Table Of Contents

Previous topic

Customizing Web Services

Next topic

Upgrading

This Page

Using spatial databases with MapFish

Since release 2.0 MapFish uses the SQLAlchemy extension GeoAlchemy that provides support for geospatial databases. By using GeoAlchemy, MapFish can also be used with all database systems supported by GeoAlchemy.

The following document is going to describe how to set up the databases and how to use them.

Using PostgreSQL/PostGIS

Installation

If you followed the installation guide, then you already have installed PostgreSQL/PostGIS and you can skip this section. If not, run the following command from a terminal to install PostgreSQL, PostGIS and the required library GEOS:

$ sudo apt-get install libgeos-3.0.0 postgresql postgis postgresql-8.3-postgis

Further information about installing PostGIS can be found in the PostGIS documentation.

Setting up a spatially-enabled database

Creating a spatially-enabled database is slightly different from creating an ordinary database. Run the following commands to create the database gis:

sudo su postgres
createdb -E UNICODE gis
createlang plpgsql gis
psql -d gis -f /usr/share/postgresql-8.3-postgis/lwpostgis.sql
psql -d gis -f /usr/share/postgresql-8.3-postgis/spatial_ref_sys.sql
psql -d gis -c "SELECT postgis_full_version()"
exit

The last query will inform you about the version of your PostGIS installation.

In most cases you do not want to access your database from a web application as root. The following commands create a user www-data and grant access rights to this user for the database gis:

sudo su postgres
createuser -P www-data
psql gis
grant all on database gis to "www-data";
grant select on spatial_ref_sys to "www-data";
grant all on geometry_columns to "www-data";
\q
exit

Hint

Depending on if your user should be allowed to create tables, to insert/update rows or just to read, you may want to adjust the privileges (see the PostgreSQL documentation for further information).

Now it is time to create tables in your database. You can either let SQLAlchemy/GeoAlchemy do that for you (see Using ‘paster setup-app’ to create your database tables) or you can manually create tables (see Creating a Spatial Table in the PostGIS documentation).

You can also use the tool shp2pgsql to create a table from a Shapefile. For example to create the table for this countries Shapefile (countries.zip), you would have to do this:

unzip countries.zip
sudo su postgres
shp2pgsql -W utf8 -s 4326 countries.shp  countries | psql -d gis
exit

And to check that everything is ok, we can query the row count. Additionally we need to grant the access to this table to user www-data in order to allow access from MapFish server in a web environment:

sudo su postgres
psql gis
\d
select count(1) from countries;
GRANT ALL ON countries TO "www-data";
\q
exit

Configuration

Once the database is set up, you only have to change the database connection string in the configuration file of your MapFish application. Open your configuration file, for example development.ini, and replace the line:

sqlalchemy.url = sqlite:///%(here)s/development.db

by this one:

sqlalchemy.url = postgresql://www-data:www-data@localhost/gis

Using MySQL

Installation

On Debian-based systems, MySQL can be installed with:

$ sudo apt-get install mysql-server mysql-client

Hint

More information about installing MySQL can be found in the MySQL manual.

You will also have to install a Python driver for MySQL. Run the following command inside the virtual environment to install the library mysql-python:

(venv) $ easy_install mysql-python

Setting up a spatially-enabled database

For MySQL setting up a spatial database is the same as setting up an ordinary database:

mysql -u root -p
create database gis;
quit

Note

When creating a database, or directly when creating a table, you can specify the storage engine (MyISAM/InnoDB/..) that should be used for the tables. Note that currently only MyISAM uses a R-tree for spatial indexes that optimize spatial queries (see Optimizing Spatial Analysis in the MySQL manual).

To create a database user that can be used for accessing the database from a web application, execute the following statements:

mysql -u root -p
create user 'www-data' identified by 'www-data';
grant all on gis.* to 'www-data';
quit

Tables with geometry columns can be created like any other table using the geometry type as column data type (see Creating Spatial Columns in the MySQL manual), for example:

mysql -u root -p
use gis;
CREATE TABLE points (id INTEGER AUTO_INCREMENT,
        name VARCHAR(40),
        geom POINT NOT NULL,
        SPATIAL INDEX(geom),
        PRIMARY KEY(id));
quit

Note

If you want to use a spatial index for your geometry column, the column must be declared as NOT NULL.

