[feature request] Data sizes

[ygrange]

Ah, finally now that the paper is accepted (for which I congratulate you) I feel like I can ethically open this issue.

I'm not sure how appropriate this proposal is so I'd rather start an issue before hacking it in and doing a pull request that may be refused anyhow (e.g. because of scope creep).

I would like to add units of information (which seems to be the base quantity of the units Byte, bit, GiB, Mbit, etc).

It would be neat if I could use unyt to answer questions like "how much time does it take to transfer 5TiB of data using a 10Gbit network link?", without having to struggle with factors of 8 and weird prefixes.

As far as I can see before doing any work, that would require adding a few prefixes (ki to Yi), two base units (bit, byte) and two dimensions (information, bandwidth).

What is your opinion on this? Would it be worth the effort?

[ngoldbaum]

I think this would be fine to add. We have prior art of this in that angle is a base dimension even though it's technically not an SI dimension. For the existing unit systems it would probably be best to just choose a sensible default for the dimensions you're adding and use that default for all the unit systems.

[ygrange]

Ok coo. I have already been playing in a fork. Will probably take some time due to holidays :)

[ngoldbaum]

It also occurs to me that you shouldn't be able to create a Mig (e.g. a mibigram), so you'd need to add handling to the code that sets up SI prefixes to do special things for units with dimensions of data.

[ygrange]

I'll have a look at it in the coming period. Until now I took the crude approach (if the user wants a Mig, they get a Mig). I am a bit torn between two options: throwing a warning (I am not sure if you have any warning system implemented in the first place anyhhow) or an exception in this case. Do you have an opinion on this?

[ngoldbaum]

It should throw the same error as the error you get if you ask for any other unit name that doesn't exist:

In [2]: unyt.Unit('foobar')
---------------------------------------------------------------------------
UnitParseError                            Traceback (most recent call last)
<ipython-input-2-ac9713246324> in <module>()
----> 1 unyt.Unit('foobar')

~/Documents/unyt/unyt/unit_object.py in __new__(cls, unit_expr, base_value, base_offset, dimensions, registry, latex_repr)
    308         else:
    309             # lookup the unit symbols
--> 310             unit_data = _get_unit_data_from_expr(unit_expr, registry.lut)
    311             base_value = unit_data[0]
    312             dimensions = unit_data[1]

~/Documents/unyt/unyt/unit_object.py in _get_unit_data_from_expr(unit_expr, unit_symbol_lut)
    869
    870     if isinstance(unit_expr, Symbol):
--> 871         return _lookup_unit_symbol(str(unit_expr), unit_symbol_lut)
    872
    873     if isinstance(unit_expr, Pow):

~/Documents/unyt/unyt/unit_registry.py in _lookup_unit_symbol(symbol_str, unit_symbol_lut)
    316     # no dice
    317     raise UnitParseError("Could not find unit symbol '%s' in the provided "
--> 318                          "symbols." % symbol_str)
    319
    320

UnitParseError: Could not find unit symbol 'foobar' in the provided symbols.

[ngoldbaum]

(maybe not the same error text, just raising UnitParseError is sufficient to keep things sane :) )