Usando Camadas Vetor

Esta seção lista várias operações que podem ser realizadas com camadas vetoriais.

Retrieving informations about attributes

You can retrieve informations about the fields associated with a vector layer by calling pendingFields() on a QgsVectorLayer instance:

# "layer" is a QgsVectorLayer instance
for field in layer.pendingFields():
    print, field.typeName()

Selecionando características

In QGIS desktop, features can be selected in different ways, the user can click on a feature, draw a rectangle on the map canvas or use an expression filter. Selected fatures are normally higlighted in a different color (default is yellow) to draw user’s attention on the selection. Sometimes can be useful to programmatically select features or to change the default color.

To change the selection color you can use setSelectionColor() method of QgsMapCanvas as shown in the following example:

iface.mapCanvas().setSelectionColor( QColor("red") )

To add add features to the selected features list for a given layer, you can call setSelectedFeatures() passing to it the list of features IDs:

# Get the active layer (must be a vector layer)
layer = iface.activeLayer()
# Get the first feature from the layer
feature = layer.getFeatures().next()
# Add this features to the selected list

To clear the selection, just pass an empty list:


Interagindo sobre camada vetor

Iterating over the features in a vector layer is one of the most common tasks. Below is an example of the simple basic code to perform this task and showing some information about each feature. the layer variable is assumed to have a QgsVectorLayer object

iter = layer.getFeatures()
for feature in iter:
    # retrieve every feature with its geometry and attributes
    # fetch geometry
    geom = feature.geometry()
    print "Feature ID %d: " %

    # show some information about the feature
    if geom.type() == QGis.Point:
        x = geom.asPoint()
        print "Point: " + str(x)
    elif geom.type() == QGis.Line:
        x = geom.asPolyline()
        print "Line: %d points" % len(x)
    elif geom.type() == QGis.Polygon:
        x = geom.asPolygon()
        numPts = 0
        for ring in x:
        numPts += len(ring)
        print "Polygon: %d rings with %d points" % (len(x), numPts)
        print "Unknown"

    # fetch attributes
    attrs = feature.attributes()

    # attrs is a list. It contains all the attribute values of this feature
    print attrs

Acessando atributos

Atributos podem ser refenciados pelo nome

print feature['name']

Alternatively, attributes can be referred to by index. This is will be a bit faster than using the name. For example, to get the first attribute:

print feature[0]

Iteração sobre os feições selecionadas

if you only need selected features, you can use the selectedFeatures() method from vector layer:

selection = layer.selectedFeatures()
print len(selection)
for feature in selection:
    # do whatever you need with the feature

Outra opção é o metódo de processamento features()

import processing
features = processing.features(layer)
for feature in features:
    # do whatever you need with the feature

By default, this will iterate over all the features in the layer, in case there is no selection, or over the selected features otherwise. Note that this behavior can be changed in the Processing options to ignore selections.

Iterando sobre um subconjunto de feições

Se você quer iterar sobre um conjunto de feições em uma camada, como por exemplo em uma determinada área, você precisa adicionar um objeto QgsFeatureRequest para a função getFeatures(). Segue um exemplo

request = QgsFeatureRequest()
for feature in layer.getFeatures(request):
    # do whatever you need with the feature

If you need an attribute-based filter instead (or in addition) of a spatial one like shown in the example above, you can build an QgsExpression object and pass it to the QgsFeatureRequest constructor. Here’s an example

# The expression will filter the features where the field "location_name" contains
# the word "Lake" (case insensitive)
exp = QgsExpression('location_name ILIKE \'%Lake%\'')
request = QgsFeatureRequest(exp)

The request can be used to define the data retrieved for each feature, so the iterator returns all features, but returns partial data for each of them.

# Only return selected fields
# More user friendly version
# Don't return geometry objects


If you only need a subset of the attributes or you don’t need the geometry informations, you can significantly increase the speed of the features request by using QgsFeatureRequest.NoGeometry flag or specifying a subset of attributes (possibly empty) like shown in the example above.

Modificando Camadas Vetoriais

Most vector data providers support editing of layer data. Sometimes they support just a subset of possible editing actions. Use the capabilities() function to find out what set of functionality is supported

caps = layer.dataProvider().capabilities()

By using any of the following methods for vector layer editing, the changes are directly committed to the underlying data store (a file, database etc). In case you would like to do only temporary changes, skip to the next section that explains how to do modifications with editing buffer.


