[GRASS-dev] from v6 to v7: G_get_cellhd -> Rast_get_cellhd ?

[Glynn Clements]

Yann Chemin wrote:

> some confusion you may clarify.
> 
> Looking into some v6 to v7 coding translation, how would you test for
> readability of raster cell header in v7
> 
> <from grass-addons: i.landsat.toar>
> 
> 	    mapset = G_find_cell2(band_in, "");
> 	    if (mapset == NULL) {
> 		G_warning(_("Raster map <%s> not found"), band_in);
> 		continue;
> 	    }
> 	    if ((infd = G_open_cell_old(band_in, mapset)) < 0)
> 		G_fatal_error(_("Unable to open raster map <%s>"), band_in);
> 	    if (G_get_cellhd(band_in, mapset, &cellhd) < 0)
> 		G_fatal_error(_("Unable to read header of raster map <%s>"), band_in);
> 	    if (G_set_window(&cellhd) < 0)
> 		G_fatal_error(_("Cannot reset current region"));
> 
> The Last 4 lines I am trying to see if Rast_get_cellhd should be used...
> 
>             if (Rast_get_cellhd(band_in, "", &cellhd) < 0)
>                 G_fatal_error(_("Unable to read header of raster map <%s>"), band_in);
>             if (G_set_window(&cellhd) < 0)
>                 G_fatal_error(_("Cannot reset current region"));
> 
> These 4 lines above are returning errors because the two functions
> Rast_get_cellhd and G_set_window both return void, and they are tested
> for a return value.
> 
> How would you rewrite that initial code within version 7?

Remove the test. If a 6.x function returned a status and the 7.x
version doesn't return anything, it means that the 7.x version calls
G_fatal_error() itself.

I made this change for almost every lib/raster function where the only
consequence of an error return was to call G_fatal_error() (assuming
that the code bothered to check the return value in the first place).

Also, note that Rast_open_* will only ever return a valid descriptor. 
If something goes wrong, it results in a fatal error, so there's no
point in explicitly checking for "fd < 0".

Functions which read metadata files which may or may not exist
(categories, colour tables, etc) still return status codes.

-- 
Glynn Clements <glynn at gclements.plus.com>