Configuration

Set the database connection string in the configuration file of your MapFish application (for example development.ini) by replacing the line:

sqlalchemy.url = sqlite:///%(here)s/development.db

by this one:

sqlalchemy.url = mysql://www-data:www-data@localhost/gis

Limitations

  • MySQL does not support coordinate system transformations for geometries. All your spatial data must be in the same spatial reference system.
  • Not all methods of the OpenGIS® Simple Features Specifications For SQL are supported by MySQL, for example distance() or buffer() are not part of the stable release. A list of these functions can be found here. Other functions only operate on the minimum bounding rectangle (MBR) of the geometries, a list of these functions can be found in the MySQL manual. Because of that, features queried through the MapFish Protocol are also selected using the MBR.

Using SQLite/Spatialite

Installation

Installation of Spatialite

Spatialite requires the libraries GEOS and PROJ4, which in most cases you will have already installed together with PostGIS:

$ sudo apt-get install libgeos-c1 proj

Note

Spatialite expects libgeos 3.1.1, but it can also be used with any 3.0.x release. You just have to create a symbolic link:

sudo ln /usr/lib/libgeos-3.0.0.so /usr/lib/libgeos-3.1.1.so

Now download the precompiled Spatialite library libspatialite from the Spatialite download page and unzip the archive to /usr/local/lib/libspatialite or into a folder of your convenience:

wget http://www.gaia-gis.it/spatialite/libspatialite-linux-x86-2.3.1.tar.gz
sudo tar -xvf libspatialite-linux-x86-2.3.1.tar.gz -C /usr/local/lib/
sudo mv /usr/local/lib/libspatialite-* /usr/local/lib/libspatialite

Hint

On Ubuntu 9.10+ you can install the Spatialite library as package libspatialite2 directly from the repositories.

Note

The precompiled libraries for Spatialite only work on 32-bit systems, if you are using a 64-bit system you will have to compile by yourself. To do so, download the source code for libspatialite-amalgamation from the Spatialite website. Make sure that you also have installed the package libgeos-dev. Unzip the source archive and compile by using the following commands (you may want to change the prefix path, the compiled library will be copied there):

./configure --prefix=/home/c2c/libspatialite --with-geos-lib=/usr/lib --with-proj-lib=/usr/lib
make install

If you are receiving the error message cannot find -lstdc++, you may have to create a symbolic link for this library:

sudo ln /usr/lib/libstdc++.so.6 /usr/lib/libstdc++.so

Further information about compiling libspatialite can be found here: How to build libspatialite.

Installation of pysqlite2

Even though Python 2.5+ contains the SQLite driver pysqlite2, you have to compile it by yourself. The Spatialite library is used in SQLite as extension, and by default loading external extensions is disabled in pysqlite2.

To compile pysqlite2 you will have to install the SQLite header files:

sudo apt-get install libsqlite3-dev

Download the pysqlite2 source code from pysqlite2 - Downloads and unzip it into your MapFish virtual environment.

Then open the file setup.cfg and comment out the line define=SQLITE_OMIT_LOAD_EXTENSION:

[build_ext]
#define=
#include_dirs=/usr/local/include
#library_dirs=/usr/local/lib
libraries=sqlite3
#define=SQLITE_OMIT_LOAD_EXTENSION

Now you can compile and setup pysqlite2 with:

(venv) $ python setup.py install

Note

If you are running into Segmentation Fault errors using this build, you can try to do a static build. This will download the latest SQLite3 amalgation file and link it internally:

(venv) $ python setup.py buildstatic install

Setting up a spatially-enabled database

Creating a database can be done using spatialite-gui or by using the CLI client spatialite. In the following we will use spatialite, but you can also use spatialite-gui to execute the commands.

Hint

You can get spatialite-gui from the Spatialite website, also take a look at the Quickguide for spatialite-gui (PDF).

First you have to download the package spatialite-tools from Spatialite Downloads and the script init_spatialite-2.3.sql from Spatialite Ressources. The scripts creates the geometry_columns and spatial_ref_sys metadata tables and also inserts a collection of spatial reference systems.

