Configurando aplicativos externos

The processign framework can be extended using additional applications. Currently, SAGA, GRASS, OTB(Orfeo Toolbox) and R are supported, along with some other command-line applications that provide spatial data analysis functionalities. Algorithms relying on an external application are managed by their own algorithm provider.

This chapter will show you how to configure the processing framework to include these additional applications, and will explain some particular features of the algorithm based on them. Once you have correctly configured the system, you will be able to execute external algorithms from any component like the toolbox or the graphical modeler, just like you do with any other geoalgorithm.

By default, all algorithms that rely on an external appplication not shipped with QGIS are not enabled. You can enable them in the configuration dialog. Make sure that the corresponding application is already installed in your system. Enabling an algorithm provider without installing the application it needs will cause the algorithms to appear in the toolbox, but an error will be thrown when you try to execute them.

This is because the algorithm descriptions (needed to create the parameters dialog and provide the information needed about the algorithm) are not included with each application, but with QGIS instead. That is,they are part of QGIS, so you have them in your installation even if you have not installed any other software. Running the algorithm, however, needs the application binaries to be installed in your system.

A note for Windows users

If you are not an advanced user and you are running QGIS on windows, you might not be interested in reading the rest of this chapter. Make sure you install QGIS in your system using the OSGeo4W application. That will automatically install SAGA, GRASS and OTB in your system, and configure them so they can be run from QGIS. All the algorithms in the simplified view of the toolbox will be ready to be run, without needing any further configuration.

If you want to know more about how these providers work, or want to use some algorithms not included in the simplified toolbox (such as R scripts), keep on reading.

Notas sobre os formatos de arquivos

When using an external software, opening a file in QGIS does not mean that it can be opened and processed as well on that other software. In most cases, it can read what you have opened in QGIS, but in some cases, that might not be the case. When using databases or uncommon file formats, whether for raster of vector layers, problems might arise. If that happens, try to use well known file formats that you are sure that are understood by both programs, and check to console output (in the history and log dialog) for knowing more about what is going wrong.

Using GRASS raster layers is, for instance, one case in which you might have trouble and not be able to complete your work if you call an external algorithm using such a layer as input. For this reason, these layers will not appear as available to algorithms.

You should, however, find no problems at all with vector layers, since QGIS automatically converts from the original file format to one accepted by the external application before passing the layer to it. This adds an extra processing time, which might be significant if the layer has a large size, so do not be surprised if it takes more to process a layer from a DB connection that one of a similar size stored in a shapefile.

Providers not using external applications can process any layer that you can open in QGIS, since they open it for analysis through QGIS.

Regarding output formats, all formats supported by QGIS as output can be used, both for raster and vector layers. Some provider do not support certain formats, but all can export to common formats raster layers that can be later transformed by QGIS automatically. As in the case of input layers, if this conversion is needed, that might increase the processing time.

If the extension of the filename specified when calling an algorithm does not match the extension of any of the formats supported by QGIS, then a suffix will be added to set a default format. In the case of raster layers, the tif extension is used, while shp is used for vector layer.

Notas sobre seleção de camadas vetoriais

External applications are also aware of the selection that exist in vector layers within QGIS. However, that requires rewritting all input vector layers, just as if they were originally in a format not supported by the external application. Only when no selection exist, or the Use only selected features option is not enabled in the processing general configuration, a layer can be directly passed to an external application.

In other cases, exporting only selected features is needed, which causes execution times to be longer.


SAGA algorithms can be run from QGIS if you have SAGA installed in your system and you configure the processing framework properly so it can find SAGA executables. In particular, the SAGA command-line executable is needed to run SAGA algorithms.

In case of running Windows, the standalone installer or the OSGeo4W installer, both install SAGA along with QGIS, and the path is automatically configured, so there is no need to do anything else.

If you have installed SAGA yourself (remember, you need version 2.1), the path to the SAGA executable must be configured. To do it, open the configuration dialog. In the SAGA block you will find a setting named SAGA Folder. Enter the path to the folder where SAGA is installed. Close the configuration dialog and now you are ready to run SAGA algorithms from QGIS.