If you are working inside QGIS (either from the console or from a plugin), it might be necessary to force a redraw of the map canvas in order to see the changes you’ve done to the geometry, to the style or to the attributes:

# If caching is enabled, a simple canvas refresh might not be sufficient
# to trigger a redraw and you must clear the cached image for the layer
if iface.mapCanvas().isCachingEnabled():

Adicionar feições

Create some QgsFeature instances and pass a list of them to provider’s addFeatures() method. It will return two values: result (true/false) and list of added features (their ID is set by the data store)

if caps & QgsVectorDataProvider.AddFeatures:
    feat = QgsFeature()
    feat.addAttribute(0, 'hello')
    feat.setGeometry(QgsGeometry.fromPoint(QgsPoint(123, 456)))
    (res, outFeats) = layer.dataProvider().addFeatures([feat])

Excluir feições

Para excluir algumas características, apenas providencie a lista com a identificação das características

if caps & QgsVectorDataProvider.DeleteFeatures:
    res = layer.dataProvider().deleteFeatures([5, 10])

Modificar Feições

It is possible to either change feature’s geometry or to change some attributes. The following example first changes values of attributes with index 0 and 1, then it changes the feature’s geometry

fid = 100   # ID of the feature we will modify

if caps & QgsVectorDataProvider.ChangeAttributeValues:
    attrs = { 0 : "hello", 1 : 123 }
    layer.dataProvider().changeAttributeValues({ fid : attrs })

if caps & QgsVectorDataProvider.ChangeGeometries:
    geom = QgsGeometry.fromPoint(QgsPoint(111,222))
    layer.dataProvider().changeGeometryValues({ fid : geom })


If you only need to change geometries, you might consider using the QgsVectorLayerEditUtils which provides some of useful methods to edit geometries (translate, insert or move vertex etc.)

Adicionando e Removendo Campos

To add fields (attributes), you need to specify a list of field definitions. For deletion of fields just provide a list of field indexes.

if caps & QgsVectorDataProvider.AddAttributes:
    res = layer.dataProvider().addAttributes([QgsField("mytext", QVariant.String), QgsField("myint", QVariant.Int)])

if caps & QgsVectorDataProvider.DeleteAttributes:
    res = layer.dataProvider().deleteAttributes([0])

After adding or removing fields in the data provider the layer’s fields need to be updated because the changes are not automatically propagated.


Modificando Camadas Vetoriais com um Buffer

Durante a edição de vetores com o QGIS, você precisa primeiramente colocar a camada alvo em modo de edição, então faça algumas modificações e finalmente envie (ou desfaça) as mudanças. Todas as alterações realizadas até antes do envio, permanecerão num buffer de edição em memória. É possível usar esta funcionalidade programaticamente, isso é apenas um outro métodos de edição de vetores que complementam o uso direto de provedores de dados. Use esta opção quando desenvolvendo alguma ferramenta de interface (GUI) para edição de camadas vetoriais, desde que você permita ao usuário decidir enviar/desfazer e desfazer/refazer. Quando enviar as alterações, toda edição no buffer de edição em memória será salvo em uma provedor de dados.

To find out whether a layer is in editing mode, use isEditing() — the editing functions work only when the editing mode is turned on. Usage of editing functions

# add two features (QgsFeature instances)
# delete a feature with specified ID

# set new geometry (QgsGeometry instance) for a feature
layer.changeGeometry(fid, geometry)
# update an attribute with given field index (int) to given value (QVariant)
layer.changeAttributeValue(fid, fieldIndex, value)

# add new field
layer.addAttribute(QgsField("mytext", QVariant.String))
# remove a field

In order to make undo/redo work properly, the above mentioned calls have to be wrapped into undo commands. (If you do not care about undo/redo and want to have the changes stored immediately, then you will have easier work by editing with data provider.) How to use the undo functionality

layer.beginEditCommand("Feature triangulation")

# ... call layer's editing methods ...

if problem_occurred:

# ... more editing ...


The beginEditCommand() will create an internal “active” command and will record subsequent changes in vector layer. With the call to endEditCommand() the command is pushed onto the undo stack and the user will be able to undo/redo it from GUI. In case something went wrong while doing the changes, the destroyEditCommand() method will remove the command and rollback all changes done while this command was active.