Start the Spatialite client by calling:

spatialite gis.sqlite

This will create the file gis.sqlite, if it does not exist already. Then execute the script init_spatialite-2.3.sql:

.read init_spatialite-2.3.sql ASCII

Now you can create a table with a geometry column. This is done in two steps: First we create a plain SQLite table without the geometry column, and then we add the geometry column using the function AddGeometryColumn():

CREATE TABLE points (id INTEGER PRIMARY KEY AUTOINCREMENT, name VARCHAR(40));
SELECT AddGeometryColumn('points', 'geom', 4326, 'POINT', 2);

Hint

You can also create a table from a Shapefile with .loadshp (see also Creating a new SpatiaLite db and populating it):

.loadshp countries Countries utf-8

And you can even execute queries directly on Shapefiles without copying the data into a table (see also Performing SQL queries directly on shapefiles):

CREATE VIRTUAL TABLE virtual_countries USING VirtualShape('/home/c2c/data/countries', utf-8, 4326);
select count(*) from virtual_countries where
        MBRWithin(Geometry, GeomFromText('POLYGON((0 0, 40 0, 40 40, 0 40, 0 0))', 4326));

Currently only read operations are supported, but still virtual tables are a good option to publish a Shapefile with MapFish.

Configuration

When using spatialite-gui and spatialite the Spatialite library is automatically loaded as extension. But when connecting to a Spatialite database using a ordinary SQLite driver, you have to load the Spatialite library manually.

In MapFish, database connections are managed by SQLAlchemy. Every time SQLAlchemy opens a new connection to a Spatalite database, the Spatialite library must be loaded. This can be done by setting up a PoolListener.

Open the file model/__init__.py and modify the method init_model(engine), so that it looks like this:

# ...

from sqlalchemy.dialects.sqlite.base import SQLiteDialect
from sqlalchemy.interfaces import PoolListener

# ...

def init_model(engine):
    sm = orm.sessionmaker(autoflush=True, autocommit=False, bind=engine)

    meta.engine = engine
    meta.Session = orm.scoped_session(sm)

    if isinstance(engine.dialect, SQLiteDialect):
        class SpatialiteConnectionListener(PoolListener):
            def connect(self, dbapi_con, con_record):
                dbapi_con.enable_load_extension(True)
                dbapi_con.execute("select load_extension('/usr/local/lib/libspatialite/lib/libspatialite.so')")
                dbapi_con.enable_load_extension(False)

        engine.pool.add_listener(SpatialiteConnectionListener())

        # ...

Now you just have to set the database connection string in the configuration file of your MapFish application (for example development.ini) by replacing the line:

sqlalchemy.url = sqlite:///%(here)s/development.db

by this one:

sqlalchemy.url = sqlite:////home/c2c/data/gis.sqlite

Hint

The number of slashs to the right of sqlite: depends on if you are using a relative or absolute path, see also SQLite: Connect Strings.

Using Oracle

Installation

The Python driver cx_Oracle requires an Oracle client or server installation. If your MapFish application is running on the same system as your Oracle database, you can skip the section Installation of Oracle Instant Client.

Note

The installation of Oracle database server software is not covered in this tutorial, please refer to the Oracle Database Documentation.

For guidance on installing Oracle on Debian based systems, take a look at these two tutorials:

Installation of Oracle Instant Client

The easiest way to get an Oracle Client is installing Oracle Instant Client. Download the following two packages for your operating system from Instant Client Downloads:

  • Instant Client Package - Basic
  • Instant Client Package - SDK

In the following we are using RPM files for an installation on a Debian based system.

First install the required packages alien and libaio1:

sudo apt-get install alien libaio1

Install the two RPM packages:

sudo alien -i oracle-instantclient11.2-basic-11.2.0.1.0-1.i386.rpm
sudo alien -i oracle-instantclient11.2-devel-11.2.0.1.0-1.i386.rpm

To add the installed libraries to the system search path, create the file /etc/ld.so.conf.d/oracle.conf and insert the path to your installation, for example:

/usr/lib/oracle/11.2/client/lib/

Then run ldconfig to update the library cache:

sudo ldconfig
ldconfig -p | grep oracle

The last command should print out the Oracle library files.

Installation of cx_Oracle

Installer files for various operating systems and Oracle versions can be found on the cx_Oracle website. In the following we are building cx_Oracle from source. To do so, download the source code archive from cx_Oracle - Files.

