symbol for categorized raster looks like a mask

[Andreas Janz]

Find test file attached. The default renderer is a PalettedRenderer because of the QML sidecar. In the Data Views panel category legend looks good. Not sure why it is concidered a mask (?)

‌

‌

[Andreas Janz]

Also, opening the raster via “Open in new map… / Default Colors” is not using the default style from the QML:

‌

[Andreas Janz]

Note that drag&drop from Data Sources panel into a map view is working fine.

[Benjamin Jakimow]

Mhhh. @Andreas Rabe the classification.tif cannot be downloaded from Bitbucket

‌

[Benjamin Jakimow]

resolves #674 - DataSourceRaster().createUnregistereMapLayer() loads default style - DataSourceTreeView().openInMap( ... DEFAULT) tries to get default style from map layer first, GDAL second.

→ <<cset e0eb121d141d>>

[Andreas Janz]

@Benjamin Jakimow looks good now, thanks!

Regarding GDAL Categories: I think you do not have to handle GDAL categories at all. If GDAL categories are available, QGIS will default to a paletted raster renderer. Do you see any case where checking the GDAL categories explicitely is required? At leased I’m fully relying only on QGIS inside my algorithms. Maybe I’m missing something(?)

[Benjamin Jakimow]

Nope, without a GDAL color table and categories only, a QgsRasterLayer will get a QgsSingleBandGrayRenderer. Try this:

‌

import numpy as np
from osgeo import gdal_array, gdal
from qgis.core import QgsRasterLayer, QgsPalettedRasterRenderer
array = np.asarray([[1,2,3,2],
                    [0,1,1,2]]).astype(np.uint8)
path = '/vsimem/myimage.tif'
ds: gdal.Dataset = gdal_array.SaveArray(array, path)
band: gdal.Band = ds.GetRasterBand(1)
band.SetCategoryNames(['none', 'A', 'B', 'C', 'D'])

# set True to add colors and get a QgsPalettedRasterRenderer instead
if False:
    ct = gdal.ColorTable()
    ct.SetColorEntry(0, (0, 0, 0))
    ct.SetColorEntry(1, (0, 255, 0))
    ct.SetColorEntry(3, (255, 0, 0))
    band.SetColorTable(ct)

ds.FlushCache()
lyr = QgsRasterLayer(path)
assert lyr.isValid()
renderer = lyr.renderer()
print(renderer)

‌

[Andreas Janz]

You have to prepare a QML file of course

See classification.qml attached.

[Benjamin Jakimow]

relates to #166