Talk:Live GIS Disc Testing

From OSGeo
Jump to navigation Jump to search
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.
darkblue_B: hi MartinSpott
darkblue_B: we were working on the OSGeo liveDVD last night
darkblue_B: I moved over a PostGIS database
darkblue_B: one of the others asked if I could write up a note on "How To"
MartinSpot: Hi darkblue_B, "ho to move a PosiGIS DB" ?
darkblue_B: its not that much.. but, since the database was created from Templte, spatial_ref_sys and geometry_colums already existed
darkblue_B: yes
darkblue_B: there were about hmm 15 tables?
darkblue_B: in Postgres its work to do anyhting to them.. like change the owner..
darkblue_B: I'm sure you have ways you have developed to do such things
darkblue_B: your database is so well maintained!
MartinSpot: Really ?
darkblue_B: you and the others I should say
darkblue_B: ;-)
darkblue_B: only you know the bad parts
darkblue_B: to me, it looks great
MartinSpot: Well, I'm casting everything into scripts, thus I have a pretty good overview about what's actually done on the DB.
darkblue_B: right
MartinSpot: Actually everything started by having one table (plus two little helpers) that contain everything I need to build the DB structure from
MartinSpot: ....  plus a couple of cascaded shell scripts to generate the respective SQL maintenance scripts.
MartinSpot: Now, is there anything missing for your HowTo ?
darkblue_B: Its not written@
darkblue_B: !!
darkblue_B: I have 'real work' too 
MartinSpot: Ah, sure
darkblue_B: well we made a template_postgis
darkblue_B: with spatial_ref_sys populated
darkblue_B: then for the 15 tables, and 4 views.. I listed them in psql
darkblue_B: removed the seqeunce names
darkblue_B: used regular expressions to generate a list of the tables (and views)
darkblue_B: created a change owner script
darkblue_B: created a pg_dump -t name1 -t name2... db --inserts line
darkblue_B: including geometry_columns but not including spatial ref sys
MartinSpot: Maybe adding an index or two is usually a good idea ....
darkblue_B: then edited out the create table statment by hand
darkblue_B: and saved it
MartinSpot: Ah good
darkblue_B: hmm indexes
darkblue_B: didnt think of that.. yes
darkblue_B: fortunately all the commands worked for the views as if they were tables
MartinSpot: I've so far been pretty happy with creating a primary key on the 'ogc_fid' column (or however you name it)
darkblue_B: I have done a similar thing for 100+ tables in a database once..
darkblue_B: I didnt do anything special with the key
MartinSpot: plus a GIST index on the geometry column .....  clustered onto the table
darkblue_B: yes, indexes.. I didnt do that
darkblue_B: this is for the FOSS4G live DVD
MartinSpot: I'm not entirely certain which one it was, but one of the tools I use requires a primary key on every table
darkblue_B: quite a few of the project will read it as a data source
darkblue_B: well, PgAdmin editing, for one
darkblue_B: in the window
MartinSpot: Either 'ogr2ogr' or 'mapserver' or 'pgsql2shp' or ....  I don't know by now
darkblue_B: ah ok
MartinSpot: Maybe as a sanity check put a constraint or two onto every table (enforce_geotype_wkb_geometry, enforce_srid_wkb_geometry)
darkblue_B: yes.. mleslie supplied the original
darkblue_B: so there were constraints and such already
MartinSpot: Fine
darkblue_B: perhaps I will just add this chat to the wiki and call it done
MartinSpot: ....  and don't forget to maintain the geometry_columns properly. 4 of 5 strange faults I usually see are due to typos or mismatches in the geometry_columns  :-)
MartinSpot: (typically my fault ....)
darkblue_B: yes - geometry_columns seems to be in good order.. all in the same projection
darkblue_B: EPSG:2207 if I recall
darkblue_B: in our US State of Oregon
darkblue_B: 2270