Rc/0.3.0

[jayckaiser]

Release Candidate 0.3.0

Description & motivation

This is the next major release candidate of Earthmover with a series of features, bugfixes, and behind-the-scenes refactors.

PR Merge Priority:

• [ ] Low
• [ ] Medium
• [x] High

We will merge this as soon as possible to keep with our internal deadline.

Changes

• feature: add project composition using packages keyword in template file (see README)
• feature: add installation extras for optional libraries, and improve error logging to notify which is missing
• feature: GroupByWithRankOperation cumulatively sums record counts by group-by columns
• feature: setting log_level: DEBUG in template configs or setting debug: True for a node displays the head of the node mid-run
• feature: add optional_fields key to all Sources to add optional empty columns when missing from schema
• feature: add optional ignore_errors and exact_match boolean flags to DateFormatOperation
• internal: force-cast a dataframe to string-type before writing as a Destination
• internal: remove attempted directory-hashing when a source is a directory (i.e., Parquet)
• internal: refactor project to standardize import paths for Node and Operation
• internal: add Node.full_name attribute and Node.set_upstream_source() method
• internal: unify graph-building into compilation
• internal: refactor compilation and execution code for cleanliness
• internal: unify Node.compile() into initialization to ease Node development
• internal: Remove unused group_by_with_count and group_by_with_agg operations

Tests and QC done:

All Earthmover tests have run successfully, and the code has been used for SC assessment loading.