Ok, let me describe in details what I do, because the switch did not help.
open FR image (level-1 OLCI and SLSTR)
subset it to the region of interest
extract 3*3 pixel values around one coordinate somewhere in this region of interest
I suppose, accurate coordinates of extracted pixels are important for me (accuracy +/-0.001).
The subset is saved as “NetCDF4-BEAM” and has lat and lon (as you said, it is added by your writer). But now the reader (even if per-pixel geo-coding is specified) treats only lat and lon as coordinates.
Therefore pixel extraction (step 3) occurs with lat, lon not with per-pixel values.
The situation is even worse with SLSTR, because it has ao, an, bo, bn, co, cn… latitudes and longitudes. I assume they are also more accurate than lat, lon.
The question is: how can I extract pixels, according to their per-pixel coordinates from subset image?
Your were right that SNAP and snappy share common options, but somehow it is not possible to subset by per-pixel coordinates with snappy.jpy.get_type(‘org.esa.snap.core.gpf.common.SubsetOp’). It gives an error
RuntimeError: org.esa.snap.core.gpf.OperatorException: latBand.getProduct().getSceneRasterWidth() < 2 || latBand.getProduct().getSceneRasterHeight() < 2