Improve coordinate equality comparison

As mentioned in #1832, coordinate comparisons currently just check that the internal angles and distances are the same:

>>> c1 = ICRS(1, 2, unit=['deg','deg'])
>>> c2 = FK5(1, 2, unit=['deg','deg'])
>>> c1 == c2
True

A more nuanced implementation of equality is needed, details TBD in APE5.

FYI, this is not explicitly addressed in the APE5 API. But my current thinking is that for low-level classes, equality means must be the same frame, all meatdata and data the same. For high-level classes it's a bit more ambiguous, but that can be addressed later. So this issue will be a reminder to get that figured out.

@taldcroft

@taldcroft - do you consider this closed by APE5, or should we push it to 1.0 and consider changing the equality operation then?

@eteq

@eteq - from what I can tell equality comparison has gone to the other extreme of just not being implemented. Am I missing something??

In [15]: c2 = ICRS(1*u.deg, 2*u.deg)

In [16]: c1 = ICRS(1*u.deg, 2*u.deg)

In [17]: c1 == c2
Out[17]: False

In [18]: c1.__eq__
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-18-b85dbdd0e83c> in <module>()
----> 1 c1.__eq__

AttributeError: 'ICRS' object has no attribute '__eq__'

@taldcroft

@taldcroft - That's right - I thought we concluded this is the safest for coordinates, given that it's ambiguous what is actually meant? It may be that's what I decided, and we didn't really talk about it, though... my memory is a bit hazy.

I thought we were going with "equal if all data and attributes are equal" as a safe option for low-level. I do think that c2 = copy(c1); assert c1 == c2 should work for 0.4.

What about having a coordinates.allclose method that transforms c1 into the c2 frame and checks that the sky separations are all less than a separation tolerance (angle) and the distances are the same within some rtol and atol (distance).

It seems this issue has become two: (1) there should be a way to test for strict equality; (2) there should be a way to test coordinates are close. I think the former should indeed exist but in practice is not very useful, while the latter will take quite a bit of work to get it really right. So, I suggest relabeling this issue to 0.4.1 to deal with equality only, and raise another for 1.0 for dealing with allclose.

p.s. Sorry, in case it wasn't clear from other issues: the goal is to get the list of 0.4 issues and PRs small, so that the overdue release can happen.

@mhvk

@mhvk - done. This is now 0.4.1 and I opened a feature request for allclose for 1.0 (#2650)

Is this still something that needs to be resolved?

The original snippet is not working any more, and

c1 = SkyCoord(1, 2, unit='deg',frame='icrs')
c2 = SkyCoord(1, 2, unit='deg',frame='fk5')
c1 == c2
False

so I wonder whether this can be closed now?

I'm closing this because it is a duplicate of #3182 and I think the discussion there is more relevant now.

Navigation Menu

Repository navigation

Improve coordinate equality comparison #2154

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Improve coordinate equality comparison #2154

Description

Activity

eteq commented on Mar 25, 2014

eteq commented on May 27, 2014

taldcroft commented on May 27, 2014

eteq commented on May 30, 2014

taldcroft commented on May 31, 2014

mhvk commented on Jun 17, 2014

mhvk commented on Jun 17, 2014

astrofrog commented on Jun 22, 2014

embray commented on Dec 2, 2014

bsipocz commented on Jul 31, 2017

adrn commented on Sep 19, 2017

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions