Paul Ramsey wrote:
On Jan 23, 2008, at 3:39 AM, Peter Hopfgartner wrote:
we were looking at the tables generated by shp2pgsql. As it seems,
numeric(m, 0) is mapped, if possible to integers, and numeric(m,n) is
mapped to numeric.
Do you mean, mapped to float?
Usually yes, but in our case the original column was a a numeric(19,
11), so the if statement in line 559 choses numeric instead of float8,
which makes perfectly sense.
We would like to optionally maintain the numeric length and precision
for our data imports. Looking at shp2pgsql.c, this should be rather
simple. Basically, we could add an option, let's say '-K' for keep
precision, and modify the code around line 550.
This is possible. You need to be aware of the *massive* performance
penalty that comes with using numeric instead of integer/float. As
long as you are clear on that, it's certainly acceptable.
Yes, this is true. In almost all cases the current behavior of shp2pgsql
and pgsql2shp is preferable. But in this specific project we are working
on, the customer, a public administration, insists on data formats.
Since we have -i for "force to integer" how about -n for "use numeric
for all numbers"?
Sounds good.
A similar change could be done for pgsql2shp, modifying the SQL
query, which gets the type information from PostgreSQL and changing
the code generating the dbf header.
If such a change is welcomed, we can try to have it done soon.
Regards,
Peter Hopfgartner and Sergio Segala
R3 GIS
http://www.r3-gis.com
_______________________________________________
postgis-users mailing list
[email protected]
http://postgis.refractions.net/mailman/listinfo/postgis-users
_______________________________________________
postgis-users mailing list
[email protected]
http://postgis.refractions.net/mailman/listinfo/postgis-users
Regards,
Peter & Sergio
R3 GIS
http://www.r3-gis.com
_______________________________________________
postgis-users mailing list
[email protected]
http://postgis.refractions.net/mailman/listinfo/postgis-users
