BUG/CLN: Fix predicates on Selections on Joins

The motivation for this PR is twofold    1) fix a bug where filters on
joined data would either fail or give the wrong result, depending on
which version of pandas you're on  2) refactor the
`execute_selection_dataframe` function because it was getting rather
large and unreadable.    The bug in ibis-project#1136 was occurring due to the
fact that the predicates in the `Selection` were not being evaluated
against the `data` argument  which is the result of the joined data.
This doesn't fail on non-`Join` selections because there's no possible
ambiguity regarding which column to select. The solution is to map
each root table in each predicate to `data`, therefore evaluating the
predicate against the joined data.    Closes ibis-project#1136.

Author: Phillip Cloud <cpcloud@gmail.com>

Closes ibis-project#1149 from cpcloud/fix-selection-join-predicates and squashes the following commits:

d76b2a8 [Phillip Cloud] BUG/CLN: Fix predicates on Selections on Joins

ibis/expr/api.py

@@ -1998,6 +1998,7 @@ def between_time(arg, lower, upper, timezone=None):
     'any_inner': _ops.AnyInnerJoin,
     'any_left': _ops.AnyLeftJoin,
     'outer': _ops.OuterJoin,
+    'right': _ops.RightJoin,
     'left_semi': _ops.LeftSemiJoin,
     'semi': _ops.LeftSemiJoin,
     'anti': _ops.LeftAntiJoin,
@@ -2019,6 +2020,7 @@ def join(left, right, predicates=(), how='inner'):
       - 'inner': inner join
       - 'left': left join
       - 'outer': full outer join
+      - 'right': right outer join
       - 'semi' or 'left_semi': left semi join
       - 'anti': anti join
 

ibis/pandas/core.py

@@ -13,7 +13,7 @@
 import ibis.expr.types as ir
 import ibis.expr.datatypes as dt
 
-import ibis.pandas.context as ctx
+import ibis.pandas.aggcontext as agg_ctx
 from ibis.pandas.dispatch import (
     execute, execute_node, execute_first, data_preload
 )
@@ -92,7 +92,7 @@ def execute_with_scope(expr, scope, context=None, **kwargs):
         return scope[op]
 
     if context is None:
-        context = ctx.Summarize()
+        context = agg_ctx.Summarize()
 
     try:
         computed_args = [scope[t] for t in op.root_tables()]
@@ -175,6 +175,6 @@ def execute_without_scope(
     return execute(
         expr,
         new_scope,
-        context=context if context is not None else ctx.Summarize(),
+        context=context if context is not None else agg_ctx.Summarize(),
         **kwargs
     )

ibis/pandas/execution/__init__.py

@@ -0,0 +1,4 @@
+from ibis.pandas.execution.generic import *  # noqa: F401,F403
+from ibis.pandas.execution.selection import *  # noqa: F401,F403
+from ibis.pandas.execution.join import *  # noqa: F401,F403
+from ibis.pandas.execution.window import *  # noqa: F401,F403

ibis/pandas/execution/constants.py

@@ -0,0 +1,74 @@
+"""Constants for the pandas backend.
+"""
+
+import operator
+
+import numpy as np
+import pandas as pd
+
+import ibis
+import ibis.expr.datatypes as dt
+import ibis.expr.operations as ops
+
+
+JOIN_TYPES = {
+    ops.LeftJoin: 'left',
+    ops.RightJoin: 'right',
+    ops.InnerJoin: 'inner',
+    ops.OuterJoin: 'outer',
+}
+
+
+LEFT_JOIN_SUFFIX = '_ibis_left_{}'.format(ibis.util.guid())
+RIGHT_JOIN_SUFFIX = '_ibis_right_{}'.format(ibis.util.guid())
+JOIN_SUFFIXES = LEFT_JOIN_SUFFIX, RIGHT_JOIN_SUFFIX
+ALTERNATE_SUFFIXES = {
+    LEFT_JOIN_SUFFIX: RIGHT_JOIN_SUFFIX,
+    RIGHT_JOIN_SUFFIX: LEFT_JOIN_SUFFIX,
+}
+
+
+IBIS_TYPE_TO_PANDAS_TYPE = {
+    dt.float: np.float32,
+    dt.double: np.float64,
+    dt.int8: np.int8,
+    dt.int16: np.int16,
+    dt.int32: np.int32,
+    dt.int64: np.int64,
+    dt.string: str,
+    dt.timestamp: 'datetime64[ns]',
+}
+
+
+IBIS_TO_PYTHON_LITERAL_TYPES = {
+    dt.double: float,
+    dt.float: float,
+    dt.int64: int,
+    dt.int32: int,
+    dt.int16: int,
+    dt.int8: int,
+    dt.string: str,
+    dt.date: lambda x: pd.Timestamp(x).to_pydatetime().date(),
+}
+
+
+BINARY_OPERATIONS = {
+    ops.Greater: operator.gt,
+    ops.Less: operator.lt,
+    ops.LessEqual: operator.le,
+    ops.GreaterEqual: operator.ge,
+    ops.Equals: operator.eq,
+    ops.NotEquals: operator.ne,
+
+    ops.And: operator.and_,
+    ops.Or: operator.or_,
+    ops.Xor: operator.xor,
+
+    ops.Add: operator.add,
+    ops.Subtract: operator.sub,
+    ops.Multiply: operator.mul,
+    ops.Divide: operator.truediv,
+    ops.FloorDivide: operator.floordiv,
+    ops.Modulus: operator.mod,
+    ops.Power: operator.pow,
+}