[GRASS-dev] [GRASS-SVN] r73995 - grass/trunk/scripts/v.db.addtable

[Markus Metz]

On Sat, Jan 26, 2019 at 2:50 PM Huidae Cho <grass4u at gmail.com> wrote:
>
> Markus,
>
> If there is a linked table, v.db.addtable stops in line 106. If not, this
script doesn't create a unique index and calls v.db.connect. v.db.connect
adds a db link in line 317 and creates a unique index (db_create_index2) in
line 334 if linking was successful. SQLite didn't complain when both
modules created the same unique index, but PostgreSQL failed in
v.db.connect (2nd time creating the same unique index). Not sure which code
was added first/later. I think it's more of how we design both modules.
v.db.connect will fail if we try to link a table with a unique index. Isn't
v.db.connect supposed to "just" connect a table to a layer (without
creating any database objects like index)? Which module should be
responsible for creating unique indices?
>
> Before this commit:
> 1. v.db.addtable creates unique index
> 2. v.db.connect tries to create unique index again ==> fatal error
>
> After this commit:
> 1. v.db.addtable doesn't create unique index
> 2. v.db.connect creates unique index
>
> Probably, it should be:
> 1. v.db.addtable should create unique index
> 2. v.db.connect shouldn't try to create unique index? Just "connect"...

Yes, this is the preferred solution. Consider v.db.connect -d followed by
v.db.connect, i.e. disconnecting a table and then reconnecting the same
table: in this case v.db.connect should also fail with PostgreSQL. IMHO, a
unique index should only be created when the table is created.

Markus M

>
> Regards,
> Huidae
>
> On Sat, Jan 26, 2019 at 7:39 AM Markus Neteler <neteler at osgeo.org> wrote:
>>
>> Hi,
>>
>> On Tue, Jan 22, 2019 at 3:51 AM <svn_grass at osgeo.org> wrote:
>> >
>> > Author: hcho
>> > Date: 2019-01-21 18:51:33 -0800 (Mon, 21 Jan 2019)
>> > New Revision: 73995
>> >
>> > Modified:
>> >    grass/trunk/scripts/v.db.addtable/v.db.addtable.py
>> > Log:
>> > v.db.addtable: Do not create unique index from this script;
v.db.connect will try to create it again causing some drivers to fail
(PostgreSQL)
>> >
>> > Modified: grass/trunk/scripts/v.db.addtable/v.db.addtable.py
>> > ===================================================================
>> > --- grass/trunk/scripts/v.db.addtable/v.db.addtable.py  2019-01-21
22:37:59 UTC (rev 73994)
>> > +++ grass/trunk/scripts/v.db.addtable/v.db.addtable.py  2019-01-22
02:51:33 UTC (rev 73995)
>> > @@ -139,16 +139,6 @@
>> >          except CalledModuleError:
>> >              grass.fatal(_("Unable to create table <%s>") % table)
>> >
>> > -        # create index, see db/driver/*/index.c
>> > -        if driver != "dbf":
>> > -            sql = "CREATE UNIQUE INDEX %s_%s ON %s (%s)" % (table,
key, table, key)
>> > -            try:
>> > -                grass.run_command('db.execute',
>> > -                                  database=database, driver=driver,
sql=sql)
>> > -            except:
>> > -                grass.warning(_("Unable to create index on table
<%s>") % table)
>> > -                pass
>> > -
>> >      # connect the map to the DB:
>> >      if schema:
>> >          table = '{schema}.{table}'.format(schema=schema, table=table)
>>
>> just a conceptual question:
>> ... are we sure that this index creation removal never leads to a
>> table without index?
>>
>> Markus
>
>
>
> --
> Huidae Cho, Ph.D., GISP, PE (MD), CFM, M.ASCE
> Open Source GIS Developer, GRASS GIS Development Team
> _______________________________________________
> grass-dev mailing list
> grass-dev at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/grass-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-dev/attachments/20190127/9db04254/attachment-0001.html>