Before building cx_Oracle three environment variables have to be set:

export ORACLE_HOME=/usr/lib/oracle/11.2/client/
export LD_LIBRARY_PATH=$ORACLE_HOME/lib
export PATH=$ORACLE_HOME/bin:$PATH

Then start the setup inside your virtual environment:

(venv) $ python setup.py install

To test if the installation was succesfull, start a Python interpreter and try to import the cx_Oracle module:

(venv) $ python -i
>>> import cx_Oracle
>>>

Hint

If you are using Buildout you may want to take a look at the recipe gocept.cxoracle, which automatically installs cx_Oracle and Oracle Instant Client.

Configuration

When using cx_Oracle inside MapFish, the environment variables ORACLE_HOME and LD_LIBRARY have to be set before the cx_Oracle module is used from SQLAlchemy. You can do this in the file config/environment.py before the SQLAlchemy engine is created from the configuration:

def load_environment(global_conf, app_conf):

   # ...

   # Set the evironment variables required for cx_Oracle
   os.environ['ORACLE_HOME'] = '/usr/lib/oracle/11.2/client/'
   os.environ['LD_LIBRARY'] = '/usr/lib/oracle/11.2/client/lib'

   # Setup the SQLAlchemy database engine
   engine = engine_from_config(config, 'sqlalchemy.')
   init_model(engine)

Then set the database connection string in your Pylons configuration file development.ini:

sqlalchemy.url = oracle://www-data:www-data@localhost:1521/gis

Notes about the Oracle dimension information array

Oracle requires a Dimension Information Array (DIMINFO) in its geometry metadata table for every spatial column (see Oracle® Spatial User’s Guide and Reference: Geometry Metadata Views).

If you are creating your tables with SQLAlchemy/GeoAlchemy (see Using ‘paster setup-app’ to create your database tables), you will have to specify a DIMINFO in your model files. GeoAlchemy then will make an entry in the view USER_SDO_GEOM_METADATA (see also Oracle® Spatial User’s Guide and Reference: DIMINFO).

Example definition (model/places.py):

from sqlalchemy import Column, types
from sqlalchemy.schema import Sequence

from geoalchemy import GeometryColumn, Point, GeometryDDL

from mapfish.sqlalchemygeom import GeometryTableMixIn
from mapfishsample.model.meta import engine, Base

diminfo = "MDSYS.SDO_DIM_ARRAY("\
            "MDSYS.SDO_DIM_ELEMENT('LONGITUDE', -180, 180, 0.000000005),"\
            "MDSYS.SDO_DIM_ELEMENT('LATITUDE', -90, 90, 0.000000005)"\
            ")"

class Place(Base, GeometryTableMixIn):
    __tablename__ = 'places'
    id = Column(types.Integer, Sequence('place_id_seq'), primary_key=True)
    name = Column(types.String(40))

    the_geom = GeometryColumn(Point(dimension=2, srid=4326, diminfo=diminfo))

GeometryDDL(Place.__table__)

For Oracle MapFish in general uses the operator SDO_WITHIN_DISTANCE for spatial filter queries. If the filter geometry (Lat/Lon, BBox or arbitrary geometry) uses a different SRID, the geometry column has to be reprojected to this SRID. In this case the Oracle function SDO_GEOM.WITHIN_DISTANCE has to be used which either requires dimension information arrays or a tolerance value. These parameters have to be set on custom filters for the method index() inside the controller classes.

Example (controllers/places.py):

class PlacesController(BaseController):
    # ..

    def index(self, format='json'):
        """GET /: return all features."""
        filter = create_default_filter(request, Place, additional_params={'tol': '0.005'})
        return self.protocol.read(request, filter=filter)
    # ..

The tolerance value will be passed to SDO_GEOM.WITHIN_DISTANCE. Alternatively you can set a DIMINFO for the reprojected geometry column and the filter query using the keywords dim1 and dim2.

Note that this tolerance is not the one used in the MapFish Protocol to specify a distance in which features should be queried.

Hint

More information about using Oracle can be found in the GeoAlchemy documentation.

Notes about using a tolerance value for MapFish Web Services in Oracle

Requests to a MapFish web service can contain a tolerance parameter, which specifies within which distance features should be queried. Usually the unit of this value is the unit associated with the coordinate system in use (for example degree for EPSG:4326). But for geodetic coordinate systems (like EPSG:4326) Oracle uses meter as unit. You have to keep this in mind when developing applications for Oracle.