To start editing mode, there is startEditing() method, to stop editing there are commitChanges() and rollback() — however normally you should not need these methods and leave this functionality to be triggered by the user.

Using Spatial Index

Spatial indexes can dramatically improve the performance of your code if you need to do frequent queries to a vector layer. Imagine, for instance, that you are writing an interpolation algorithm, and that for a given location you need to know the 10 closest points from a points layer, in order to use those point for calculating the interpolated value. Without a spatial index, the only way for QGIS to find those 10 points is to compute the distance from each and every point to the specified location and then compare those distances. This can be a very time consuming task, especially if it needs to be repeated for several locations. If a spatial index exists for the layer, the operation is much more effective.

Think of a layer without a spatial index as a telephone book in which telephone numbers are not ordered or indexed. The only way to find the telephone number of a given person is to read from the beginning until you find it.

Spatial indexes are not created by default for a QGIS vector layer, but you can create them easily. This is what you have to do.

  1. criar um índice espacial — o seguinte código cria um índice vazio.

    index = QgsSpatialIndex()
  2. add features to index — index takes QgsFeature object and adds it to the internal data structure. You can create the object manually or use one from previous call to provider’s nextFeature()

  3. uma vez que o índice espacial é preenchido com alguns valores, você pode fazer algumas consultas

    # returns array of feature IDs of five nearest features
    nearest = index.nearestNeighbor(QgsPoint(25.4, 12.7), 5)
    # returns array of IDs of features which intersect the rectangle
    intersect = index.intersects(QgsRectangle(22.5, 15.3, 23.1, 17.2))

Writing Vector Layers

Você pode escrever um arquivo de camada vetorial utilizando a classe QgsVectorFileWriter. Ela atende a qualquer outro tipo de arquivo vetorial com suporte OGR (shapefiles, GeoJSON, KML e outros).

Há duas possibilidades de exportação de camadas vetoriais:

  • from an instance of QgsVectorLayer

    error = QgsVectorFileWriter.writeAsVectorFormat(layer, "my_shapes.shp", "CP1250", None, "ESRI Shapefile")
    if error == QgsVectorFileWriter.NoError:
        print "success!"
    error = QgsVectorFileWriter.writeAsVectorFormat(layer, "my_json.json", "utf-8", None, "GeoJSON")
    if error == QgsVectorFileWriter.NoError:
        print "success again!"

    The third parameter specifies output text encoding. Only some drivers need this for correct operation - shapefiles are one of those — however in case you are not using international characters you do not have to care much about the encoding. The fourth parameter that we left as None may specify destination CRS — if a valid instance of QgsCoordinateReferenceSystem is passed, the layer is transformed to that CRS.

    For valid driver names please consult the supported formats by OGR — you should pass the value in the “Code” column as the driver name. Optionally you can set whether to export only selected features, pass further driver-specific options for creation or tell the writer not to create attributes — look into the documentation for full syntax.

  • diretamente das características

    # define fields for feature attributes. A list of QgsField objects is needed
    fields = [QgsField("first", QVariant.Int),
              QgsField("second", QVariant.String)]
    # create an instance of vector file writer, which will create the vector file.
    # Arguments:
    # 1. path to new file (will fail if exists already)
    # 2. encoding of the attributes
    # 3. field map
    # 4. geometry type - from WKBTYPE enum
    # 5. layer's spatial reference (instance of
    #    QgsCoordinateReferenceSystem) - optional
    # 6. driver name for the output file
    writer = QgsVectorFileWriter("my_shapes.shp", "CP1250", fields, QGis.WKBPoint, None, "ESRI Shapefile")
    if writer.hasError() != QgsVectorFileWriter.NoError:
        print "Error when creating shapefile: ", writer.hasError()
    # add a feature
    fet = QgsFeature()
    fet.setAttributes([1, "text"])
    # delete the writer to flush features to disk (optional)
    del writer

provedor de Memória

O Gerenciador de memória foi desenvolvido para ser utilizado principalmente por plugins ou aplicativos desenvolvidos por terceiros. Estes não são armazenados no disco, permitindo aos desenvolvedores utilizá-los como um rápido backend para algumas camadas temporárias.

