Introduction¶

PYQGIS_STARTUP environment variable¶

You can run Python code just before QGIS initialization completes by setting the PYQGIS_STARTUP environment variable to the path of an existing Python file.

This method is something you will probably rarely need, but worth mentioning here because it is one of the several ways to run Python code within QGIS and because this code will run before QGIS initialization is complete. This method is very useful for cleaning sys.path, which may have undesireable paths, or for isolating/loading the initial environ without requiring a virt env, e.g. homebrew or MacPorts installs on Mac.

The startup.py file¶

Every time QGIS starts, the user’s Python home directory (usually: .qgis2/python) is searched for a file named startup.py, if that file exists, it is executed by the embedded Python interpreter.

Using PyQGIS in custom application¶

Note: do not use qgis.py as a name for your test script — Python will not be able to import the bindings as the script’s name will shadow them.

First of all you have to import qgis module, set QGIS path where to search for resources — database of projections, providers etc. When you set prefix path with second argument set as True, QGIS will initialize all paths with standard dir under the prefix directory. Calling initQgis() function is important to let QGIS search for the available providers.

from qgis.core import *

# supply path to where is your qgis installed
QgsApplication.setPrefixPath("/path/to/qgis/installation", True)

# load providers
QgsApplication.initQgis()

Now you can work with QGIS API — load layers and do some processing or fire up a GUI with a map canvas. The possibilities are endless :-)

When you are done with using QGIS library, call exitQgis() to make sure that everything is cleaned up (e.g. clear map layer registry and delete layers):

QgsApplication.exitQgis()

Running Custom Applications¶

You will need to tell your system where to search for QGIS libraries and appropriate Python modules if they are not in a well-known location — otherwise Python will complain:

>>> import qgis.core
ImportError: No module named qgis.core

This can be fixed by setting the PYTHONPATH environment variable. In the following commands, qgispath should be replaced with your actual QGIS installation path:

• on Linux: export PYTHONPATH=/qgispath/share/qgis/python
• on Windows: set PYTHONPATH=c:\qgispath\python

The path to the PyQGIS modules is now known, however they depend on qgis_core and qgis_gui libraries (the Python modules serve only as wrappers). Path to these libraries is typically unknown for the operating system, so you get an import error again (the message might vary depending on the system):

>>> import qgis.core
ImportError: libqgis_core.so.1.5.0: cannot open shared object file: No such file or directory

Fix this by adding the directories where the QGIS libraries reside to search path of the dynamic linker:

• on Linux: export LD_LIBRARY_PATH=/qgispath/lib
• on Windows: set PATH=C:\qgispath;%PATH%

These commands can be put into a bootstrap script that will take care of the startup. When deploying custom applications using PyQGIS, there are usually two possibilities:

• require user to install QGIS on his platform prior to installing your application. The application installer should look for default locations of QGIS libraries and allow user to set the path if not found. This approach has the advantage of being simpler, however it requires user to do more steps.
• package QGIS together with your application. Releasing the application may be more challenging and the package will be larger, but the user will be saved from the burden of downloading and installing additional pieces of software.

The two deployment models can be mixed - deploy standalone application on Windows and Mac OS X, for Linux leave the installation of QGIS up to user and his package manager.