Serialisation error (with dicts derived from DataFrame)

Anvil Q&A
1

What I’m trying to do:
I have a Pandas dataframe, generated from a crosstab.

ct = pd.crosstab(
    [df.verwijzing_type],
    [df.verwijzing_jaar],
    margins=True)

I wanted to get this table to the client. Usually I would do this with ct.to_dict(‘records’) and usually works fine. The output of ct.to_dict(‘records’) is something like:

[{2012.0: 2896,
  2013.0: 2825,
  2014.0: 3179,
  2015.0: 3638,
  2016.0: 3670,
  2017.0: 3956,
  2018.0: 3117,
  2019.0: 3265,
  2020.0: 3170,
  2021.0: 1120,
  'All': 30836},
 {2012.0: 1000,
  2013.0: 1182,
  2014.0: 896,
  2015.0: 846,
  2016.0: 956,
  2017.0: 1076,
  2018.0: 1135,
  2019.0: 1124,
  2020.0: 1025,
  2021.0: 330,
  'All': 9570},
:
:

This yields two problems:

First a very vague error (at first):

TypeError: '<' not supported between instances of 'str' and 'float'

Turns out that the column 'All", added by the crosstab margins=True, confuses Anvil in the serialisation step. This is related to the next problem.
By setting margins=False I removed the All column.
Then we get the error:

anvil.server.SerializationError: Cannot serialize return value from function. Cannot serialize dictionaries with keys that aren't strings at ...['crosstab_type_year'][0][2012.0]

In my crosstab the columns are years and are named as such. These turn out to be are floats: 2018.0, 2019.0, 2020.0 etc. Perfectly fine in python/Pandas.

But it seems Anvil gets confused.
I can probably work around this problem, but is there any reason for Anvil not to fully support python dicts?

2

If I recall correctly, it’s mentioned elsewhere (on this forum) that Anvil can serialize a dict only when all of its keys are strings. The problem case was a lot like yours, where one or more keys were numbers. The workaround was to convert all keys into numbers.