The provider supports string, int and double fields.

The memory provider also supports spatial indexing, which is enabled by calling the provider’s createSpatialIndex() function. Once the spatial index is created you will be able to iterate over features within smaller regions faster (since it’s not necessary to traverse all the features, only those in specified rectangle).

A memory provider is created by passing "memory" as the provider string to the QgsVectorLayer constructor.

The constructor also takes a URI defining the geometry type of the layer, one of: "Point", "LineString", "Polygon", "MultiPoint", "MultiLineString", or "MultiPolygon".

The URI can also specify the coordinate reference system, fields, and indexing of the memory provider in the URI. The syntax is:

Specifies the coordinate reference system, where definition may be any of the forms accepted by QgsCoordinateReferenceSystem.createFromString()

Especifica que o provedor irá usar o index espacial

Specifies an attribute of the layer. The attribute has a name, and optionally a type (integer, double, or string), length, and precision. There may be multiple field definitions.

O exemplo seguinte de URL incorpora todas estas opções


The following example code illustrates creating and populating a memory provider

# create layer
vl = QgsVectorLayer("Point", "temporary_points", "memory")
pr = vl.dataProvider()

# add fields
pr.addAttributes([QgsField("name", QVariant.String),
                    QgsField("age",  QVariant.Int),
                    QgsField("size", QVariant.Double)])
vl.updateFields() # tell the vector layer to fetch changes from the provider

# add a feature
fet = QgsFeature()
fet.setAttributes(["Johny", 2, 0.3])

# update layer's extent when new features have been added
# because change of extent in provider is not propagated to the layer

Finally, let’s check whether everything went well

# show some stats
print "fields:", len(pr.fields())
print "features:", pr.featureCount()
e = layer.extent()
print "extent:", e.xMiniminum(), e.yMinimum(), e.xMaximum(), e.yMaximum()

# iterate over features
f = QgsFeature()
features = vl.getFeatures()
for f in features:
    print "F:",, f.attributes(), f.geometry().asPoint()

Aparencia (Simbologia) de Camadas de Vetor

When a vector layer is being rendered, the appearance of the data is given by renderer and symbols associated with the layer. Symbols are classes which take care of drawing of visual representation of features, while renderers determine what symbol will be used for a particular feature.

A renderização para uma dada camada pode ser obtida como mostrada abaixo

renderer = layer.rendererV2()

And with that reference, let us explore it a bit

print "Type:", rendererV2.type()

Existem muitos tipos de renderização conhecidas disponíveis na biblioteca principal do QGIS




singleSymbol QgsSingleSymbolRendererV2

Renderiza todas as características com o mesmo símbolo

categorizedSymbol QgsCategorizedSymbolRendererV2

Renderiza características usando um símbolo diferente para cada categoria

graduatedSymbol QgsGraduatedSymbolRendererV2

Renderiza caracter´sticas usando diferents símbolos para cada limite de valores

There might be also some custom renderer types, so never make an assumption there are just these types. You can query QgsRendererV2Registry singleton to find out currently available renderers:

# Prints:

It is possible to obtain a dump of a renderer contents in text form — can be useful for debugging

print rendererV2.dump()

Single Symbol Renderer

You can get the symbol used for rendering by calling symbol() method and change it with setSymbol() method (note for C++ devs: the renderer takes ownership of the symbol.)

You can change the symbol used by a particular vector layer by calling setSymbol() passing an instance of the appropriate symbol instance. Symbols for point, line and polygon layers can be created by calling the createSimple() function of the corresponding classes QgsMarkerSymbolV2, QgsLineSymbolV2 and QgsFillSymbolV2.

The dictionary passed to createSimple() sets the style properties of the symbol.

For example you can change the symbol used by a particular point layer by calling setSymbol() passing an instance of a QgsMarkerSymbolV2 as in the following code example:

symbol = QgsMarkerSymbolV2.createSimple({'name': 'square', 'color': 'red'})

name indicates the shape of the marker, and can be any of the following:

  • circulo

  • quadrado

  • retângulo`

  • diamante

  • pentágono

  • triângulo

  • triângulo_equilateral

  • estrela

  • regular_star
  • flecha

  • filled_arrowhead

Categorized Symbol Renderer

You can query and set attribute name which is used for classification: use classAttribute() and setClassAttribute() methods.

