[FB] [PI-3478] Add resolve doc-strings #3842
Conversation
| if self.lhs_cube.ndim >= self.rhs_cube.ndim: | ||
| self.map_rhs_to_lhs = True | ||
| else: | ||
| self.map_rhs_to_lhs = False |
There was a problem hiding this comment.
Decided to make a slight infrastructure change to support the documentation of Resolve attribute members.
This involves pretty much taking the contents of _init and relocating it to __init__ so that sphinx captures the doc-strings (#:) of Resolve attribute members (this must physically happen in __init__ and not indirectly via calling _init from __init__, otherwise sphinx won't document them), and ensuring any other necessary logic is placed here in __call__.
| self._metadata_prepare() | ||
|
|
||
| return self | ||
|
|
There was a problem hiding this comment.
Returning self supports method chaining after making a call to a Resolve instance, which I wanted to do in the doc-tests e.g., now supports such patterns like
resolve = Resolve()
resolve(c1, c2).map_rhs_to_lhswhich is pretty handy for certain workflows
| free_flip = len(tgt_free) > len(src_free) | ||
| free_flip = src_cube.ndim == tgt_cube.ndim and len(tgt_free) > len( | ||
| src_free | ||
| ) |
There was a problem hiding this comment.
This was merged to the feature-branch cube-arithmetic-post as part of #3805
| self.category_common = _CategoryItems( | ||
| items_dim=[], items_aux=[], items_scalar=[] | ||
| ) | ||
|
|
There was a problem hiding this comment.
Do this here, rather in the now defunct _init.
And we don't want to do this within __init__, as we want it to happen per __call__
|
|
||
| @property | ||
| def _src_cube(self): | ||
| assert self.map_rhs_to_lhs is not None |
There was a problem hiding this comment.
Defensive code to ensure the _init restructuring to __init__ et al, hasn't had any side-effects that we don't know about.
But in general, it's a healthy assert to adopt IMHO
| result = None | ||
| if self.mapping is not None: | ||
| result = self._src_cube.ndim == len(self.mapping) | ||
| return result |
There was a problem hiding this comment.
Adds a bit more rigour to deal with the case where the user attempt to determine mapped, but no operands have actually been provisioned to Resolve
| data = np.zeros(cube1.shape) | ||
| data1 = data * 10 | ||
| data2 = data * 20 | ||
| data3 = data * 30 |
There was a problem hiding this comment.
Note that, the testsetup instances created here for this doc-string is available for all other doc-strings, thanks to the magic of sphinx.
Hence, there is no need to replicate the testsetup directive per doc-string test.
| STASH: m01s03i236 | ||
| source: Data from Met Office Unified Model 6.05 | ||
| Cell methods: | ||
| mean: time (6 hour) |
There was a problem hiding this comment.
The doc-test equivalent of a picture to 1000 words... there is no need to describe what's happening to a detailed level... just show it. Works for me as a philosophy 😉
|
thanks for the hangout @bjlittle to walk thru this. LGTM. |
|
@tkknight Awesome, thanks 👍 |
* Add resolve doc-strings * reorder some doctest statements
2 participants
This PR adds doc-strings and doc-tests primarily for
iris.common.resolve.ResolveNote that, full test coverage for
Resolveis to follow as part of the roadmap for cube arithmetic - as agreed offline with other core devs