Template.analyze's representation of variables is lossy

[brian-vanderford]

>>> from liquid import Environment
>>> env = Environment()      
>>> env.from_string("{{thing['foo.bar']}}").analyze().variables # property "foo.bar" of thing
{'thing.foo.bar': [('<string>', 1)]}
>>> env.from_string("{{thing.foo.bar]}}").analyze().variables   # property bar of property foo of thing
{'thing.foo.bar': [('<string>', 1)]}

thing['foo.bar'] and thing.foo.bar are distinct, but end up with the same representation. It would be helpful to be able to get a more structured representation (maybe with a keyword arg to analyze) so that thing['foo.bar'] is something like ('thing', 'foo.bar') and thing.foo.bar ('thing', 'foo', 'bar').

[jg-rp]

Hi @brian-vanderford,

The string representation of variables has been fixed in version 1.6.0, and those strings now include a parts property, being a tuple representation of the variable. No extra arguments to analyze() required.

from liquid import Environment

env = Environment()
template = env.from_string("{{ thing.foo.bar }} {{ thing['foo.bar'] }}")

for var, location in template.analyze().variables.items():
    for template_name, line_number in location:
        print(f"{var} {var.parts} found in '{template_name}' on line {line_number}")

output

thing.foo.bar ('thing', 'foo', 'bar') found in '<string>' on line 1
thing["foo.bar"] ('thing', 'foo.bar') found in '<string>' on line 1