Jaidev Deshpande

A data `source` can also be a sputnik package (http://github.com/spacy-io/sputnik)

Some columns in a table might be encrypted. Add support to decrypt them.

Add tests for the MySQL functionality. Explore both real MySQL databases and Mock.

Specifying delimiters in the schema should properly infer parsers.

Before adding support for remote files, the Project class and the validators need to be refactored. They are doing a lot of things. Refactor them based on how they handle...

Just as the `pandas.read_*` signatures are inferred from the schema, `DataFrame.to_*` signatures should also be readable from the schema.

- [ ] Given a file, generating a minimal, safe schema should be possible. - [ ] Given a dataframe, generate a schema such that the dataframe can be reproduced.

It would help to have a motivation section in the docs, written with the assumption that the reader doesn't know pandas or even Python.

Most of the traits interface needs to be heavily inspected and if necessary, rewritten. Especially need to examine the need for custom traits, now that excepts in traits are understood....