How should one parse unit names that are not valid identifiers?

[jameshiebert]

I'm using pint to do checking and conversion of units for some database insertion code and I'm a bit confused about how to define the unit '%'.

Basically, I have a set of weather observations coming in from multi-source input files that are attributed with units, and I'm stuffing them into a database table that may be attributed with different (but compatible) units.

In particular, we're pulling in relative humidity data that are attributed with the units "%" in the database.

Based on the comments in #226, I assumed that the new parser would allow units like " ' and %. But then in pint's test suite the test for this case seems to be intended to fail?

    @unittest.expectedFailure
    def test_issue25(self):
        x = ParserHelper.from_string('10 %')

Or am I reading that wrong?

Defining the unit '%' doesn't error out, but then trying to parse it does:

>>> import pint
>>> ureg = pint.UnitRegistry()
>>> ureg.define('% = percent')
>>> ureg.parse_expression('%')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/hiebert/code/crmprtd/env36/lib/python3.6/site-packages/pint/registry.py", line 840, in parse_expression
    return build_eval_tree(gen).evaluate(lambda x: self._eval_token(x,
AttributeError: 'NoneType' object has no attribute 'evaluate'

Is there some other way that I should be doing this? Thanks for any info that you can provide.

[hgrecco]

The new parser should definitely allow this as we have decoupled from Python's eval. But we are still using the Python tokenizer. Notice this:
The interesting thing is that:

>>> ureg.parse_expression('10%')
10

I think that the quickest way in your case will be to introduce a replacement rule '%' -> '* 0.01'.

[dalito]

Also interesting:

>>> ureg.parse_expression(u'2%4')
8                               
>>> 2%4                         
2                               

[jameshiebert]

Thanks @hgrecco , I'll give that a shot.

>>> 2%4 
2

Well... that's 2 [mod] 4, so that doesn't surprise me.

[hgrecco]

You can now add your own preprocessors See #911

[jameshiebert]

Perfect, thanks!