[postgis-devel] shp2pgsql patch

strk at refractions.net strk at refractions.net
Mon Apr 4 07:33:56 PDT 2005


On Mon, Apr 04, 2005 at 04:27:40PM +0200, Markus Schaber wrote:
> Hi, Strk,
> 
> strk at refractions.net schrieb:
> > Oh.. I've found the transaction issue.
> > Since I remember the shp2pgsql code splitted the load into 
> > multiple transactions. This actually makes it hard
> > to use the transaction as an integrity check as only the
> > block of records in the transaction containing a failur will
> > be skipped while the END; BEGIN; statements will start a fresh
> > one.
> >
> > Maybe Jeff knows why it is like that, or Paul (but he's out of
> > office for a couple of weeks).
> 
> Hmm, the splitting is some optimization thing if you happen to have
> concurrently running processes on the same table, I guess. Maybe we
> should benchmark this and/or discuss it on the performance list.
> 
> It did not hit me yet as I always use the -D(ump) format. I usually have
> large bunches of data (hundred thousands up to millions of rows), and -D
> turned out to be 5 - 7 times faster.
> 
> >>I'd include your patch as it is, w/out requiring RC7.
> >>Does it sound azardous ?
> 
> Hmm - no risk, no fun? The patch at least passed my tests, and
> production upload of about 40 shapefiles (35 million rows) here. AFAIK,
> there are no regression tests yet, right?
> 
> But then I would also vote to include Alex' manpages as well as the
> debian/ subdirectory, as they are only file additions that both should
> not be able to hurt anything else.
> 
> Markus

Man pages are in place. Added install rule today.
I did not get the debian directory package.
For shp2pgsql patch probably better delay so we handle the
transaction's cleanup as well (btw: I think I'll change the load_data
function: it takes too many args ;)
--strk;




> _______________________________________________
> postgis-devel mailing list
> postgis-devel at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-devel




More information about the postgis-devel mailing list