EVOLUTION-MANAGER

Edit File: drv_pg.html

<html>
<head>
<title>PostgreSQL / PostGIS</title>
</head>

<h1>PostgreSQL / PostGIS</h1>

This driver implements support for access to spatial tables in
PostgreSQL extended with the
<a href="http://postgis.net/">PostGIS</a> spatial data support.
Some support exists in the driver for use with PostgreSQL without PostGIS
but with less functionalities.

This driver requires a connection to a Postgres database. If you want to
prepare a SQL dump to inject it later into a Postgres database, you can
instead use the <a href="drv_pgdump.html">PostgreSQL SQL Dump driver</a> (GDAL/OGR >= 1.8.0)

You can find additional information on the driver in the <a href="drv_pg_advanced.html">Advanced OGR PostgreSQL driver Information</a> page.

<h2>Connecting to a database</h2>

To connect to a Postgres datasource, use a connection string specifying the database name,
with additional parameters as necessary 
<blockquote><pre>PG:dbname=databasename</pre> or<pre>PG:"dbname='databasename' host='addr' port='5432' user='x' password='y'"</pre></blockquote>
It's also possible to omit the database name and connect
to a default database, with the same name as the user name. 
Note: We use PQconnectdb() to make the connection, so any other options and defaults
that would apply to it, apply to the name here (refer to the documentation of the PostgreSQL server.
Here for <a href="http://www.postgresql.org/docs/8.4/interactive/libpq-connect.html">PostgreSQL 8.4</a>).
The PG: prefix is used to mark the name as a postgres connection string.

<h2>Geometry columns</h2>

If the geometry_columns table exists (i.e. PostGIS is enabled for the accessed
database), then all tables and named views listed in the geometry_columns table
will be treated as OGR layers. Otherwise (PostGIS disabled for the accessed database),
all regular user tables and named views will be treated as layers.

Starting with GDAL 1.7.0, the driver also supports the
<a href="http://postgis.refractions.net/documentation/manual-svn/ch04.html#PostGIS_Geography">geography</a>
column type introduced in PostGIS 1.5.

Starting with GDAL 2.0, the driver also supports reading and writing the
following non-linear geometry types :CIRCULARSTRING, COMPOUNDCURVE, CURVEPOLYGON, MULTICURVE and MULTISURFACE

<h2>SQL statements</h2>

The PostgreSQL driver passes SQL statements directly to PostgreSQL by default,
rather than evaluating them internally when using the ExecuteSQL() call on the
OGRDataSource, or the -sql command option to ogr2ogr. Attribute query
expressions are also passed directly through to PostgreSQL.
It's also possible to request the ogr Pg driver to handle SQL commands
with the <a href="ogr_sql.html">OGR SQL</a> engine, by passing "OGRSQL"
string to the ExecuteSQL() method, as the name of the SQL dialect.

The PostgreSQL driver in OGR supports the OGRDataSource::StartTransaction(),
OGRDataSource::CommitTransaction() and OGRDataSource::RollbackTransaction()
calls in the normal SQL sense.

<h2>Creation Issues</h2>

The PostgreSQL driver does not support creation of new datasets (a database
within PostgreSQL), but it does allow creation of new layers within an
existing database.

As mentioned above the type system is impoverished, and many OGR types
are not appropriately mapped into PostgreSQL.

If the database has PostGIS types loaded (i.e. the geometry type), newly
created layers will be created with the PostGIS Geometry type. Otherwise
they will use OID.

By default it is assumed that text being sent to Postgres is in the UTF-8
encoding. This is fine for plain ASCII, but can result in errors for
extended characters (ASCII 155+, LATIN1, etc). While OGR provides no direct
control over this, you can set the PGCLIENTENCODING environment variable
to indicate the format being provided. For instance, if your text is
LATIN1 you could set the environment variable to LATIN1 before using OGR
and input would be assumed to be LATIN1 instead of UTF-8.
An alternate way of setting the client encoding is to issue the following SQL command
with ExecuteSQL() : "SET client_encoding TO encoding_name" where encoding_name is LATIN1, etc.
Errors can be caught by enclosing this command with a CPLPushErrorHandler()/CPLPopErrorHandler() pair.

