Feature request #18444

Memory layer does not cast attribute types

Added by Norwin Roosen about 6 years ago. Updated about 5 years ago.

Status:Feedback
Priority:Normal
Assignee:-
Category:Data Provider
Pull Request or Patch supplied:No Resolution:
Easy fix?:Yes Copied to github as #:26332

Description

Expected Behaviour:
Writing values of incorrect type to a memory layer field should cast them to the correct value or throw an error.

Actual behaviour:
In QGIS 2.18.16, writing a string into a double field of a memory layer writes the value unchanged.
Reading the attribute later returns the original value of type unicode.

The following python code illustrates this:

from qgis.core import QgsVectorLayer, QgsFeature

layer =  QgsVectorLayer("Point?field=floatfield:double", "poc", "memory")
feat = QgsFeature()
feat.setAttributes(['10'])
layer.dataProvider().addFeatures([feat])
QgsMapLayerRegistry.instance().addMapLayer(layer)

# in my usecase I actually loaded the layer from within another processing algo
for f in layer.getFeatures():
    print f['floatfield'], type(f['floatfield'])

# prints
#   10 <type 'unicode'>
# should obviously be
#   10.0 <type 'float'>

History

#1 Updated by Jürgen Fischer over 5 years ago

  • Status changed from Open to Feedback

Please test with QGIS 3.4 - QGIS 2.18 reached it's end of life.

#2 Updated by matteo ghetta about 5 years ago

Tested on QGIS 3.4 (with some small changes to the script):

from qgis.core import QgsVectorLayer, QgsFeature

layer =  QgsVectorLayer("Point?field=floatfield:double", "poc", "memory")
feat = QgsFeature()
feat.setAttributes(['10'])
layer.dataProvider().addFeatures([feat])
QgsProject.instance().addMapLayer(layer)

# in my usecase I actually loaded the layer from within another processing algo
for f in layer.getFeatures():
    print(f['floatfield'], type(f['floatfield']))

same behavior of your description. IMHO it is not a bug that QGIS casts automatically the data type. Maybe a warning could be thrown, but this could be a feature request?

#3 Updated by Norwin Roosen about 5 years ago

Thanks for testing in QGIS 3.4 @matteo!
To me this issue is about semantic correctness of the API & predictability of its results.
Putting a string in a float field is never semantically meaningful, so raising an error may be appropriate.
We shure can argue if following the semantics of the API is a feature or a bugfix, but what matters is that there is working software in the end ;)

#4 Updated by matteo ghetta about 5 years ago

  • Tracker changed from Bug report to Feature request

thanks Norwin Roosen. OK then I'll mark it as feature request. So it will stay open as reminder.

Also available in: Atom PDF