EXP: For big cube, be a minimalist #8
Conversation
and avoid loading uncert and correct units [ci skip] [rtd skip]
kecnry
force-pushed
the
cubeviz-spec-extract-through-plugin
branch
3 times, most recently
from
b7873f5 to
a39bb5a
Compare
| if not is_big_cube: | ||
| sc = _return_spectrum_with_correct_units( | ||
| flux, wcs, metadata, data_type, hdulist=hdulist) | ||
| else: | ||
| sc = Spectrum1D(flux=flux, wcs=wcs, meta=metadata) |
There was a problem hiding this comment.
will this also skip any necessary units stuff? Can we just set/pass hdulist=None and still use _return_spectrum_with_correct_units or is that still too expensive?
There was a problem hiding this comment.
_return_spectrum_with_correct_units creates a new Spectrum1D object. I have not looked into whether it does a copy or what. I was just trying some possible low hanging fruits to see if anything helps. Given the ticket is only 2 points, I didn't dig too deep here. There is a chance we do not have to disable this (and I don't think we want to disable if we can avoid it).
It skips parsing uncert and mask but flux should still be loaded. |
|
My diff here is only exploratory. It didn't completely fix the lag in loading and slice. I tried profiling the code on |
Gotcha 👍 |
|
Follow up tickets created. This is no longer necessary. |
Companion to:
Seems a little more performant when using spacetelescope#2827 instead of
mainwith these changes, but slider still too laggy to be usable (but not as laggy asmain). Subset creation at least possible (not possible onmain).🐱