[GRASS-dev] r.external no cats file?

[Markus Metz]

On Wed, Jun 29, 2011 at 11:21 AM, Markus Metz
<markus.metz.giswork at googlemail.com> wrote:
> On Wed, Jun 29, 2011 at 12:23 AM, Glynn Clements
> <glynn at gclements.plus.com> wrote:
>>
>> Markus Metz wrote:
>>
>>> When linking external raster maps with r.external, I have to manually
>>> create a corresponding cats file, otherwise a number of modules, e.g.
>>> d.rast do not work properly. Is there any reason why no corresponding
>>> cats file is produced?
>>
>> IMHO, modules should not be required to explicitly write a cats file.
>> GRASS 4.x is ancient history, and raster maps are no longer inherently
>> category-based.
>>
> I think this is true for all modules creating regular raster maps
> because a cats file is automatically generated by the library when a
> new raster map is closed. r.external does not use Rast_close() and
> creates all the support files itself, but not the cats file.
>
>> Raster modules should only write a cats file if they actually have
>> something meaningful to put in it.
>>
>> Either Rast_read_cats() should silently return an empty category
>> structure if no cats file is present, or Rast_close() should
>> automatically generate an empty cats file if no cats file was
>> explicitly written.
>>
> This is what Rast_close() does, but r.external does not call
> Rast_close(). Maybe it should?
>
r.external can not use Rast_close() because it never opens a raster
map with Rast_open_new(), thus the module has to do itself what
Rast_close() does for all other modules that create a new raster map?

Markus M
>
>> I'm inclined towards the former, unless there might be cases where the
>> distinction between "no category list" and "empty category list" would
>> matter.
>>
>> --
>> Glynn Clements <glynn at gclements.plus.com>
>>
>