zgbjgg / jun Goto Github PK

Not operator != is not working with legacy assignment since str are evaluated directly to ==

Support for python 3 since python 2.7 lifecycle is ending.

In order to support PY3 we need this changes:

• All atoms now will be used as binaries since mapping data types is translated to bytes() and atoms are limited to 255 chars.
• Check a new decoder to change bytes() to str() from python side
• Allow worker start with py3, py2 or custom py path
• Fix tests to work with new implementation
• New version of JUN will be published as 0.2.0
• Support to manage engines: pandas or koalas, since migrating to py3 we are able to use koalas as engine to improve performance of JUN tasks.

If new tasks appears in the process of update the repo it will be added here. All work done in this release should be put in py3 branch.

Allow usage of binaries (as atoms) in order to avoid errors when using large atoms, so it will use binaries instead of.

This is related to issue in Kaa: zgbjgg/kaa#2

When return values from PY to Erl check for NaN, maybe eval using isnull from pandas and then convert nan to num

Check if can change opaque term representing dataframes or series to a single key in the python process, so dataframe will live only in python side.

Running large datasets maybe will cause invalid_protocol_data but I'm not sure about, however this will be a great improve.