To get a list of categories

for cat in rendererV2.categories():
    print "%s: %s :: %s" % (cat.value().toString(), cat.label(), str(cat.symbol()))

Where value() is the value used for discrimination between categories, label() is a text used for category description and symbol() method returns assigned symbol.

The renderer usually stores also original symbol and color ramp which were used for the classification: sourceColorRamp() and sourceSymbol() methods.

Graduated Symbol Renderer

This renderer is very similar to the categorized symbol renderer described above, but instead of one attribute value per class it works with ranges of values and thus can be used only with numerical attributes.

To find out more about ranges used in the renderer

for ran in rendererV2.ranges():
    print "%f - %f: %s %s" % (

you can again use classAttribute() to find out classification attribute name, sourceSymbol() and sourceColorRamp() methods. Additionally there is mode() method which determines how the ranges were created: using equal intervals, quantiles or some other method.

If you wish to create your own graduated symbol renderer you can do so as illustrated in the example snippet below (which creates a simple two class arrangement)

from qgis.core import *

myVectorLayer = QgsVectorLayer(myVectorPath, myName, 'ogr')
myTargetField = 'target_field'
myRangeList = []
myOpacity = 1
# Make our first symbol and range...
myMin = 0.0
myMax = 50.0
myLabel = 'Group 1'
myColour = QtGui.QColor('#ffee00')
mySymbol1 = QgsSymbolV2.defaultSymbol(myVectorLayer.geometryType())
myRange1 = QgsRendererRangeV2(myMin, myMax, mySymbol1, myLabel)
#now make another symbol and range...
myMin = 50.1
myMax = 100
myLabel = 'Group 2'
myColour = QtGui.QColor('#00eeff')
mySymbol2 = QgsSymbolV2.defaultSymbol(
myRange2 = QgsRendererRangeV2(myMin, myMax, mySymbol2 myLabel)
myRenderer = QgsGraduatedSymbolRendererV2('', myRangeList)


Trabalhando com Símbolos

Para representação de símbolos, existe QgsSymbolV2 que é a classe básica com três classes derivadas.

  • QgsMarkerSymbolV2 — for point features
  • QgsLineSymbolV2 — for line features
  • QgsFillSymbolV2 — for polygon features

Every symbol consists of one or more symbol layers (classes derived from QgsSymbolLayerV2). The symbol layers do the actual rendering, the symbol class itself serves only as a container for the symbol layers.

Having an instance of a symbol (e.g. from a renderer), it is possible to explore it: type() method says whether it is a marker, line or fill symbol. There is a dump() method which returns a brief description of the symbol. To get a list of symbol layers

for i in xrange(symbol.symbolLayerCount()):
    lyr = symbol.symbolLayer(i)
    print "%d: %s" % (i, lyr.layerType())

To find out symbol’s color use color() method and setColor() to change its color. With marker symbols additionally you can query for the symbol size and rotation with size() and angle() methods, for line symbols there is width() method returning line width.

Por padrão, tamanho e largura são em milimetros e ângulos em graus.

Working with Symbol Layers

As said before, symbol layers (subclasses of QgsSymbolLayerV2) determine the appearance of the features. There are several basic symbol layer classes for general use. It is possible to implement new symbol layer types and thus arbitrarily customize how features will be rendered. The layerType() method uniquely identifies the symbol layer class — the basic and default ones are SimpleMarker, SimpleLine and SimpleFill symbol layers types.

You can get a complete list of the types of symbol layers you can create for a given symbol layer class like this

from qgis.core import QgsSymbolLayerV2Registry
myRegistry = QgsSymbolLayerV2Registry.instance()
myMetadata = myRegistry.symbolLayerMetadata("SimpleFill")
for item in myRegistry.symbolLayersForType(QgsSymbolV2.Marker):
    print item



A classe QgsSymbolLayerV2Registry gerencia um banco de dados com todos os tipo de símbolo de camadas.