In case you are running linux, SAGA binaries are not included with SEXTANTE, so you have to download and install the software yourself. Please check the SAGA website for more information. SAGA 2.1 is needed.

In this case there is no need to configure that, and you will not see those folders. Instead, you must make sure that SAGA is properly installed and its folder is added to the PATH environment variable. Just open a console and type saga_cmd to check that the system can find where SAGA binaries are located.

Sobre as limitações do sistema SAGA

Most of SAGA algorithms that require several input raster layers, require them to have the same grid system. That is, to cover the same geographic area and have the same cellsize, so their corresponding grids match. When calling SAGA algorithms from QGIS, you can use any layer, regardless of its cellsize and extent. When multiple raster layers are used as input for a SAGA algorithm, QGIS resamples them to a common grid system and then passes them to SAGA (unless the SAGA algorithm can operate with layers from different grid systems).

A definição do sistema de projeção comum é controlado pelo usuário, você vai encontrar vários parâmetros no grupo SAGA da janela de configuração para definí-lo. Existem duas maneiras de definir o sistema de grid de destino:

  • Setting it manually. You define the extent setting the values of the following parameters:

    • Resampling min X
    • Resampling max X
    • Resampling min Y
    • Resampling max Y
    • Resampling cellsize

    Notice that QGIS will resample input layers to that extent, even if they do not overlap with it.

  • Setting it automatically from input layers. To select this option, just check the Use min covering grid system for resampling option. All the other settings will be ignored and the minimum extent that covers all the input layers will be used. The cellsize of the target layer is the maximum of all cellsizes of the input layers.

Para algoritmos que não usam múltiplas camadas do tipo raster, ou para aquelas camadas que não precisam de um sistema de projeção único de entrada, esses parametros serão executados sem reprojeção antes de chamar o SAGA.

Limitações para multiplas bandas de camadas

Unlike QGIS, SAGA has no support for multi-band layers. If you want to use a multiband layer (such as an RGB or multispectral image), you first have to split it into single-banded images. To do so, you can use the ‘SAGA/Grid - Tools/Split RGB image’ algorithm (which creates 3 images from an RGB image) or the ‘SAGA/Grid - Tools/Extract band’ algorithm (to extract a single band).

Limitações na resolução espacial (cellsize)

SAGA assumes that raster layers have the same cellsize in the X and Y axis. If you are working with a layer with different values for its horizontal and vertical cellsizes, you might get unexcepted results. In this case, a warning will be added to the processing log, indicating that an input layer might not be suitable to be processed by SAGA.


When QGIS calls SAGA, it does it using its command-line interface, thus passing a set of commands to perform all the required operation. SAGA show its progress by writing information to the console, which includes the percentage of processing already done, along with additional content. This output is filtered and used to update the progress bar while the algorithm is running.

Both the commands sent by QGIS and the additional information printed by SAGA can be logged along with other processing log messages, and you might find them useful to track in detailed what is going on when QGIS runs a SAGA algorithm. you will find two settings, namely Log console output and Log execution commands to activate that logging mechanism.

Most other providers that use an external application and call it through the command-line have similar options, so you will find them as well in other places in the processing settings list.

R, Criando scripts em R

R integration in QGIS is different from that of SAGA in that there is not a predefined set of algorithms you can run (except for a few examples). Instead, you should write your scripts and call R commands, much like you would do from R, and in a very similar manner to what we saw in the chapter dedicated to processing scripts. This chapter shows you the syntax to use to call those R commands from QGIS and how to use QGIS objects (layers, tables) in them.

The first thing you have to do, as we saw in the case of SAGA, is to tell QGIS where you R binaries are located. You can do so using the R folder entry in the processing configuration dialog. Once you have set that parameter, you can start creating your own R scripts and executing them.

Mais uma vez, esta configuração é diferente no Linux, e você só tem que ter certeza de que a pasta R está incluído no ambiente PATH. Para isso você pode começar apenas digitando R no console, e assim R está pronto para uso.

To add a new algorithm that calls an R function (or a more complex R script that you have developed and you would like to have available from QGIS), you have to create a script file that tells the processing framework how to perform that operation and the corresponding R commands to do so.

