def test_stress_block_sizes(self):
# Test a number of small block sizes to stress block stitching
data_base, expected = make_random_json(num_cols=2, num_rows=100)
read_options = ReadOptions()
parse_options = ParseOptions()
for data in [data_base, data_base.rstrip(b'\r\n')]:
for newlines_in_values in [False, True]:
parse_options.newlines_in_values = newlines_in_values
for block_size in [22, 23, 37]:
read_options.block_size = block_size
table = self.read_bytes(data, read_options=read_options,
parse_options=parse_options)
assert table.schema == expected.schema
if not table.equals(expected):
# Better error output
assert table.to_pydict() == expected.to_pydict()
def test_block_sizes(self):
rows = b'{"a": 1}\n{"a": 2}\n{"a": 3}'
read_options = ReadOptions()
parse_options = ParseOptions()
for data in [rows, rows + b'\n']:
for newlines_in_values in [False, True]:
parse_options.newlines_in_values = newlines_in_values
read_options.block_size = 4
with pytest.raises(ValueError,
match="try to increase block size"):
self.read_bytes(data, read_options=read_options,
parse_options=parse_options)
# Validate reader behavior with various block sizes.
# There used to be bugs in this area.
for block_size in range(9, 20):
read_options.block_size = block_size
table = self.read_bytes(data, read_options=read_options,
parse_options=parse_options)
assert table.to_pydict() == {'a': [1, 2, 3]}
def test_reconcile_accross_blocks(self):
# ARROW-12065: reconciling inferred types accross blocks
first_row = b'{ }\n'
read_options = ReadOptions(block_size=len(first_row))
for next_rows, expected_pylist in [
(b'{"a": 0}', [None, 0]),
(b'{"a": []}', [None, []]),
(b'{"a": []}\n{"a": [[1]]}', [None, [], [[1]]]),
(b'{"a": {}}', [None, {}]),
(b'{"a": {}}\n{"a": {"b": {"c": 1}}}',
[None, {"b": None}, {"b": {"c": 1}}]),
]:
table = self.read_bytes(first_row + next_rows,
read_options=read_options)
expected = {"a": expected_pylist}
assert table.to_pydict() == expected
# Check that the issue was exercised
assert table.column("a").num_chunks > 1
def read_json(self, *args, **kwargs):
read_options = kwargs.setdefault('read_options', ReadOptions())
read_options.use_threads = True
table = read_json(*args, **kwargs)
table.validate()
return table
