[postgis-devel] Loader cleanup

[Markus Schaber]

Hi, Strk,

strk at refractions.net schrieb:

> 	- GID omission.
> 
> 	  Current loader explicitly set the gid when NOT
> 	  in append mode. I'd make it *always* act like that
> 	  and rely on the 'serial' type. Do you see any
> 	  problem with that ?

I think this is a good change as it simplifies code.

> 	- TRANSACTION
> 
> 	  Current loader breaks insert lines in multiple
> 	  transaction blocks. It has been noted that this
> 	  should speed things up, but I'm afraid it would
> 	  add half-shapes (skipping the bogus transaction
> 	  block and continue with the rest). Would it be
> 	  ok for you wrapping the whole code in a single
> 	  transaction relying on dump mode (-D) for speed
> 	  reasons ?

I think this is a good idea, too.

> 	- NULL shapes
> 
> 	  Current loader checks shapefile and exits with
> 	  an error if all shapes are null (or contain no
> 	  vertexes). This requires two shapefile scans
> 	  (check scan, load scan).

When multiple transaction blocks are removed, we can do the checking on
the fly. When we detect a NULL value, we can do
  if (dump_format) printf("\\.\n");
  printf("ABORT;\n")
and spit out an error message on stderr.

>         Since the load scan
> 	  copes cleanly with null shapes what about dropping
> 	  the first scan allowing for a all-nulls shapefile
> 	  to produce an all-nulls geometry features table ?
> 	  (attributes would still be loaded!)

You could invent a command line switch whether NULL geometries are to be
allowed. But this could also be done on the SQL level using a NOT NULL
constraint, this is what we currently do. Keeping consistency checks
inside the database is always a good thing. :-)

Markus
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 256 bytes
Desc: OpenPGP digital signature
URL: <http://lists.osgeo.org/pipermail/postgis-devel/attachments/20050406/2ce59357/attachment.sig>