Script files have the extension .rsx and creating them is pretty easy if you just have a basic knowledge of R syntax and R scripting. They should be stored in the R-scripts folder. You can set this folder in the R settings group (available from the processing settings dialog), just like you do with the folder for regular processing scripts.

Let’s have a look at a very simple file script file, which calls the R method spsample to create a random grid within the boundary of the polygons in a given polygon layer. This method belong to the maptools package. Since almost all the algorithms that you might like to incorporate into QGIS will use or generate spatial data, knowledge of spatial packages like maptools and, specially, sp, is mandatory.

##numpoints=number 10
##output=output vector

The first lines, which start with a double Python comment sign (##), tell QGIS the inputs of the algorithm described in the file and the outputs that it will generate. They work exactly with the same syntax as the SEXTANTE scripts that we have already seen, so they will not be described here again. Check the processing_scripts section for more information.

When you declare an input parameter, QGIS uses that information for two things: creating the user interface to ask the user for the value of that parameter and creating a corresponding R variable that can be later used as input for R commands.

In the above example, we are declaring an input of type vector named polyg. When executing the algorithm, QGIS will open in R the layer selected by the user and store it in a variable also named polyg. So the name of a parameter is also the name of the variable that we can use in R for accesing the value of that parameter (thus, you should avoid using reserved R words as parameter names).

Spatial elements such as vector and raster layers are read using the readOGR() and brick() commands (you do not have to worry about adding those commands to your description file, QGIS will do it) and stored as Spatial*DataFrame objects. Table fields are stored as strings containing the name of the selected field.

Tabelas são abertas usando o comando read.csv(). Se uma tabela inserida pelo usuário não está no formato CSV, será convertida antes de importá-la para R.

Additionally, raster files can be read using the readGDAL() command instead of brick(), by using the ##usereadgdal.

If you are an advanced user and do not want QGIS to create the object representing the layer, you can use the ##passfilename tag to indicate that you prefer a string with the filename instead. In this case, it is up to you to open the file before performing any operation on the data it contains.

With the above information, we can now understand the first line of our first example script (the first line not starting with a Python comment).


A variável polygon já contém um objeto SpatialPolygonsDataFrame, para que possa ser usado para chamar o método spsample, assim como um numpoints, o que indica o número de pontos a serem adicionados para a grade de amostra criada.

Como declaramos uma saída vetorial, chamada out, temos que criar uma variável chamada out e criar nela um grupo de objetos Spatial*DataFrame` (neste caso, um ``SpatialPointsDataFrame). Você pode usar qualquer nome para as variáveis ​​intermediárias. Apenas certifique-se que a variável que armazena o resultado final tem o mesmo nome que você usou para declará-la, e contém um valor adequado.

In this case, the result obtained from the spsample method has to be converted explicitly into a SpatialPointsDataFrame object, since it is itself an object of class ppp, which is not a suitable class to be returned to QGIS.

If your algorithm generates raster layers, the way they are saved will depend on whether you have used or not the #dontuserasterpackage option. In you have used it, layers are saved using the writeGDAL() method. If not, the writeRaster() method from the raster package will be used.

If you have used the #passfilename option, outputs are generated using the raster package (with writeRaster()), even though it is not used for the inputs.

If you algorithm does not generate any layer, but a text result in the console instead, you have to indicate that you want the console to be shown once the execution is finished. To do so, just start the command lines that produce the results you want to print with the > (‘greater’) sign. The output of all other lines will not be shown. For instance, here is the description file of an algorithm that performs a normality test on a given field (column) of the attributes of a vector layer:

##field=field layer

The output ot the last line is printed, but the output of the first is not (and neither are the outputs from other command lines added automatically by QGIS).

Se o seu algoritmo cria qualquer tipo de gráfico (usando o método plot(), adicione a seguinte linha:


This will cause QGIS to redirect all R graphical outputs to a temporary file, which will be later opened once R execution has finished.

Both graphics and console results will be shown in the processing results manager.

