Support selecting a percent of total flux given a clean mask file #54
Conversation
sjperkins
force-pushed
the
clean-mask-file
branch
from
4a8d79e to
4bb2803
Compare
| @@ -66,6 +66,8 @@ def create_parser(): | |||
| help="Fraction of system RAM that can be used. " | |||
| "Used when setting automatically the " | |||
| "chunk size. Default in 0.1.") | |||
| p.add_argument("--clean-mask-file", required=False, default="", | |||
| help="Clean Mask File. If supplied") | |||
There was a problem hiding this comment.
I should update the help description
| if m := re.match("^SoFiA (?P<version>\\d+\\.\\d+\\.\\d+)$", origin): | ||
| major, _, _ = map(int, m.group("version").split(".")) | ||
| if major < 2: | ||
| raise ValueError(f"SoFiA major version is less than 2: {origin}") |
There was a problem hiding this comment.
Check that the Clean Mask file is produce by SoFiA 2.
| try: | ||
| assert header["CTYPE1"].strip() == "RA---SIN" | ||
| assert header["CTYPE2"].strip() == "DEC--SIN" | ||
| assert header["CTYPE3"].strip() == "VRAD" |
There was a problem hiding this comment.
Assume radio velocity quantity, but potentially other types could be supported.
|
|
||
| freq = SpectralQuantity(wsclean_comps["ReferenceFrequency"] * u.Hz, | ||
| doppler_rest=wcs.wcs.restfrq * u.Hz, | ||
| doppler_convention="radio") |
There was a problem hiding this comment.
It would be nice to find some better way to do this conversion as a Radio doppler convention os assumed here. The restfreq is on the wcs attribute for example. I intuit that the wcs.spectral atttribute (which is a wcs object for the Spectral coordinate axis) should be used, but a simple wcs.spectral.world_to_pixel didn't work for me..
| integrated_fluxes = np.array([flux[sid == source_ids].sum() for sid in source_id_range]) | ||
| broadcast_fluxes = integrated_fluxes[source_ids - 1] | ||
|
|
||
| comp_sort_idx = np.lexsort((flux, broadcast_fluxes))[::-1] |
There was a problem hiding this comment.
Broadcast the total flux for each source over the number of a components and lexically sort. This allows us to partially select components of a source.
| (2, 4, 5, 100), | ||
| (2, 3, 4, -100), | ||
| (2, 3, 2, -100), | ||
| (3, 0, 1, -25)] |
There was a problem hiding this comment.
Model 3 sources
- 3 components
- 2 components
- 1 component
|
|
||
|
|
||
| def test_clean_mask(wsclean_model_and_clean_mask): | ||
| # 3 sources |
There was a problem hiding this comment.
Test that the appropriate components are chosen given the percentage flux. I should change this to check that the appropriate sources are selected.
|
|
||
| source_ids = clean_mask[z, y, x] | ||
|
|
||
| if(np.any(source_ids == 0)): |
There was a problem hiding this comment.
It's possible that the wsclean components could register to the edge of a clean mask pixel, in which case coordinates could map to the wrong source_id.
I suspect this won't be a problem in practice, but I guess the only way to find out is to try this PR with real data.
Given a clean mask file produced by Sofia 2, [registers] and groups (https://en.wikipedia.org/wiki/Image_registration) a wsclean component list by source ID's in a SoFiA2 Clean Mask file. Components are lexicographically sorted by
(total source flux, component flux)and a percentage of components contributing to total flux are selected.Tests added / passed
$ py.test --flake8 -v -s .If the flake8 tests fail, the quickest way to correct
this is to run
autopep8and thenflake8to fix the remaining issues.