Functional units in FITS and VOTable

This is to remind me/us that now that function units are present (#1894), it should also be possible to write/read them to FITS and VOTable, at least for the subset for which the standard allows this.

EDIT: For instance, in fits dex(erg/s) can be written as log(erg/s).

The general writer is addressed in #3618.

👍

@mhvk

I removed the milestone, @mhvk . FYI.

@olebole

@olebole would also like a new u.log unit. See #16540 (comment)

I'd like to bring this back to the top again, because it seems to me the issue is almost complete now:

we already can generally parse functional units, see Enable parsing of mag, dB, dex #3554
we can generally write functional units

This should be rather trivial, not? I could try to create a PR here, but before I would like to clarify what prevented us from implementing this also for FITS and VO, at least for dimensionless mag and for dex (which can be represented with the log function in FITS and VO)?

i.e.

>>> u.mag(1).to_string("vounit")
'mag'
>>> u.dex(1).to_string("vounit")
'log(1)'
>>> u.dex(u.solLum).to_string("vounit")
'log(solLum)'

The first one doesn't pass a roundtrip, as u.Unit("mag", format="vounit") doesn't return a magUnit, but the log/dex representation should be simple, not?

@eerovaher

Also cc @eerovaher and @nstarman who also work on units a lot lately.

I don't know enough about function units to contribute to this meaningfully.

I would like to clarify what prevented us from implementing this also for FITS and VO, at least for dimensionless mag and for dex

I think that was before my time.
Seems like something we should support, and if it's easy to do now, that's great!

It never was particularly difficult, just needs someone to put in the time to think how exactly it should work and then implement it -- apparently, I didn't find that time in 9 years... If someone can write a series of tests, that would be a great start...

This is something I have just bumped into as well, somewhat related to the above #17297. I would love to see this as well.

In our case, it was from another package that was using astropy.units to validate FITS and VO unit strings. I was very surprised when I found that VOUnit("log(solMass)") did not pass as valid, when the syntax is allowed by the standard.

Yes, it would be good. Not sure I'll get around it myself soon, but there are some hints here and especially in #17297 on how to implement it. As I wrote here, even just having a good set of tests would be a great start.

Perhaps the original issue was that log/ln/exp() function units are mathematically (very) different from "ordinary" units, which are inherently multiplicative. Note that they are not at all like sqrt() -- e.g., sqrt(m) is still a multiplicative unit.

Is the issue not one of string-parsing, but rather one of what Astropy is supposed to do when a unit conversion is requested? Perhaps that code isn't ready for this?

@gpdf

@gpdf - astropy supports these "functional" units, which are used for Magnitude, Decibel, etc. The issue here is just about correctly translating these to FITS and VOTable (as is already possible for CDS format).

@gpdf

@gpdf Yes, I will admit my problem here is slightly reductionist. The angle from which I encountered this issue was not unit conversion or mathematical support, but straight parsing.

My naive confusion was simply that VOUnit.parse() should be able to parse (in some form) any VOUnit. The fact that an exception was thrown when parsing a syntactically correct VOUnit is what threw me for a loop.

Whether or not it can correctly map log(solMass) to the correct astropy function/unit is less worrisome to me than the fact that it should not throw a:

>>> VOUnit.parse("log(solMass)")
ValueError: 'log' is not a recognized function

error for a correct VOUnit.

@jwfraustro

@jwfraustro Yes, understood. So perhaps the situation is just "haven't gotten to it" rather than "difficult to implement correctly"?

Navigation Menu

Repository navigation

Functional units in FITS and VOTable #3822

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Functional units in FITS and VOTable #3822

Description

Activity

embray commented on Jun 5, 2015

pllim commented on Oct 24, 2019

pllim commented on Jun 5, 2024

olebole commented on Oct 22, 2024

pllim commented on Oct 22, 2024

eerovaher commented on Oct 22, 2024

nstarman commented on Oct 22, 2024

mhvk commented on Oct 22, 2024

jwfraustro commented on Dec 13, 2024

mhvk commented on Dec 13, 2024

gpdf commented on Dec 17, 2024

mhvk commented on Dec 17, 2024

jwfraustro commented on Dec 17, 2024

gpdf commented on Dec 17, 2024

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions