Rainer M Krug wrote: > > Monitoring the filesystem is likely to be unreliable, as there's no > > straightforward way to determine when files are in a consistent state. > > Thanks for pointing this out and saving me some work - otherwise I might > have chosen that path...
Also, doing this at the library level would have similar issues, albeit less severe. E.g. when a module finishes writing the actual raster data (i.e. G_close_cell() or Rast_close()), often the next thing it does is to write metadata (categories, colour table, history). There is no specific "event" which signals that the output raster is "complete". Generating notifications on "close" operations would ensure that individual files are consistent, but that doesn't mean that the raster as a whole is consistent. This is something that has been discussed in the past, but without resolution. Improvements in this area are certainly desirable; e.g. reading a map in another mapset relies upon luck, as there is no mechanism to ensure that the owner of that mapset isn't modifying the map in question at the time. Unfortunately, the issues are complex, making it hard to decide what should be done, let alone doing it. -- Glynn Clements <[email protected]> _______________________________________________ grass-user mailing list [email protected] http://lists.osgeo.org/mailman/listinfo/grass-user