To access symbol layer data, use its properties() method that returns a key-value dictionary of properties which determine the appearance. Each symbol layer type has a specific set of properties that it uses. Additionally, there are generic methods color(), size(), angle(), width() with their setter counterparts. Of course size and angle is available only for marker symbol layers and width for line symbol layers.

Creating Custom Symbol Layer Types

Imagine you would like to customize the way how the data gets rendered. You can create your own symbol layer class that will draw the features exactly as you wish. Here is an example of a marker that draws red circles with specified radius

class FooSymbolLayer(QgsMarkerSymbolLayerV2):

  def __init__(self, radius=4.0):
      self.radius = radius
      self.color = QColor(255,0,0)

  def layerType(self):
     return "FooMarker"

  def properties(self):
      return { "radius" : str(self.radius) }

  def startRender(self, context):

  def stopRender(self, context):

  def renderPoint(self, point, context):
      # Rendering depends on whether the symbol is selected (QGIS >= 1.5)
      color = context.selectionColor() if context.selected() else self.color
      p = context.renderContext().painter()
      p.drawEllipse(point, self.radius, self.radius)

  def clone(self):
      return FooSymbolLayer(self.radius)

The layerType() method determines the name of the symbol layer, it has to be unique among all symbol layers. Properties are used for persistence of attributes. clone() method must return a copy of the symbol layer with all attributes being exactly the same. Finally there are rendering methods: startRender() is called before rendering first feature, stopRender() when rendering is done. And renderPoint() method which does the rendering. The coordinates of the point(s) are already transformed to the output coordinates.

For polylines and polygons the only difference would be in the rendering method: you would use renderPolyline() which receives a list of lines, resp. renderPolygon() which receives list of points on outer ring as a first parameter and a list of inner rings (or None) as a second parameter.

Usually it is convenient to add a GUI for setting attributes of the symbol layer type to allow users to customize the appearance: in case of our example above we can let user set circle radius. The following code implements such widget

class FooSymbolLayerWidget(QgsSymbolLayerV2Widget):
    def __init__(self, parent=None):
        QgsSymbolLayerV2Widget.__init__(self, parent)

        self.layer = None

        # setup a simple UI
        self.label = QLabel("Radius:")
        self.spinRadius = QDoubleSpinBox()
        self.hbox = QHBoxLayout()
        self.connect(self.spinRadius, SIGNAL("valueChanged(double)"), \

    def setSymbolLayer(self, layer):
        if layer.layerType() != "FooMarker":
        self.layer = layer

    def symbolLayer(self):
        return self.layer

    def radiusChanged(self, value):
        self.layer.radius = value

This widget can be embedded into the symbol properties dialog. When the symbol layer type is selected in symbol properties dialog, it creates an instance of the symbol layer and an instance of the symbol layer widget. Then it calls setSymbolLayer() method to assign the symbol layer to the widget. In that method the widget should update the UI to reflect the attributes of the symbol layer. symbolLayer() function is used to retrieve the symbol layer again by the properties dialog to use it for the symbol.

On every change of attributes, the widget should emit changed() signal to let the properties dialog update the symbol preview.

Now we are missing only the final glue: to make QGIS aware of these new classes. This is done by adding the symbol layer to registry. It is possible to use the symbol layer also without adding it to the registry, but some functionality will not work: e.g. loading of project files with the custom symbol layers or inability to edit the layer’s attributes in GUI.

We will have to create metadata for the symbol layer

class FooSymbolLayerMetadata(QgsSymbolLayerV2AbstractMetadata):

  def __init__(self):
    QgsSymbolLayerV2AbstractMetadata.__init__(self, "FooMarker", QgsSymbolV2.Marker)

  def createSymbolLayer(self, props):
    radius = float(props[QString("radius")]) if QString("radius") in props else 4.0
    return FooSymbolLayer(radius)

  def createSymbolLayerWidget(self):
    return FooSymbolLayerWidget()


You should pass layer type (the same as returned by the layer) and symbol type (marker/line/fill) to the constructor of parent class. createSymbolLayer() takes care of creating an instance of symbol layer with attributes specified in the props dictionary. (Beware, the keys are QString instances, not “str” objects). And there is createSymbolLayerWidget() method which returns settings widget for this symbol layer type.