If you want to use a different unit, you can set it as parameter in your controller files. This parameter is passed to the database function call without further checks.

Example (controllers/places.py):

# ..

def index(self, format='json'):
    """GET /: return all features."""
    filter = create_default_filter(request, Place, additional_params={'params': 'unit=KM'})
    return self.protocol.read(request, filter=filter)

# ..

Valid units are listed in the view SDO_DIST_UNITS.

Sperical Mercator (900913)

For many map tiles (including OpenStreetMap and Google Maps) Sperical Mercator (EPSG: 900913) is used as projection. For Spatialite and PostGIS in releases prior to PostGIS 1.4, Sperical Mercator is not supported by default and you will receive an error message like the following, when you try to work with geometries using Sperical Mercator as spatial reference system (SRS):

(InternalError) AddToPROJ4SRSCache: Cannot find SRID (900913) in spatial_ref_sys [..]

To enable support for Sperical Mercator, you first will have to update the library proj, because on Ubuntu proj comes without datum shifting files which are required for transformations with the Sperical Mercator projection. Run the following commands to update your installation (see also Notes on proj):

wget http://download.osgeo.org/proj/proj-datumgrid-1.4.tar.gz
mkdir nad
cd nad
tar xzf ../proj-datumgrid-1.4.tar.gz
nad2bin null < null.lla
sudo cp null /usr/share/proj

Then for PostGIS run the following statements to insert the reference definition in PostGIS’ spatial_ref_sys table:

sudo su postgres
psql gis
INSERT INTO spatial_ref_sys (srid, auth_name, auth_srid, srtext, proj4text) VALUES (900913, 'EPSG', 900913, 'PROJCS["unnamed",GEOGCS["unnamed ellipse",DATUM["unknown",SPHEROID["unnamed",6378137,0]], PRIMEM["Greenwich",0],UNIT["degree",0.0174532925199433]], PROJECTION["Mercator_2SP"],PARAMETER["standard_parallel_1",0],PARAMETER["central_meridian",0],PARAMETER["false_easting",0], PARAMETER["false_northing",0],UNIT["Meter",1],EXTENSION["PROJ4","+proj=merc +a=6378137 +b=6378137 +lat_ts=0.0 +lon_0=0.0 +x_0=0.0 +y_0=0 +k=1.0 +units=m +nadgrids=@null +wktext  +no_defs"]]', '+proj=merc +a=6378137 +b=6378137 +lat_ts=0.0 +lon_0=0.0 +x_0=0.0 +y_0=0 +k=1.0 +units=m +nadgrids=@null +wktext  +no_defs');
\q
exit

The same can be done for Spatialite:

spatialite gis.sqlite
INSERT INTO spatial_ref_sys (srid, auth_name, auth_srid, ref_sys_name, proj4text) VALUES (900913, 'EPSG', 900913, 'Google Sperical Mercator', '+proj=merc +a=6378137 +b=6378137 +lat_ts=0.0 +lon_0=0.0 +x_0=0.0 +y_0=0 +k=1.0 +units=m +nadgrids=@null +wktext  +no_defs');
.quit

Table mapping

Declarative and non-declarative mapping

With SQLAlchemy table mappings can be configured either by defining a table and a class separetly (non-declarative) or by doing this at once (declarative). MapFish supports both ways. The following two examples show the mapping for a table spots, one time as non-declarative and one time as declarative mapping.

Non-declarative mapping:

from sqlalchemy import Column, Table, Integer, Numeric
from sqlalchemy.orm import mapper

from geoalchemy import (Geometry, GeometryColumn,
GeometryDDL, GeometryExtensionColumn)

from mapfish.sqlalchemygeom import GeometryTableMixIn

from mapfishsample.model.meta import metadata

# table definition
spots_table = Table('spots', metadata,
        Column('spot_id', Integer, primary_key=True),
        Column('spot_height', Numeric(asdecimal=False)),
        GeometryExtensionColumn('spot_location', Geometry(2)))

# class definition
class Spot(GeometryTableMixIn):
    __table__ = spots_table

    def __init__(self, spot_id=None, spot_height=None, spot_location=None):
        self.spot_id = spot_id
        self.spot_height = spot_height
        self.spot_location = spot_location

