d.rast: multiple calling G_find_raster2()

[martinl]

When rendering raster data using d.rast the function G_find_raster2() is called multiple times, this could be probably reduced... ???

D1/1: Cairo_Set_window: 0.000000 480.000000 0.000000 640.000000
D1/1: G_find_raster2(): name=dmt mapset=
D1/1: G_find_raster2(): name=dmt mapset=
D1/1: G_find_raster(): name=dmt mapset=
D1/1: G_find_raster(): name=MASK mapset=pg
D1/1: Cairo_Set_window: 0.000000 480.000000 0.000000 640.000000
D1/1: Cairo_Set_window: 0.000000 480.000000 80.000000 560.000000
D1/1: G_find_raster2(): name=dmt mapset=
D1/1: G_find_raster2(): name=dmt mapset=
D1/1: G_find_raster2(): name=dmt mapset=PERMANENT
D1/1: G_find_raster2(): name=dmt mapset=
D1/1: G_find_raster2(): name=dmt mapset=
D1/1: G_find_raster2(): name=dmt mapset=PERMANENT
D1/1: G_find_raster2(): name=dmt mapset=
D1/1: G_find_raster2(): name=dmt mapset=PERMANENT
D1/1: G_find_raster(): name=MASK mapset=pg

[glynn]

Replying to martinl:

When rendering raster data using d.rast the function G_find_raster2() is called multiple times, this could be probably reduced... ???

Maybe. Although most of this isn't specific to d.rast, but applies to any module which accesses raster maps. Most lib/raster functions which take a map name as an argument end up calling G_find_raster2().

Taking d.rast as an example, setting a breakpoint on G_find_raster2() shows:

#0  G_find_raster2 () at find_rast.c:78
#1 Rast_map_is_fp () at open.c:742
#2 main () at main.c:100

#0  G_find_raster2 () at find_rast.c:78
#1 Rast_map_is_fp () at open.c:742
#2 Rast_read_colors () at color_read.c:67
#3 display () at display.c:22
#4 main () at main.c:109

#0  G_find_raster2 () at find_rast.c:78
#1 Rast__open_old () at open.c:174
#2 Rast_open_old () at open.c:113
#3 cell_draw () at display.c:59
#4 display () at display.c:34
#5 main () at main.c:109

#0  G_find_raster2 () at find_rast.c:78
#1 Rast_map_type () at open.c:775
#2 Rast__open_old () at open.c:204
#3 Rast_open_old () at open.c:113
#4 cell_draw () at display.c:59
#5 display () at display.c:34
#6 main () at main.c:109

#0  G_find_raster2 () at find_rast.c:78
#1 Rast_get_gdal_link () at gdal.c:251
#2 Rast__open_old () at open.c:251
#3 Rast_open_old () at open.c:113
#4 cell_draw () at display.c:59
#5 display () at display.c:34
#6 main () at main.c:109

#0  G_find_raster2 () at find_rast.c:78
#1 Rast_map_type () at open.c:775
#2 Rast_get_gdal_link () at gdal.c:254
#3 Rast__open_old () at open.c:251
#4 Rast_open_old () at open.c:113
#5 cell_draw () at display.c:59
#6 display () at display.c:34
#7 main () at main.c:109

Six calls, of which one is from Rast_map_is_fp(), one from Rast_read_colors() (via Rast_map_is_fp()), and the other four from Rast_open_old().

Some of those could be eliminated. E.g. Rast_get_gdal_link() could probably skip the explicit G_find_raster2() call and move the Rast_map_type() call to after the G_fopen_old_misc() call (i.e. the check for whether the map is actually GDAL-linked).

A more general solution would be to replace the use of (name, mapset) with a map handle ("fd"). This would require opening maps earlier, but would allow information to be retrieved from the fileinfo structure. E.g. calls to Rast_map_type() could be replaced with Rast_get_map_type(), which is just:

RASTER_MAP_TYPE Rast_get_map_type(int fd)
{
    struct fileinfo *fcb = &R__.fileinfo[fd];

    return fcb->map_type;
}

[neteler]

Milestone renamed

[neteler]

Ticket retargeted after milestone closed

[neteler]

Ticket retargeted after milestone closed

[martinl]

Ticket retargeted after milestone closed