O último passo para adicionar este símbolo de camada para o registro — e estamos prontos.

Creating Custom Renderers

It might be useful to create a new renderer implementation if you would like to customize the rules how to select symbols for rendering of features. Some use cases where you would want to do it: symbol is determined from a combination of fields, size of symbols changes depending on current scale etc.

The following code shows a simple custom renderer that creates two marker symbols and chooses randomly one of them for every feature

import random

class RandomRenderer(QgsFeatureRendererV2):
  def __init__(self, syms=None):
    QgsFeatureRendererV2.__init__(self, "RandomRenderer")
    self.syms = syms if syms else [QgsSymbolV2.defaultSymbol(QGis.Point), QgsSymbolV2.defaultSymbol(QGis.Point)]

  def symbolForFeature(self, feature):
    return random.choice(self.syms)

  def startRender(self, context, vlayer):
    for s in self.syms:

  def stopRender(self, context):
    for s in self.syms:

  def usedAttributes(self):
    return []

  def clone(self):
    return RandomRenderer(self.syms)

The constructor of parent QgsFeatureRendererV2 class needs renderer name (has to be unique among renderers). symbolForFeature() method is the one that decides what symbol will be used for a particular feature. startRender() and stopRender() take care of initialization/finalization of symbol rendering. usedAttributes() method can return a list of field names that renderer expects to be present. Finally clone() function should return a copy of the renderer.

Like with symbol layers, it is possible to attach a GUI for configuration of the renderer. It has to be derived from QgsRendererV2Widget. The following sample code creates a button that allows user to set symbol of the first symbol

class RandomRendererWidget(QgsRendererV2Widget):
  def __init__(self, layer, style, renderer):
    QgsRendererV2Widget.__init__(self, layer, style)
    if renderer is None or renderer.type() != "RandomRenderer":
      self.r = RandomRenderer()
      self.r = renderer
    # setup UI
    self.btn1 = QgsColorButtonV2("Color 1")
    self.vbox = QVBoxLayout()
    self.connect(self.btn1, SIGNAL("clicked()"), self.setColor1)

  def setColor1(self):
    color = QColorDialog.getColor(self.r.syms[0].color(), self)
    if not color.isValid(): return

  def renderer(self):
    return self.r

The constructor receives instances of the active layer (QgsVectorLayer), the global style (QgsStyleV2) and current renderer. If there is no renderer or the renderer has different type, it will be replaced with our new renderer, otherwise we will use the current renderer (which has already the type we need). The widget contents should be updated to show current state of the renderer. When the renderer dialog is accepted, widget’s renderer() method is called to get the current renderer — it will be assigned to the layer.

The last missing bit is the renderer metadata and registration in registry, otherwise loading of layers with the renderer will not work and user will not be able to select it from the list of renderers. Let us finish our RandomRenderer example

class RandomRendererMetadata(QgsRendererV2AbstractMetadata):
  def __init__(self):
    QgsRendererV2AbstractMetadata.__init__(self, "RandomRenderer", "Random renderer")

  def createRenderer(self, element):
    return RandomRenderer()
  def createRendererWidget(self, layer, style, renderer):
    return RandomRendererWidget(layer, style, renderer)


Similarly as with symbol layers, abstract metadata constructor awaits renderer name, name visible for users and optionally name of renderer’s icon. createRenderer() method passes QDomElement instance that can be used to restore renderer’s state from DOM tree. createRendererWidget() method creates the configuration widget. It does not have to be present or can return None if the renderer does not come with GUI.

To associate an icon with the renderer you can assign it in QgsRendererV2AbstractMetadata constructor as a third (optional) argument — the base class constructor in the RandomRendererMetadata __init__() function becomes

       "Random renderer",
       QIcon(QPixmap("RandomRendererIcon.png", "png")))

The icon can be associated also at any later time using setIcon() method of the metadata class. The icon can be loaded from a file (as shown above) or can be loaded from a Qt resource (PyQt4 includes .qrc compiler for Python).

Further Topics

creating/modifying symbols working with style (QgsStyleV2) working with color ramps (QgsVectorColorRampV2) rule-based renderer (see this blogpost) exploring symbol layer and renderer registries