def __getitem__(self, obj):
"""Expression evaluation against groups.
Given a single object or list, the GroupBy will interpret it as a set
of SELECT expressions to evaluate in the context of the GROUP BY.
Given a tuple of length 2, the first element will be interpreted for
group selection (i.e., a HAVING clause), while the second element will
be interpreted as a set of expressions to evaluate against the groups.
"""
ast = copy(self._grouped_ast)
if isinstance(obj, tuple) and len(obj) == 2:
if not isinstance(obj[0], Expr):
raise ValueError("The group filter (obj[0]) must be Expr type")
ast._having = obj[0]
obj = obj[1]
# obj is now the SELECT portion
if not isinstance(obj, (list, tuple)):
obj = [obj]
select_list = []
for elt in obj:
if isinstance(elt, SelectItem):
select_list.append(elt)
elif isinstance(elt, basestring):
select_list.append(SelectItem(expr=Literal(elt)))
elif isinstance(elt, Expr):
select_list.append(SelectItem(expr=elt))
ast._select_list = select_list
return BigDataFrame(self._ic, ast)
def from_sql_table(ic, table):
"""Create a BDF from a table name usable in Impala"""
table_name = _to_TableName(table)
table_ref = BaseTableRef(table_name)
schema = _get_table_schema_hack(ic._cursor, table_ref.to_sql())
select_list = tuple(
[SelectItem(expr=Literal(col)) for (col, ty) in schema])
return BigDataFrame(ic, SelectStmt(select_list, table_ref))
def from_sql_query(ic, query, alias=None):
"""Create a BDF from a SQL query executed by Impala"""
query_alias = alias if alias else _random_id('inline_', 4)
table_ref = InlineView(query, query_alias)
schema = _get_table_schema_hack(ic._cursor, table_ref.to_sql())
select_list = tuple(
[SelectItem(expr=Literal(col)) for (col, ty) in schema])
return BigDataFrame(ic, SelectStmt(select_list, table_ref))
def take(self, n):
"""Return `n` rows as a pandas `DataFrame`
Distributed and no notion of order, so not guaranteed to be
reproducible.
"""
alias = _random_id('inline_', 4)
table_ref = InlineView(self._query_ast.to_sql(), alias)
# SELECT alias.*
select_list = [SelectItem(table_name=TableName(table_ref.name))]
limit_elt = LimitElement(Literal(n), None)
ast = SelectStmt(select_list, table_ref, limit=limit_elt)
bdf = BigDataFrame(self._ic, ast)
return as_pandas(bdf.__iter__())
def group_by(self, by):
"""Group the BDF
`by` is `string`, `Expr`, or `list/tuple[string/Expr]`
"""
if not isinstance(by, (tuple, list)):
by = (by, )
if not all([isinstance(e, (basestring, Expr)) for e in by]):
raise ValueError("must supply only strings or Exprs")
by = tuple([e if isinstance(e, Expr) else Literal(e) for e in by])
table_ref = InlineView(self._query_ast.to_sql(), 'inner_tbl')
# invalid AST; to be used by GroupBy
incomplete_ast = SelectStmt([], table_ref, group_by=by)
return GroupBy(self._ic, incomplete_ast)