PostGIS requirement, moved on branch feature_Multipoint

[Piero Campalani]

Now that PostGIS is a requirement for initialization of Petascope, it should be added to the ​requirements in the ([RasdamanQuickInstall quick]) [Install installation].

A draft wiki page is already set up: MultiPointCoverages.
NOTE: it should be also highlighted that PostGIS 2.0 is required. See compatibility issues with previous PostGIS versions here: ​http://postgis.org/documentation/manual-2.1SVN/PostGIS_FAQ.html#legacy_faq_gist.

Moreover, PostGIS should be installed to petascopedb in update_petascopedb.sh.in.

[Dimitar Misev]

I strongly suggest we try to make PostGIS an optional requirement, perhaps via a configure option --enable-multi or similar. Point clouds, etc. is exotic stuff and I doubt it will be used so much, at least right now.

And this should be handled by Alireza?

[Dimitar Misev]

I.e: I should not be required to install PostGIS if I want to initialize and use petascope.

[Piero Campalani]

Yes Dimitar, that would be clearly better.

[Dimitar Misev]

I think the task can be broken in these steps:

• remove postgis from your system, drop petascopedb and RASBASE
• try to install rasdaman and petascope: if it stops somewhere, then fix with —disable-postgis option in configure.ac
• after installed, initialize with create_db.sh and update_petascopedb.sh
• start rasdaman and petascope: if petascope fails to start, fix the code to e.g. read some config param from petascope.properties
• then run systemtest, and fix again if it fails because postgis is missing

[Dimitar Misev]

On many systems it's hard to impossible to install PostGIS 2.0, e.g. on kahlua we failed installing it and finally decided to use a VM.

[Piero Campalani]

Indeed, me neither could install PostGIS 2.0 without being tangled in a long list og dependencies.
If there is a back-compatible way to create that INDEX in schema9.sql, hence without using geom gist_geometry_ops_2d, that would be better.

[Peter Baumann]

if I understand it correctly we can either try to live without PostGIS, or make it configurable. For the latter case, I suggest —with-postgis as option (ie, disabled by default) and serious runtime checks to return an appropriate error code if Discrete Coverage functionality (currently: point clouds) are invoked, but not configured.

[Dimitar Misev]

I think the deadline was to release beta1 tomorrow, but implementing a --with-postgis option will cause a delay.

I created a new branch feature_Multipoint; my suggestion is that we remove multipoint code from master while Alireza can continue to implement the --with-postgis in the branch and later merge it with master for beta2 or rc1, etc.

Does this sound ok for everyone?

[abeccati]

Sounds good to me, we also have to avoid "holes" in dbupdates for petascope when postGIS was not enabled but it gets enabled later on so definitely needs more time to get engineered and keeping it into a feature branch seems quicker option to both allow release and also early access to this experimental feature.
If we want to discuss details I think its better to move to the dev list to keep the ticket tidy.

[Piero Campalani]

update_petascopedb.sh enhanced in changeset:6aa5441.

[Piero Campalani]

CRS:1 crs was set by default to a MultipointCoverage.
Now configured CRS is set (changeset:2d44821).