SpectralProfile fields information is getting lost when vector layer is passed to a processing algorithm

[Andreas Rabe]

Hi @Benjamin Jakimow ,

the current test that decides if a field is an actual SpectralProfile field, checks against the EditorWidgetSetup-Type:

‌

This information seams to be missing, after the spectral library layer is passed to a processing algorithm.

Have you any idea, how to circumvent this?

Of course, the user could manually select all SpectralProfile fields as an input parameter to the algorithm, but that would be very inconvinient.

[Andreas Rabe]

@Benjamin Jakimow , what I also don’t like about the current solution: the EditorWidgetSetup-Type is a property of the layer. It would make mutch more sense, if it would be specified inside the vector source.

[Benjamin Jakimow]

Please provide a code example / unit thes that reproduces this behavior. Which processing algorithm do you mean?

The editor widget type is a property of the QgsField and available at QgsFeature level as well.

‌

[Benjamin Jakimow]

By the way, there is nothing else we could check against if a QgsField is used for SpectraProfiles or not.

[Andreas Rabe]

Please provide a code example / unit thes that reproduces this behavior. Which processing algorithm do you mean?

It’s the SaveLibraryAsGeoJsonAlgorithm. I now implemented a check inside the algortithm, so you can use it to reproduce and debug:

‌

Traceback (most recent call last):
File "D:\source\QGISPlugIns\enmap-box\site-packages\typeguard\__init__.py", line 903, in wrapper
retval = func(*args, **kwargs)
File "D:\source\QGISPlugIns\enmap-box\enmapboxprocessing\algorithm\savelibraryasgeojsonalgorithm.py", line 66, in processAlgorithm
raise Exception('DEBUG Issue #1092: profile_field_list(library) returns empty list.')
Exception: DEBUG Issue #1092: profile_field_list(library) returns empty list.

‌

[Andreas Rabe]

‌

[Andreas Rabe]

To be clear, this issue is not about the SaveLibraryAsGeoJsonAlgorithm as such. Every processing algorithm that want’s to process a library layer is affected.

[Benjamin Jakimow]

The EnMAPProcessingAlgorithm fails to keep the enditor widget setup information

‌

class TestIssue1089(EnMAPBoxTestCase):

    def test_memoryConversion(self):
        slib = TestObjects.createSpectralLibrary()
        QgsProject.instance().addMapLayer(slib)
        context = QgsProcessingContext()
        context.setProject(QgsProject.instance())
        feedback = QgsProcessingFeedback()
        context.setFeedback(feedback)

        alg = EnMAPProcessingAlgorithm()
        alg.addParameterVectorLayer('memory', '')
        params = {'memory': slib}

        slib2 = alg.parameterAsVectorLayer(params, 'memory', context)

        for field in slib.fields():
            field2 = slib2.fields().field(field.name())
            self.assertEqual(field.type(), field2.type())
            self.assertEqual(field.editorWidgetSetup().type(), field2.editorWidgetSetup().type())

‌

Furthermore I don’t know if this one saves the qml as well.

‌

            # save as GeoJSON
            alg = 'native:savefeatures'
            parameters = {'INPUT': library, 'OUTPUT': filename}
            self.runAlg(alg, parameters, None, feedback2, context, True)

‌

[Andreas Rabe]

Oh no, you are right. Our fancy handling of QGIS memory layers, to make them OGR compatible, is biting us back

‌

[Andreas Rabe]

Hmm, I think it is best to not convert to OGR by default. I’ll add a “convertToOgr=False” option to the parameterAsVectorLayer method.

[Benjamin Jakimow]

I can fix this if you like

[Andreas Rabe]

No thanks, I’m already at it.

[Andreas Janz]

resolves #1092

→ <<cset d4c88ad5c98a>>