Formatting of attribute map for changeAttributeValues method using PyQGIS

Question

I have a PyQGIS script being run from within QGIS that updates attribute values in a GeoPackage, based on a mapping. It currently iterates through fields and features one by one, updating the values using QgsVectorLayer.changeAttributeValue.

This seems inefficient, particularly if there are many features. I'm looking to update it to use QgsVectorDataProvider.changeAttributeValues: https://qgis.github.io/pyqgis/3.28/core/QgsVectorDataProvider.html#qgis.core.QgsVectorDataProvider.changeAttributeValues

The changeAttributeValues method takes as a parameter "attr_map (object) – a map containing changed attributes". Unfortunately, the documentation doesn't set out the format of the map, and a Google search hasn't yielded any useful information.

The aim is to update the code to create the attribute map, and then run one operation using changeAttributeValues to update all changes at once. I am specifically looking for the format of the attribute map, though code is of course welcome.

The dummy existing code is as follows:

outfile = "c:/temp/output.gpkg"
oldFieldNames = ["oldField1", "oldField2", "oldField3"]
newFieldNames = ["newField1", "newField2", "newField3"]
mapping = [
 {0: "orange", 1: "apple", 2: "pineapple"},
 {0: "red", 1: "yellow", 2: "green", 3: "blue"},
 {0: "dog", 1: "cat", 2: "goat", 3: "cow", 4: "fish"},
]
vl = QgsVectorLayer(outfile, 'temp', 'ogr') 
pr = vl.dataProvider()
for i in range(len(oldFieldNames)):
 newFieldIndex = pr.fieldNameIndex(newFieldNames[i])
 with edit(vl):
 for feature in vl.getFeatures():
 vl.changeAttributeValue(feature.id(), newFieldIndex, mapping[i][feature[oldFieldNames[i]]])

If the initial layer attributes are as follows:

fid	oldField1	oldField2	oldField3	newField1	newField2	newField3
1	0	3	4
2	1	2	3

then the final result after mapping should be:

fid	oldField1	oldField2	oldField3	newField1	newField2	newField3
1	0	3	4	orange	blue	fish
2	1	2	3	apple	green	cow

As an aside, I have found quite regularly with the QGIS Python API documentation, that there is not enough detail to know the exact format of parameters required for some methods.

• bugmenot123: Not much help but you can look through tests in github.com/qgis/QGIS/search?q=changeAttributeValues to see how methods are used there.
• J. Monticolo: The C++ documentation gives sometimes more hints : api.qgis.org/api/… & api.qgis.org/api/…
• Bera: I'm no expert but when I have used it example, it is a dictionary of dictionaries: {feature_id: {field_index:new value}, ... }.

Answer 1

As @BERA says, the attribute map is a dictionary of dictionaries, with the first key as the feature ID, and the second key as the field index.

e.g.

attrMap = {
 featureId1: {fieldIndex1: value1, fieldIndex2: value2, fieldIndex3: value3}, 
 featureId2: {fieldIndex1: value4, fieldIndex2: value5}, 
 ...}

Updated code using changeAttributeValues (there may be more efficient ways of iterating):

outfile = "c:/temp/output.gpkg"
oldFieldNames = ["oldField1", "oldField2", "oldField3"]
newFieldNames = ["newField1", "newField2", "newField3"]
mapping = [
 {0: "orange", 1: "apple", 2: "pineapple"},
 {0: "red", 1: "yellow", 2: "green", 3: "blue"},
 {0: "dog", 1: "cat", 2: "goat", 3: "cow", 4: "fish"},
]
vl = QgsVectorLayer(outfile, 'temp', 'ogr') 
pr = vl.dataProvider()
attrMap = {}
newFieldIndices = []
for i in range(len(newFieldNames)):
 newFieldIndices.append(pr.fieldNameIndex(newFieldNames[i]))
for feature in vl.getFeatures():
 featureAttrMap = {}
 for i in range(len(oldFieldNames)):
 featureAttrMap[newFieldIndices[i]] = mapping[i][feature[oldFieldNames[i]]]
 attrMap[feature.id()] = featureAttrMap
# print(attrMap)
pr.changeAttributeValues(attrMap)

The attrMap data for the example above is:

attrMap = {
 1: {4: 'orange', 5: 'blue', 6: 'fish'}, 
 2: {4: 'apple', 5: 'green', 6: 'cow'}
}