def test_query_field_from_path():
"""
Queries with SelectionOps that reference only fields
parsed from the directory structre will rewrite
the query so that the file list is filtered without
opening/decoding the files.
"""
adapter = DirAdapter(employees=dict(root_dir="/",
pattern="{department}",
filename_column="path",
decode="auto",
schema=TEST_SCHEMA))
op = SelectionOp(LoadOp('employees'),
EqOp(Var('department'), Const('sales')))
loc = query_zipper(op).leftmost_descendant()
res = adapter.evaluate(loc)
relation = adapter.get_relation('employees')
compare(
res.root(),
Function(
'decode',
SelectionOp(
Function('extract_path',
Function('files', Const(relation.root_dir)),
Const(relation.root_dir + "{department}")),
EqOp(Var('department'), Const('sales'))), Const('auto'),
Const(TEST_SCHEMA), Const('path')))
def test_query_field_in_payload():
"""
Querying a field inside the payload should result
in the LoadOp being rewritten as
SelectionOp(Function('decode', Function('extract_path', Function('files'))))
"""
adapter = DirAdapter(employees=dict(root_dir="/",
pattern="{department}",
filename_column="path",
decode="auto",
schema=TEST_SCHEMA))
op = SelectionOp(LoadOp('employees'), GeOp(Var('salary'), Const(40000)))
loc = query_zipper(op).leftmost_descendant()
res = adapter.evaluate(loc)
relation = adapter.get_relation('employees')
compare(
res.root(),
SelectionOp(
Function(
'decode',
Function('extract_path',
Function('files', Const(relation.root_dir)),
Const(relation.root_dir + "{department}")),
Const('auto'), Const(TEST_SCHEMA), Const('path')),
GeOp(Var('salary'), Const(40000))))
def compare(op1, op2):
loc1 = query_zipper(op1).leftmost_descendant()
loc2 = query_zipper(op2).leftmost_descendant()
while True:
n1 = loc1.node()
n2 = loc2.node()
if not (compare_relation(n1,n2) or n1 == n2):
raise NodeDiffException(n1,n2)
if any((loc1.at_end(), loc2.at_end())):
# if either is at the end, they both should be
assert loc1.at_end() == loc2.at_end()
break
else:
loc1 = loc1.postorder_next()
loc2 = loc2.postorder_next()
def test_query_field_from_path_and_contents():
"""
Queries with SelectionOps that reference both fields
parsed from the directory structre and content will
rewrite the query so that the file list is filtered before
opening/decoding the files and finally filtered by the field
from the content
"""
adapter = DirAdapter(
employees=dict(root_dir="/",
pattern="{department}",
filename_column="path",
decode="auto",
schema=dict(fields=[
dict(type='STRING', name='department'),
dict(type='INTEGER', name='id'),
dict(type='STRING', name='full_name'),
dict(type='INTEGER', name='salary'),
dict(type='INTEGER', name='manager_id'),
])))
op = SelectionOp(
LoadOp('employees'),
And(
EqOp(Var('department'), Const('sales')),
GeOp(Var('salary'), Const(40000)),
))
loc = query_zipper(op).leftmost_descendant()
res = adapter.evaluate(loc)
relation = adapter.get_relation('employees')
compare(
res.root(),
SelectionOp(
Function(
'decode',
SelectionOp(
Function('extract_path',
Function('files', Const(relation.root_dir)),
Const(relation.root_dir + "{department}")),
EqOp(Var('department'), Const('sales'))), Const('auto'),
Const(TEST_SCHEMA), Const('path')),
GeOp(Var('salary'), Const(40000))))
def test_evaluate():
adapter = DirAdapter(songs=dict(
root_dir=path,
pattern="{artist}/{album}/{track}.{ext}",
filename_column="path",
))
relation = adapter.get_relation('songs')
op = LoadOp('songs')
loc = query_zipper(op).leftmost_descendant()
res = adapter.evaluate(loc)
compare(
res.root(),
Function('extract_path', Function('files', Const(relation.root_dir)),
Const(path + "/{artist}/{album}/{track}.{ext}")))
def test_evaluate():
adapter = S3Adapter(
logs = dict(
bucket = "aws-publicdatasets",
anon = True,
prefix = "/common-crawl/",
pattern = "{timestamp}/{server}"
)
)
relation = adapter.get_relation('logs')
op = LoadOp('logs')
loc = query_zipper(op).leftmost_descendant()
res = adapter.evaluate(loc)
import pdb; pdb.set_trace()
eq_(
res.root(),
Function('s3_keys', Const(relation.bucket_name), Const(relation.prefix))
)
