Bokeh: GroupFilter only accepts strings

[mrocklin]

ALL software version info (bokeh, python, notebook, OS, browser, any other relevant packages)

bokeh 0.12.14

Description of expected behavior and the observed behavior

GroupFilter only operates on string-valued columns

Complete, minimal, self-contained example code that reproduces the issue

from bokeh.models import GroupFilter

GroupFilter(column_name='visible', group=True)

Traceback

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-30-b872a7a9ba79> in <module>()
      1 
----> 2 GroupFilter(column_name='visible', group=True)

~/anaconda/lib/python3.6/site-packages/bokeh/models/filters.py in __init__(self, *args, **kw)
     72             kw["group"] = args[1]
     73 
---> 74         super(GroupFilter, self).__init__(**kw)
     75 
     76 class CustomJSFilter(Filter):

~/anaconda/lib/python3.6/site-packages/bokeh/models/filters.py in __init__(self, *args, **kw)
     23             kw["filter"] = args[0]
     24 
---> 25         super(Filter, self).__init__(**kw)
     26 
     27 class IndexFilter(Filter):

~/anaconda/lib/python3.6/site-packages/bokeh/model.py in __init__(self, **kwargs)
    217         self._id = kwargs.pop("id", make_id())
    218         self._document = None
--> 219         super(Model, self).__init__(**kwargs)
    220         default_theme.apply_to_model(self)
    221 

~/anaconda/lib/python3.6/site-packages/bokeh/core/has_props.py in __init__(self, **properties)
    234 
    235         for name, value in properties.items():
--> 236             setattr(self, name, value)
    237 
    238     def __setattr__(self, name, value):

~/anaconda/lib/python3.6/site-packages/bokeh/core/has_props.py in __setattr__(self, name, value)
    261 
    262         if name in props or (descriptor is not None and descriptor.fset is not None):
--> 263             super(HasProps, self).__setattr__(name, value)
    264         else:
    265             matches, text = difflib.get_close_matches(name.lower(), props), "similar"

~/anaconda/lib/python3.6/site-packages/bokeh/core/property/descriptors.py in __set__(self, obj, value, setter)
    501             raise RuntimeError("%s.%s is a readonly property" % (obj.__class__.__name__, self.name))
    502 
--> 503         self._internal_set(obj, value, setter=setter)
    504 
    505     def __delete__(self, obj):

~/anaconda/lib/python3.6/site-packages/bokeh/core/property/descriptors.py in _internal_set(self, obj, value, hint, setter)
    722 
    723         '''
--> 724         value = self.property.prepare_value(obj, self.name, value)
    725 
    726         old = self.__get__(obj, obj.__class__)

~/anaconda/lib/python3.6/site-packages/bokeh/core/property/bases.py in prepare_value(self, obj_or_cls, name, value)
    279                     break
    280             else:
--> 281                 raise e
    282         else:
    283             value = self.transform(value)

~/anaconda/lib/python3.6/site-packages/bokeh/core/property/bases.py in prepare_value(self, obj_or_cls, name, value)
    272     def prepare_value(self, obj_or_cls, name, value):
    273         try:
--> 274             self.validate(value)
    275         except ValueError as e:
    276             for tp, converter in self.alternatives:

~/anaconda/lib/python3.6/site-packages/bokeh/core/property/bases.py in validate(self, value)
    403         if not (value is None or isinstance(value, self._underlying_type)):
    404             raise ValueError("expected a value of type %s, got %s of type %s" %
--> 405                 (nice_join([ cls.__name__ for cls in self._underlying_type ]), value, type(value).__name__))
    406 
    407     def from_json(self, json, models=None):

ValueError: expected a value of type str, got True of type bool

[mrocklin]

Obviously I can get around this by mapping my data to strings, but I'm sending a fair amount of it, and wouldn't mind using more compact datatypes

[bryevdv]

cc @clairetang6 thoughts?

[clairetang6]

I don't see any reason why the group parameter couldn't be a boolean or number. The JS implementation just checks whether the values of the column are equal (===) to the group parameter. I think this would be an easy addition, just changing the types that are allowed for group on the python and JS sides.

[chesucr]

In my case I would need to group by a numeric values as well. Althoug I found an alternative mapping the values and adding them to the BooleanFilter:

boolean_filter = BooleanFilter(
    booleans=[True if flag == 5 else False for flag in source.data['flag']] 
)
view = CDSView(source=source, filters=[boolean_filter])

But I reckon this is less efficient