"NAN" to invalid json

[Sasha4ever]

>>> from lxml import etree
>>> from xmljson import XMLData
>>> XMLData(attr_prefix=None, dict_type=dict).data(etree.XML('<ROOT test="NAN"/>'))
{'ROOT': {'test': nan}}

[Sasha4ever]

because

>>> float('nan')
nan
>>> float('inf')
inf
>>> float('-inf')
-inf

[sanand0]

@Sasha4ever - thanks. This raises a broader question:

• Should 1E6 be converted to a float or not? It's possible that it is an ID
• Should true always be converted to boolean?
• Should we allow True to also represent a boolean?

Because of the variety of possible conversions, you can pass an xml_fromstring that uses the conversion method you prefer. For example:

>>> from lxml import etree
>>> from xmljson import XMLData
>>> XMLData(xml_fromstring=str, attr_prefix=None, dict_type=dict).data(etree.XML('<ROOT test="NAN"/>'))
{'ROOT': {'test': 'NAN'}}

Hope that resolves the issue.

[Sasha4ever]

IMHO XMLData shouldn't return invalid XML.

[dagwieers]

@Sasha4ever I don't see from your comments that XMLData is returning invalid XML. Can you elaborate ?

Ah, I think you mean return invalid JSON. Yes, that requires a fix.

The question is what would you like for these values in the resulting JSON output. To me it's unclear whether we want to fail the conversion, make it null or keep it as a string. Any insight you want to share ?