v0.0.12

User Facing Changes

• Arrange correctly resets index. #106
• Filter correctly resets index. #130
• Switch to MIT license
• SqlTranslator is now mutable, so it's easier to tweak them on the fly (e.g. in a notebook. b9c81d7
• pandas join feats / fixes
  • the implementation of semi_join was duplicating rows as standard joins do.
    this resulted in some very large results.... #115
  • implement anti_join
  • fix failures when the on arg received a mapping
  • fix full_join failing for pandas, since pandas calls it an 'outer' join
• implement sql transmute. #108
• clean up, document, add docstring tests for vector funcs and forcats. #120
• add full tests for forcats (thanks @bakera81!) .#121
• clean up and document separate(). #119
• allow raw SQL. #123
• generalize spread and gather. #129
• mutate generates fewer subqueries now. #127

As an experimental feature, I shortened the stacktraces for SQL translator errors! #125

Chores

• clean up pytest. #116

count without args returns total rows

Small fix, supporting count without args. This is a very common case against SQL dbs, since it lets you know how big (how many rows) a table has.

e.g.

tbl_something >> count()

add gather, improve case when support

• all case_when values to be Calls (e.g. case_when({True: _.a + 1}) #103
• add gather function
• add developer docs

Support connections to redshift

• Allow custom specification of table columns (#78) - this was preventing people from working with SQL when SqlAlchemy couldn't reflect column names from a database.
• Add some intro docs

document core 1 table verbs, and add fct_rev

v0.0.8

Update __init__.py

SQL support and unit tests

This release implements extensive testing for postgres and sqlite. It also sets up (but skips) pandas unit tests.

It follows from this PR: #36

SQL Improvements:

• fix distinct when following an arrange (#65)
• move collect and show_query into dply.verbs (#23)
• mutate now doesn't put redefined columns at end of table (#42)
• full join gets all values in joining columns, even if unique to only 1 table (#57)
• complete implementation of joins, including anti_ and semi_join (#58, #55, #53)
• allow mutate to accept only scalar values, e.g. mutate(a = 1). (#39)
• handle arbitrary siu expressions inside arrange (#30)
• track arrange expressions, in order to allow cumulative functions (#30)
• clearer messaging that summarize can't refer to just defined column (#49)
• summarize, and mutate better identify when referencing a variable requires a CTE (#46, #45)
• allow keyword args to group_by (#52)