<h3>Dataset open options</h3>

(GDAL &gt;= 2.0)

<ul>
<li> DBNAME=string: Database name.
<li> PORT=integer: Port.
<li> USER=string: User name.
<li> PASSWORD=string: Password.
<li> HOST=string: Server hostname.
<li> ACTIVE_SCHEMA=string: Active schema.
<li> SCHEMAS=string: Restricted sets of schemas to explore (comma separated).
<li> TABLES=string: Restricted set of tables to list (comma separated).
<li> LIST_ALL_TABLES=YES/NO: This may be "YES" to force all tables,
including non-spatial ones, to be listed.
<li> PRELUDE_STATEMENTS=string (GDAL &gt;= 2.1). SQL statement(s) to send on the
PostgreSQL client connection before any other ones. In case of several statement,
they must be separated with the semi-column (;) sign. The driver will specifically
recognize BEGIN as the first statement to avoid emitting BEGIN/COMMIT itself.
This option may be useful when using the driver with pg_bouncer in transaction pooling,
e.g. 'BEGIN; SET LOCAL statement_timeout TO "1h";'
<li> CLOSING_STATEMENTS=string (GDAL &gt;= 2.1). SQL statement(s) to send on the
PostgreSQL client connection after any other ones. In case of several statement,
they must be separated with the semi-column (;) sign. With the above example value
for PRELUDE_STATEMENTS, the appropriate CLOSING_STATEMENTS would be "COMMIT".
</ul>

<h3>Dataset Creation Options</h3>

None

<h3>Layer Creation Options</h3>

<ul>
<li>
GEOM_TYPE: The GEOM_TYPE layer creation option can be set to
one of "geometry", "geography" (PostGIS >= 1.5), "BYTEA" or "OID" to force the type of geometry used for
a table. For a PostGIS database, "geometry" is the default value.
<li> OVERWRITE: This may be "YES" to force an existing layer of the
desired name to be destroyed before creating the requested layer.
<li> LAUNDER: This may be "YES" to force new fields created on this
layer to have their field names "laundered" into a form more compatible with
PostgreSQL. This converts to lower case and converts some special characters
like "-" and "#" to "_". If "NO" exact names are preserved.
The default value is "YES". If enabled the table (layer) name will also be laundered.
<li> PRECISION: This may be "YES" to force new fields created on this
layer to try and represent the width and precision information, if available
using NUMERIC(width,precision) or CHAR(width) types. If "NO" then the types
FLOAT8, INTEGER and VARCHAR will be used instead. The default is "YES".
<li> DIM={2,3,XYM,XYZM}: Control the dimension of the layer.
Important to set to 2 for 2D layers with PostGIS 1.0+ as it has constraints
on the geometry dimension during loading.
<li> GEOMETRY_NAME: Set name of geometry column in new table. If
omitted it defaults to wkb_geometry for GEOM_TYPE=geometry, or the_geog for GEOM_TYPE=geography.
<li> SCHEMA: Set name of schema for new table.
Using the same layer name in different schemas is supported, but not in the public schema and others. Note that
using the -overwrite option of ogr2ogr and -lco SCHEMA= option at the same time will not work, as the ogr2ogr utility
will not understand that the existing layer must be destroyed in the specified schema. Use the -nln option of ogr2ogr instead,
or better the active_schema connection string. See below example.
<li> SPATIAL_INDEX: (From GDAL 1.6.0) Set to YES by default. Creates a spatial index (GiST) on the geometry column
to speed up queries. Set to FALSE to disable. (Has effect only when PostGIS is available).
<li> TEMPORARY: (From GDAL 1.8.0) Set to OFF by default. Creates a temporary table instead of a permanent one.
<li> UNLOGGED: (From GDAL 2.0) Set to OFF by default. Whether to create the table as a unlogged one.
Unlogged tables are only supported since PostgreSQL 9.1, and GiST indexes used for spatial indexing since PostgreSQL 9.3.
<li> NONE_AS_UNKNOWN: (From GDAL 1.8.1) Can bet set to TRUE to force non-spatial layers (wkbNone) to be created as
spatial tables of type GEOMETRY (wkbUnknown), which was the behaviour prior to GDAL 1.8.0. Defaults to NO, in which case
a regular table is created and not recorded in the PostGIS geometry_columns table.
<li> FID: (From GDAL 1.9.0) Name of the FID column to create. Defaults to 'ogc_fid'.
<li> FID64: (From GDAL 2.0) This may be "TRUE" to create a FID column that can support
64 bit identifiers. The default value is "FALSE".</li>
<li> EXTRACT_SCHEMA_FROM_LAYER_NAME: (From GDAL 1.9.0) Can be set to NO to avoid considering the dot character
as the separator between the schema and the table name. Defaults to YES.
<li> COLUMN_TYPES: (From GDAL 1.10) A list of strings of format field_name=pg_field_type (separated by comma)
that should be use when CreateField() is invoked on them. This will override the default choice that OGR would have made.
This can for example be used to create a column of type <a href="http://www.postgresql.org/docs/9.0/static/hstore.html">HSTORE</a>.
<li> DESCRIPTION (From GDAL 2.1) Description string to put in the pg_description system table. On
reading, if such a description is found, it is exposed in the DESCRIPTION metadata item. The description
can also be written with SetMetadataItem("DESCRIPTION", description_string). Descriptions are preserved
by default by ogr2ogr, unless the -nomd option is used.
</ul>

<h3>Configuration Options</h3>

There are a variety of
<a href="http://trac.osgeo.org/gdal/wiki/ConfigOptions">Configuration
Options</a> which help control the behavior of this driver.

