Handling of Empty and None values in dtype.py lead to load errors

[mpsonntag]

The value handling of lists of "empty" values in dtypes.py lead to different results and
potentially broken odML documents:

• int and float replace None by default values. Can be saved and loaded.

  prop = odml.Property(name="int", value=[""], dtype="int")
  leads to prop.value ... [0]

  prop = odml.Property(name="float", value=[None, None], dtype="float")
  leads to prop.value ... [0.0, 0.0]

• string related dtypes (string, text, url, person) replace None by "None". Can be saved and loaded.

  prop = odml.Property(name="str", value=["", None], dtype="string")
  leads to prop.value ... ['', 'None']

• bool, date, time and datetime save None to file but leads to an Error that is hard to identify when the file is loaded.

  prop = odml.Property(name="date", value=[None], dtype="date")
  leads on document load to ValueError: odml.Property.value: passed values are not of consistent type!

[mpsonntag]

Should we consistently

• return the defined dtype default value if an assigned list value is "" or None?
  or
• handle the currently broken dtypes case by case - since e.g. returning the current date for every "" or None entry might be annoying as well.

[lzehl]

hi! I would vote for the first solution: always return the default dtype for "" or None

[jgrewe]

I agree, return default value. Or, we could return None for any non-existing value

[mpsonntag]

• If we return the default value, for date, time and datetime this will always be "now", for bool it will be False.
• If we return None we need to handle this differently on load, currently it leads to an error.

My vote is to always return the default value.

[jgrewe]

ok, let's do it