# set up the mapping between table and class
mapper(Spot, spots_table, properties={
            'spot_location': GeometryColumn(spots_table.c.spot_location,
                                            comparator=PGComparator)})

# register table for DDL extension, so that it can be created from SQLAlchemy
GeometryDDL(spots_table)

Declarative mapping:

from sqlalchemy import Column, types

from geoalchemy import GeometryColumn, Geometry

from mapfish.sqlalchemygeom import GeometryTableMixIn
from mapfishsample.model.meta import engine, Base

class Spot(Base, GeometryTableMixIn):
    __tablename__ = 'spots'

    spot_id = Column(Integer, primary_key=True)
    spot_height = Column(Numeric(asdecimal=False))
    spot_location = GeometryColumn(Point(2), comparator=PGComparator)

GeometryDDL(Spot.__table__)

Geometry column properties

When using paster mf-layer or paster mf-model, MapFish creates a default configuration for the geometry column of your table. You may want to customize this configuration to your needs.

Example configuration for model/spots.py:

# [..]
spot_location = GeometryColumn(
        Point(dimension=2, srid=4326, spatial_index=True),
        comparator=PGComparator,
        nullable=False)
# [..]

dimension=2

The dimension of the geometry (default: 2).

srid=4326

The spatial reference system (SRS) of the geometry column as EPSG code (default: 4326).

spatial_index=True

Indicates if a spatial index is created for the geometry column (default: True).

Hint

Spatialite does not automatically make use of the spatial index when executing queries, you explicitly have to access the spatial index in your queries, see Spatial Index: using SQLite’s R*Tree.

comparator=PGComparator

You only have to set this option, when you want to use a database specific function (like AsKML in PostGIS) on a geometry column in a SQLAlchemy query (for example session.query(Spot).filter(Spot.geom.kml == '..'). Following comparators are available:

  • PostGIS: geoalchemy.postgis.PGComparator
  • MySQL: geoalchemy.mysql.MySQLComparator
  • Spatialite: geoalchemy.spatialite.SQLiteComparator
  • Oracle: geoalchemy.oracle.OracleComparator

nullable=False

Indicates if null values can be inserted into the geometry column (default: True).

Note

When using MySQL with a spatial index, the parameter nullable is ignored, because MySQL requires a NOT NULL constraint for spatial indexed columns.

Using ‘paster setup-app’ to create your database tables

When setting up a Pylons application, you often have to create database tables to run the application. SQLAlchemy/GeoAlchemy can take over that task for you, so that all your tables are created just by calling paster setup-app config.ini.

The following steps describe how to configure your application.

  1. Table mapping

    All columns that you want to be created for a table, must be enlisted in the table definition.

    Example: model/Point.py

    from sqlalchemy import Column, types
    
    from geoalchemy import GeometryColumn, Point, GeometryDDL
    
    from mapfish.sqlalchemygeom import GeometryTableMixIn
    from mapfishsample.model.meta import metadata, Base
    
    
    class Point(Base, GeometryTableMixIn):
        __tablename__ = 'points'
    
        id = Column(types.Integer, primary_key=True)
        name = Column(types.String(30), default = 'foo')
        the_geom = GeometryColumn(Point(dimension=2, srid=4326))
    
    GeometryDDL(Point.__table__)
    

    Note the last line GeometryDDL(Point.__table__), this makes sure that GeoAlchemy creates the geometry field of the table.

  2. websetup.py

    When calling paster setup-app config.ini, the method setup_app() inside the file [your_app]/websetup.py is executed. By default the method setup_app() already contains the command metadata.create_all() that creates the tables. You just have to import your model classes.

    Example: websetup.py

    """Setup the MapFishSample application"""
    import logging
    
    from mapfishsample.config.environment import load_environment
    from mapfishsample.model import meta
    
    # Import the model classes you want to create the tables for
    from mapfishsample.model import points
    
    log = logging.getLogger(__name__)
    
    def setup_app(command, conf, vars):
        """Place any commands to setup mapfishsample here"""
        load_environment(conf.global_conf, conf.local_conf)
    
        # Create the tables if they don't already exist
        meta.metadata.create_all(bind=meta.engine)
    
  3. paster setup-app

    Finally to setup your application, run the following command inside the virtual environment:

    (venv) $ paster setup-app [your_config].ini