Virtual Format

OGR Virtual Format is a driver that transforms features read from other drivers based on criteria specified in an XML control file. It is primarily used to derive spatial layers from flat tables with spatial information in attribute columns. It can also be used to associate coordinate system information with a datasource, merge layers from different datasources into a single data source, or even just to provide an anchor file for access to non-file oriented datasources.

The virtual files are currently normally prepared by hand.

Creation Issues

Before GDAL 1.7.0, the OGR VRT driver was read-only.

Since GDAL 1.7.0, the CreateFeature(), SetFeature() and DeleteFeature() operations are supported on a layer of a VRT dataset, if the following conditions are met :

the VRT dataset is opened in update mode
the underlying source layer supports those operations
the SrcLayer element is used (as opposed to the SrcSQL element)
the FID of the VRT features is the same as the FID of the source features, that is to say, the FID element is not specified

Virtual File Format

The root element of the XML control file is OGRVRTDataSource. It has an OGRVRTLayer child for each layer in the virtual datasource. That element should have a name attribute with the layer name, and may have the following subelements:

SrcDataSource (mandatory): The value is the name of the datasource that this layer will be derived from. The element may optionally have a relativeToVRT attribute which defaults to "0", but if "1" indicates that the source datasource should be interpreted as relative to the virtual file. This can be any OGR supported dataset, including ODBC, CSV, etc. The element may also have a shared attribute to control whether the datasource should be opened in shared mode. Defaults to OFF for SrcLayer use and ON for SrcSQL use.
SrcLayer (optional): The value is the name of the layer on the source data source from which this virtual layer should be derived. If this element isn't provided, then the SrcSQL element must be provided.
SrcSQL (optional): An SQL statement to execute to generate the desired layer result. This should be provided instead of the SrcLayer for statement derived results. Some limitations may apply for SQL derived layers.
FID (optional): Name of the attribute column from which the FID of features should be derived. If not provided, the FID of the source features will be used directly.
Style (optional): Name of the attribute column from which the style of features should be derived. If not provided, the style of the source features will be used directly.
GeometryType (optional): The geometry type to be assigned to the layer. If not provided it will be taken from the source layer. The value should be one of "wkbNone", "wkbUnknown", "wkbPoint", "wkbLineString", "wkbPolygon", "wkbMultiPoint", "wkbMultiLineString", "wkbMultiPolygon", or "wkbGeometryCollection". Optionally "25D" may be appended to mark it as including Z coordinates. Defaults to "wkbUnknown" indicating that any geometry type is allowed.
LayerSRS (optional): The value of this element is the spatial reference to use for the layer. If not provided, it is inherited from the source layer. The value may be WKT or any other input that is accepted by the OGRSpatialReference::SetUserInput() method. If the value is NULL, then no SRS will be used for the layer.
GeometryField (optional): This element is used to define how the geometry for features should be derived.
If not provided the geometry of the source feature is copied directly.
The type of geometry encoding is indicated with the encoding attribute which may have the value "WKT", "WKB" or "PointFromColumns".
If the encoding is "WKT" or "WKB" then the field attribute will have the name of the field containing the WKT or WKB geometry.
If the encoding is "PointFromColumns" then the x, y and z attributes will have the names of the columns to be used for the X, Y and Z coordinates. The z attribute is optional.
Starting with GDAL 1.7.0, the optional reportSrcColumn attribute can be used to specify whether the source geometry fields (the fields set in the field, x, x or z attributes) should be reported as fields of the VRT layer. It defaults to TRUE. If set to FALSE, the source geometry fields will only be used to build the geometry of the features of the VRT layer.
SrcRegion (optionnal, from GDAL 1.7.0) : This element is used to define an initial spatial filter for the source features. This spatial filter will be combined with any spatial filter explicitely set on the VRT layer with the SetSpatialFilter() method. The value of the element must be a valid WKT string defining a polygon. An optional clip attribute can be set to "TRUE" to clip the geometries to the source region, otherwise the source geometries are not modified.
Field (optional, from GDAL 1.7.0): One or more attribute fields may be defined with Field elements. If no Field elements are defined, the fields of the source layer/sql will be defined on the VRT layer. The Field may have the following attributes:
- name (required): the name of the field.
- type: the field type, one of "Integer", "IntegerList", "Real", "RealList", "String", "StringList", "Binary", "Date", "Time", or "DateTime" - defaults to "String".
- width: the field width, defaults to unknown.
- precision: the field width, defaults to zero.
- src: the name of the source field to be copied to this one. By default defaults to the value of "name".

Example: ODBC Point Layer

In the following example (disease.ovf) the worms table from the ODBC database DISEASE is used to form a spatial layer. The virtual file uses the "x" and "y" columns to get the spatial location. It also marks the layer as a point layer, and as being in the WGS84 coordinate system.

<OGRVRTDataSource>

    <OGRVRTLayer name="worms">
        <SrcDataSource>ODBC:DISEASE,worms</SrcDataSource> 
 	<SrcLayer>worms</SrcLayer> 
	<GeometryType>wkbPoint</GeometryType> 
        <LayerSRS>WGS84</LayerSRS>
	<GeometryField encoding="PointFromColumns" x="x" y="y"/> 
    </OGRVRTLayer>

</OGRVRTDataSource>

Example: Renaming attributes

It can be usefull in some circumstances to be able to rename the field names from a source layer to other names. This is particularly true when you want to transcode to a format whose schema is fixed, such as GPX (<name>, <desc>, etc.). This can be accomplished using SQL this way:

<OGRVRTDataSource>
    <OGRVRTLayer name="remapped_layer">
        <SrcDataSource>your_source.shp</SrcDataSource>
        <SrcSQL>SELECT src_field_1 AS name, src_field_2 AS desc FROM your_source_layer_name</SrcSQL>
    </OGRVRTLayer>
</OGRVRTDataSource>

This can also be accomplished (from GDAL 1.7.0) using explicit field definitions:

<OGRVRTDataSource>
    <OGRVRTLayer name="remapped_layer">
        <SrcDataSource>your_source.shp</SrcDataSource>
        <SrcLayer>your_source</SrcSQL>
        <Field name="name" src="src_field_1" />
        <Field name="desc" src="src_field_2" type="String" width="45" />
    </OGRVRTLayer>
</OGRVRTDataSource>

Example: Transparent spatial filtering (GDAL >= 1.7.0)

The following example will only return features from the source layer that intersect the (0,40)-(10,50) region. Furthermore, returned geometries will be clipped to fit into that region.

<OGRVRTDataSource>
    <OGRVRTLayer name="source">
        <SrcDataSource>source.shp</SrcDataSource>
        <SrcRegion clip="true">POLYGON((0 40,10 40,10 50,0 50,0 40))</SrcRegion>
    </OGRVRTLayer>
</OGRVRTDataSource>

Other Notes

When the GeometryField is "WKT" spatial filtering is applied after extracting all rows from the source datasource. Essentially that means there is no fast spatial filtering on WKT derived geometries.
When the GeometryField is "PointFromColumns", and a SrcLayer (as opposed to SrcSQL) is used, and a spatial filter is in effect on the virtual layer then the spatial filter will be internally translated into an attribute filter on the X and Y columns in the SrcLayer. In cases where fast spatial filtering is important it can be helpful to index the X and Y columns in the source datastore, if that is possible. For instance if the source is an RDBMS. You can turn off that feature by setting the useSpatialSubquery attribute of the GeometryField element to FALSE.
Normally the SrcDataSource is a non-spatial tabular format (such as MySQL, SQLite, CSV, OCI, or ODBC) but it can also be a spatial database in which case the geometry can be directly copied over.