Performance issue with WCS units

I’m running into a performance issue where accessing WCS.wcs.cunit is the bottleneck. For example:

In [19]: wcs = WCS(naxis=2)

In [20]: wcs.wcs.cunit = ['deg', 'm/s']

In [21]: wcs.wcs.cunit
Out[21]: ['deg', 'm s-1']

In [22]: %timeit wcs.wcs.cunit[0]
8.24 µs ± 18.8 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)

In [23]: %timeit wcs.wcs.cunit[1]
1.19 ms ± 13.9 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

Accessing the second unit takes over 100 times longer than the first one. If I initialize the same unit directly I get better performance:

In [24]: %timeit Unit('m/s')
72.9 µs ± 1.35 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)

In [25]: %timeit Unit('m s-1')
87.8 µs ± 348 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

so I’m not sure what’s going on here. In addition, I think it would be nice to have a way to optionally access the units simply as strings rather than as astropy units.

[astrofrog]

@himanshupathak21061998 - the underlying issue is not in repr, that is just a way to reproduce the performance problem. This issue is not ideal for first-time contributors as it requires understanding the WCS and units packages in detail, but if you really want to work on it, I would suggest looking into tools like pyinstrument and line_profiler to profile the relevant code to see how we might be able to speed things up.

[dstansby]

I think https://github.com/astropy/astropy/pull/12699 improves this a lot, and possibly resolves this issue?