Add Fermi-LAT 3FHL spatial models #1013
Conversation
adonath
changed the title
|
Travis-CI test fails are unrelated |
|
@adonath - Actually this one messed up the docs builds on travis, I suspect it's something in |
There was a problem hiding this comment.
I realise I'm late here (have been sick since Wednesday), but I still did a review. @adonath - Can you please do a follow-up PR taking the comments into account? (If no, I'll make a reminder issue for myself and do it next week.)
| pars['x_0'] = glon.value | ||
| pars['y_0'] = glat.value | ||
| return Delta2D(**pars) | ||
| else: |
There was a problem hiding this comment.
Maybe remove the else and dedent the rest?
| raise ValueError('Not a valid spatial model{}'.format(morph_type)) | ||
|
|
||
| @property | ||
| def pointlike(self): |
There was a problem hiding this comment.
Uncle Bob and the other clean code guys suggest to put is or has for boolean attributes, i.e. in this case pointlike -> is_pointlike. I think it makes sense and we should change here (and in other cases where needed in Gammapy.
| pars['y_mean'] = glat.value | ||
| pars['x_stddev'] = de['Model_SemiMajor'].to('deg').value | ||
| pars['y_stddev'] = de['Model_SemiMajor'].to('deg').value | ||
| pars['amplitude'] = amplitude * 1 / (2 * np.pi * pars['x_stddev'] ** 2) |
There was a problem hiding this comment.
Such amplitude to integral conversions belong on the model classes, to avoid duplicating the formulae in our codebase, no? Maybe as a static helper function to compute before __init__ or as an option in init or a method to call after init?
In this case just taking x_stddev and not considering y_stddev is a bug for asymm cases, no?
| See Also | ||
| -------- | ||
| Shell2D, Sphere2D, `~astropy.modeling.models.Gaussian2D` | ||
|
|
| template = SkyImage.read(filename) | ||
| return cls(template) | ||
|
|
||
| def evaluate(self, x, y, amplitude): |
There was a problem hiding this comment.
Add docstring and say what x / y is here. Pix or world coords? Units?
| return cls(template) | ||
|
|
||
| def evaluate(self, x, y, amplitude): | ||
| coord = SkyCoord(x, y, frame='galactic', unit='deg') |
There was a problem hiding this comment.
Hard-coding to galactic here is bad, no?
Avoid creating a SkyCoord object possible?
| return amplitude * values | ||
|
|
||
| @property | ||
| def bounding_box(self): |
There was a problem hiding this comment.
Move bounding_box to the SkyImage class?
Should be useful in other contents as well, no?
This is what we have:
http://docs.gammapy.org/en/latest/api/gammapy.image.SkyImage.html#gammapy.image.SkyImage.footprint
I thought this just computes the corners and that's what the docstring says.
Here I see that also e.g. center is computed.
@adonath - Did you add this?
I think this is bad, there can just be separate properties like center, or whatever is needed, and footprint should just return what the name says: the footprint (not center, width, height).
| center = footprint['center'] | ||
| x_0 = center.data.lon.wrap_at('180d').deg | ||
| y_0 = center.data.lat.deg | ||
| return ((y_0 - height / 2, y_0 + height / 2), |
There was a problem hiding this comment.
Tuples are bad, it's so easy to get the order wrong. And there's no docstring here explaining what this computes at all and what is returned.
| '/Templates/HESSJ1841-055.fits') | ||
| template = Template2D.read(filename) | ||
| assert_allclose(template(26.7, 0), 1.1553735159851262) | ||
|
|
There was a problem hiding this comment.
Please add an assert that exercises bounding_box and establishes the current result (to make it easier to refactor or change the code that computes it in the future).
This PR adds the
.spatial_modelattribute toSourceCatalogObject3FHL. The corresponding extended sources template archive is added in gammapy/gammapy-extra#78.