[AstroPy] astropy.cosmology

[Erik Tollerud]

(Passing on a message from Alex Conley here about a complication in
the inverse function scheme, same as Rick White pointed out.)

But I disagree that this is "bloat" - the fact that three people on
this list have already said they need it for science shows that it's a
useful thing to include.  Inverting it with scipy functions is not
necessarily trivial, and having an informative warning that
specifically states the non-monotonicity is in the cosmology module is
a reason to make it specific to astropy.cosmology .

Anyway, we'll see what Neil comes up with and see then whether it
seems like it will be difficult to maintain (the main disavantage of
"bloat").


On Wed, Dec 11, 2013 at 11:58 AM, Alex Conley <alexanderconley at gmail.com> wrote:
> That seems pretty contrived, honestly.  It doesn’t sound like you were up
> to much good when you did this.
>
> Maybe a more serious issue: luminosity distance is not monotonic in closed Universes.
> And the angular diameter distance is not monotonic in most, including the
> one we live in.  Age, at least, is always monotonic, at least in Universes with a Big Bang.
> So most of the inverses you want are ill-defined.
>
>>>> from astropy.cosmology import LambdaCDM
>>>> cos = LambdaCDM(70, 0.2, 1.4)
>>>> cos.luminosity_distance([1.3044, 3.0, 5.5])
> <Quantity [ 11800.64030641, 17066.4066551 , 11800.83252956] Mpc>
>
>>>> from astropy.cosmology import WMAP9
>>>> WMAP9.angular_diameter_distance([0.668, 3, 4.0])
> <Quantity [ 1468.8531073 , 1625.97548513, 1468.88895528] Mpc>
>
> Even assuming you overcame this problem — what about the function
> inverter says: should be specific to astropy.cosmology?  This screams
> code bloat to me.
>
> PS: I’m not subscribed to astropy at scipy.org, so it will not let me
> post.
> On 11-Dec-2013, at 9:41 AM, Erik Tollerud <erik.tollerud at gmail.com> wrote:
>
>>> I don’t see much point in implementing reverse functions of -everything- —
>>> it’s hard to imagine scenarios where most of them would be interesting.
>>> When will somebody want to know what redshift corresponds to a given
>>> luminosity distance?   Age is really the only one I can imagine having
>>> wide interest, so I would just start with that in the interest of avoiding
>>> bloat.
>>
>> Suppose you have a galaxy where you think you know what its absolute
>> magnitude should be (say its a standard candle) and have an apparent
>> magnitude, and you want to use it to know what redshift it's at.  Then
>> you need this, because the distance modulus is tied to the luminosity
>> distance (this example came to mind because I've actually needed this
>> for actual science in the past).  My point here is mainly that just
>> because we don't want all of them, doesn't mean some other users don't
>> have a need.
>>
>> That said, I see your point in that others are probably generally less
>> useful (e.g. AD distance), and duplicating every function/method does
>> make everything harder to deal with. Perhaps the solution is to add an
>> `inverse` function to `astropy.cosmology`?  You could then do
>> something like ``z = cosmology.inverse(WMAP.age, 1 * u.Gyr)``.  That
>> would keep the API cleaner, and it would be fairly trivial to
>> implement it that way using the method Juande and Alex are talking
>> about.  Then if we decide to optimize some of them later, we can add
>> those in and have the `inverse` function use those if they are
>> present.




-- 
Erik