Para mais informações, consulte os arquivos de script fornecidos com o SEXTANTE. A maioria deles são bastante simples e vai lhe ajudar muito a entender como criar as seus próprios scripts.


rgdal and maptools libraries are loaded by default so you do not have to add the corresponding library() commands (you have to make sure, however, that those two packages are installed in your R distribution). However, other additional libraries that you might need have to be explicitly loaded. Just add the necessary commands at the beginning of your script. You also have to make sure that the corresponding packages are installed in the R distribution used by QGIS. The processing framework will not take care of any package installation. If you run a script that requires an uninstalled package, the execution will fail, and SEXTANTE will try to detect which packages are missing. You must install those missing libraries manually before you can run the algorithm.


Configuring GRASS is not much different from configuring SAGA. First, the path to the GRASS folder has to be defined, but only if you are running Windows. Additionaly, a shell interpreter (usually msys.exe, which can be found in most GRASS for Windows distributions) has to be defined and its path set up as well.

By default, the processign framework tries to configure its GRASS connector to use the GRASS distribution that ships along with QGIS. This should work without problems in most systems, but if you experience problems, you might have to do it manually. Also, if you want to use a different GRASS installation, you can change that setting and point to the folder where that it is installed. GRASS 6.4 is needed for algorithms to work correctly.

Se você estiver executando o QGIS no Linux, só tem que se certificar de que o GRASS está instalado corretamente, e que ele pode ser executado sem problemas a partir de um console.

GRASS algorithms use a region for calculations. This region can be defined manually using values similar to the ones found in the SAGA configuration, or automatically, taking the minimum extent that covers all the input layers used to execute the algorithm each time. If this is the behaviour you prefer, just check the Use min covering region option in the GRASS configuration parameters.

The last parameter that has to be configured is related to the mapset. A mapset is needed to run GRASS, and the processing frmaework creates a temporary one for each execution. You have to specify if the data you are working with uses geographical (lat/lon) coordinates or projected ones.


No additional configuration is needed to run GDAL algorithms, since it is already incorporated to QGIS and algorithms can infere its configuration from it.

Orfeo ToolBox

Orfeo ToolBox (OTB) algorithms can be run from QGIS if you have OTB installed in your system and you have configured QGIS properly, so it can find all necessary files (command-line tools and libraries).

As in the case of SAGA OTB binaries are included in the standalone installer for Windows, but are not included if you are runing Linux, so you have to download and install the software yourself. Please check the OTB website for more information.

Once OTB is installed, start QGIS, open the processing configuration dialog and configure the OTB algorithm provider. In the Orfeo Toolbox (image analysis) block you will find all settings related to OTB. First ensure that algorithms are enabled.

Then configure the path to the folder where OTB command–line tools and libraries are installed:

  • nix usually OTB applications folder point to /usr/lib/otb/applications and OTB command line tools folder is /usr/bin
  • win if you use OSGeo4W installer, than install otb-bin package and enter C:\OSGeo4W\apps\orfeotoolbox\applications as OTB applications folder and C:\OSGeo4W\bin as OTB command line tools folder. This values should be configured by default, but if you have a different OTB installation, configure them to the correspondig values in your system.


To use this provider you need to install TauDEM command line tools.


Please visit TauDEM homepage for installation instructions and precompiled binaries for 32bit and 64bit systems. IMPORTANT: you need TauDEM 5.0.6 executables, version 5.2 is currently not supported.


Não existem pacotes para a distribuição Linux, então você deve compilar o TauDEM manualmente. Como TauDEM usa MPICH2, primeiro tem que instalá-lo usando algum gerenciador de pacotes de aplicativos. TauDEM também funciona bem com OpenMPI, assim é possível usá-lo em vez de MPICH2.

Download TauDEM 5.0.6 source code and extract files in some folder.

Open linearpart.h file and add after line

#include "mpi.h"

adicione uma nova linha com

#include <stdint.h>

para obter sucesso

#include "mpi.h"
#include <stdlib.h>

Save changes and close file. Now open tiffIO.h, find line #include "stdint.h" and replace quotes ("") with <>, so you’ll get

#include <stdint.h>

Salve as alterações e feche o arquivo. Crie um diretório cd de armazenamento para ele

mkdir build
cd build

Configure a sua compilação com o comando

CXX=mpicxx cmake -DCMAKE_INSTALL_PREFIX=/usr/local ..

e então compile


Finalmente, instale o TauDEM em /usr/local/bin, execute

sudo make install