20:41 schrieb Veronica Andreo:
>>  Hello Tobias,
>>
>>  The cat colum is the ID that GRASS uses for vectors, hence this is
>>  expected
>>  and right. The same happens if you import a gpgk or shape or any other
>>  vector file into GRASS. See
>>  https://grass.osgeo.org/grass78/manuals/vectorintro.html for further info
>>
>>  HTH,
>>  Vero
>> 
>> 
>>
>>  El mié, 20 oct 2021 a las 9:34, Tobias Pilz (<topilz at pik-potsdam.de>)
>>  escribió:
>>
>>>  Dear list,
>>>
>>>  when I use in R the writeVECT() command the resulting map in GRASS got
>>>  the 'cat' column silently added to its attribute table if it didn't
>>>  exist.
>>>
>>>  The vector object in question was previously imported to R via
>>>  readVECT() and had a different column defined as key column.
>>>
>>>  Is this behaviour of writeVECT() intentional? If yes, why so and
>>>  wouldn't it make sense to give a warning?
>>>
>>>  Note: I used rgrass7 version 0.2-6 together with GRASS 7.8.5 under Linux
>>>  (opensuse).
>>>
>>>  Regards,
>>>
>>>  Tobias
>>> 
>>>
>>>  _______________________________________________
>>>  grass-stats mailing list
>>>  grass-stats at lists.osgeo.org
>>>  https://lists.osgeo.org/mailman/listinfo/grass-stats
>>> 
>

-- 
Roger Bivand
Emeritus Professor
Department of Economics, Norwegian School of Economics,
Postboks 3490 Ytre Sandviken, 5045 Bergen, Norway.
e-mail: Roger.Bivand at nhh.no
https://orcid.org/0000-0003-2392-6140
https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en