Hi,
can we change the vector library to first write into
the .tmp directory (as done for raster data) and then
copy into the target directory within vector/?
Like this we also had the "transactional" file management.
Currently, if a vector command fails (crash, connection to
remote machine fails, disk full or other problems), it leaves
corrupted files which appear to be a valid vector map.
Markus
On 7/28/05, Markus Neteler <neteler@itc.it> wrote:
Hi,
can we change the vector library to first write into
the .tmp directory (as done for raster data) and then
copy into the target directory within vector/?
Like this we also had the "transactional" file management.
Currently, if a vector command fails (crash, connection to
remote machine fails, disk full or other problems), it leaves
corrupted files which appear to be a valid vector map.
Certainly not for update - Vect_open_update().
For new vectors, it would be possible if a table is not created.
If a table is created before/with geometry, we must store
somewhere the information that the table exists, so that it can
be deleted with the vector map.
If a vector map is corrupted, it is reported by library
(size of coor file).
Radim