<ul>
<li>PG_USE_COPY: This may be "YES" for using COPY for inserting data to Postgresql.
COPY is significantly faster than INSERT. Starting with GDAL 2.0, COPY is used by
default when inserting from a table that has just been created.</li>
<li>PGSQL_OGR_FID: Set name of primary key instead of 'ogc_fid'. Only used when opening a layer whose primary key cannot be autodetected.
Ignored by CreateLayer() that uses the FID creation option.</li>


<li>PG_USE_BASE64: (GDAL >= 1.8.0) If set to "YES", geometries will be fetched as BASE64 encoded EWKB instead of canonical HEX encoded EWKB.
This reduces the amount of data to be transferred from 2 N to 1.333 N, where N is the size of EWKB data. However, it might be a
bit slower than fetching in canonical form when the client and the server are on the same machine, so the default is NO.</li>
<li>OGR_TRUNCATE: (GDAL &gt;= 1.11) If set to "YES", the content of the table will be first erased with the SQL TRUNCATE command before
inserting the first feature. This is an alternative to using the -overwrite flag of ogr2ogr,
that avoids views based on the table to be destroyed.
Typical use case: "ogr2ogr -append PG:dbname=foo abc.shp --config OGR_TRUNCATE YES".
</ul>

<h3>Examples</h3>

<ul>
<li>
Simple translation of a shapefile into PostgreSQL. The table 'abc' will
be created with the features from abc.shp and attributes from abc.dbf.
The database instance (warmerda) must already exist, and the table abc must
not already exist.

<pre>
% ogr2ogr -f PostgreSQL PG:dbname=warmerda abc.shp
</pre>

<li>

This second example loads a political boundaries layer from VPF (via the
<a href="drv_ogdi.html">OGDI driver</a>), and renames the layer from the
cryptic OGDI layer name to something more sensible. If an existing table
of the desired name exists it is overwritten.

<pre>
% ogr2ogr -f PostgreSQL PG:dbname=warmerda \
 gltp:/vrf/usr4/mpp1/v0eur/vmaplv0/eurnasia \
 -lco OVERWRITE=yes -nln polbndl_bnd 'polbndl@bnd(*)_line'
</pre>
</li>

<li>

In this example we merge tiger line data from two different directories of
tiger files into one table. Note that the second invocation uses -append
and no OVERWRITE=yes.

<pre>
% ogr2ogr -f PostgreSQL PG:dbname=warmerda tiger_michigan \
 -lco OVERWRITE=yes CompleteChain
% ogr2ogr -update -append -f PostgreSQL PG:dbname=warmerda tiger_ohio \
 CompleteChain
</pre>
</li>

<li>

This example shows using ogrinfo to evaluate an SQL query statement
within PostgreSQL. More sophisticated PostGIS specific queries may also be
used via the -sql commandline switch to ogrinfo.

<pre>
ogrinfo -ro PG:dbname=warmerda -sql "SELECT pop_1994 from canada where province_name = 'Alberta'"
</pre>
</li>

<li>

This example shows using ogrinfo to list PostgreSQL/PostGIS layers on a different host.

<pre>
ogrinfo -ro PG:'host=myserver.velocet.ca user=postgres dbname=warmerda'
</pre>
</li>

<li>

This example shows use of PRELUDE_STATEMENTS and CLOSING_STATEMENTS as
destination open options of ogr2ogr.

<pre>
ogrinfo "pg:dbname=mydb" poly.shp -doo "PRELUDE_STATEMENTS=BEGIN; SET LOCAL statement_timeout TO '1h';" -doo CLOSING_STATEMENTS=COMMIT
</pre>
</li>

</ul>

<ul>
<li> Why can't I see my tables? PostGIS is installed and I have data 
You must have permissions on all tables you want to read and geometry_columns and spatial_ref_sys. 
Misleading behavior may follow without an error message if you do not have permissions to these tables. Permission
issues on geometry_columns and/or spatial_ref_sys tables can be generally confirmed if you can see the tables
by setting the configuration option PG_LIST_ALL_TABLES to YES. (e.g. ogrinfo --config PG_LIST_ALL_TABLES YES PG:xxxxx)
</ul>

<ul>
<li> <a href="drv_pg_advanced.html">Advanced OGR PostgreSQL driver Information</a>
<li> <a href="drv_pgdump.html">OGR PostgreSQL SQL Dump driver Page</a>
<li> <a href="http://www.postgresql.org/">PostgreSQL Home Page</a>
<li> <a href="http://postgis.net/">PostGIS</a>
<li> <a href="http://trac.osgeo.org/postgis/wiki/UsersWikiOGR">PostGIS / OGR Wiki Examples Page</a>
</ul>

</body>
</html>