Adds support for -b/--band to avoid reading or writing irrelevant bands #37
Conversation
alpha-beta-soup
commented
Oct 24, 2025
alpha-beta-soup
commented
- Built upon Adds support for the A5 DGGS #36
- Closes Only index some bands #16
- Closes Support for multi-band rasters with nulls in some (but not all) bands #12
- puts CLI help text in consts to ease update burden - edge case handling for mode aggregation to handle all-nan series
alpha-beta-soup
added
bug
alpha-beta-soup
changed the title
ndemaio
left a comment
ndemaio
left a comment
There was a problem hiding this comment.
The parameter proliferation continues! :D But the optimisation does make sense. I think you're the best person to decide at which point the command line arguments get too involved. (I think Linux shell commands have some truly terrible examples...) I don't know if I've got a good alternative off the top of my head either - it feels like something that needs to be looked at on a case by case basis. But it doesn't seem to be a problem just yet so all good.
| with rio.Env(CHECK_WITH_INVERT_PROJ=True): | ||
| with rio.open(raster_input) as src: | ||
| with rio.open(raster_input, mode='r', sharing=False) as src: | ||
| LOGGER.debug("Source CRS: %s", src.crs) |
There was a problem hiding this comment.
This looks far more controlled, does it fix a specific thing as well?
There was a problem hiding this comment.
No, it's actually the default args. I was just making it explicit (and I guess also protecting against the defaults changing). I'm not actually sure that we've parallelised this properly, I need to look into it.
There was a problem hiding this comment.
There are definitely some adjustments to be made here. A start is #38, another could be to reconfigure reads so there isn't a shared WarpedVRT that requires a dask.utils.SerializableLock. Chunks of windows could be read in separate processes. Something for another day unfortunately.
- puts CLI help text in consts to ease update burden - edge case handling for mode aggregation to handle all-nan series
My reference point is GDAL, so I think I'm in good company... https://gdal.org/programs/ogr2ogr.html |
