Adding PARSEC/COLIBRI + MIST stellar model support

[lcjohnso]

Work on stellar model expansion: Issues https://github.com/karllark/beast/issues/25 and https://github.com/karllark/beast/issues/22.

[mfouesneau]

1 bug in the imports: they need to be relative

a few warnings (code and science)

[mfouesneau]

I would add in the doc string that 0.015 comes with the new Zsun adopted in the recent Padova models. Just for completeness with documentation

[lcjohnso]

Done.

[mfouesneau]

I found this not to work all the time. Maybe being more explicit?

if oiso is None: 
    oiso = isochrone.PadovaWeb()

[mfouesneau]

I see that all classes derive from Isochrone class. That's good. So maybe my previous comment should even be

if oiso is None: 
    oiso = isochrone.PadovaWeb()
else if not isinstance(oiso, Isochrone):
    raise AttributeError("oiso is not of Isochrone type")

[lcjohnso]

@mfouesneau -- Can you elaborate on "not working all the time"? I haven't tested this yet due to the need to rebase before fully implementing changes back to datamodel.py. I was following format used for stellib objects in make_spectra(), but happy to make a change if necessary.

[mfouesneau]

I did not find what goes wrong sometimes and not all the time. I know I used it many times and not surprisingly in stellibs. I found safer to any case my alternative even if less compact.

[mfouesneau]

This comes from Yumi's needs. It filters out PMS data points. Maybe moving that back into the PadovaWeb class as an option?

[lcjohnso]

I moved this cut back into PadovaWeb, in _filter_iso_points(), but have all cuts commented out. I need to enable keyword passing (filterPMS & filterBad boolean variables) to this function so that these cuts can be optionally applied.

[mfouesneau]

Maybe just document the filter function and remove the _ in front (that indicates private/don't use by convention). The important part is to tell Yumi too. It makes sense that you don't want it to be generic for all projects.

[galaxyumi]

I think filtering PMS leaves as an option is a good idea.

[mfouesneau]

Need to correct the imports

from .ezpadova import parsec
from .ezmist import mist

[lcjohnso]

Done.

[mfouesneau]

Careful, that new for Padova, they predict the initial and current Z values meaning Z(tage) is variable like the current mass in a consistent way (was not before)
We should document that somewhere that we only give Zini

[lcjohnso]

To my knowledge, Z==Zini is assumed throughout the codebase, but we should make that explicit in datamodel.py -- that's on my ToDo.

[mfouesneau]

I do not think it will change a lot in the content but worth mentioning that the reported Z values are Zinit not current Z as given by Padova (I don't know if MESA gives current)

[lcjohnso]

@mfouesneau -- Can you comment on this? Acceptable to remove these attributes of the PadovaWeb object?

[mfouesneau]

Don't you need these to query the website?

[lcjohnso]

No -- I successfully ran a model generation all the way through to fitting with these lines commented. The values necessary to define the grid are passed as parameters directly to the object's _get_t_isochrones() function within make_iso_table() in models.py, and these are never touched.

[mfouesneau]

Then you're good to me. No need to keep it. They come from the header of the files from the website in case someone needs to access to it somewhere.

[lcjohnso]

To clarify: these object attributes (self.logtmin, self.logtmax, self.dlogt) for the ezIsoch object are used -- perhaps that's what you remembered.