def convert_header_with_schema(self, header):
self.types = [None] * self.column_count # Value type of every column.
for idx, field in enumerate(header):
pair = field.split(':')
# Multiple colons found in column name, emit error.
# TODO might need to check for backtick escapes
if len(pair) > 2:
raise CSVError("%s: Field '%s' had %d colons" %
(self.infile.name, field, len(field)))
# Convert the column type.
col_type = convert_schema_type(pair[1].upper().strip())
# If the column did not have a name but the type requires one, emit an error.
if len(pair[0]) == 0 and col_type not in (Type.ID, Type.START_ID,
Type.END_ID,
Type.IGNORE):
raise SchemaError(
"%s: Each property in the header should be a colon-separated pair"
% (self.infile.name))
else:
# We have a column name and a type.
# Only store the name if the column's values should be added as properties.
if len(pair[0]) > 0 and col_type not in (Type.START_ID,
Type.END_ID,
Type.IGNORE):
column_name = pair[0].strip()
self.column_names[idx] = column_name
# Store the column type.
self.types[idx] = col_type
def validate_row(self, row):
# Each row should have the same number of fields
if len(row) != self.column_count:
raise CSVError(
"%s:%d Expected %d columns, encountered %d ('%s')" %
(self.infile.name, self.reader.line_num, self.column_count,
len(row), self.config.separator.join(row)))
def process_schemaless_header(self, header):
if self.column_count < 2:
raise CSVError(
"Relation file '%s' should have at least 2 elements in header line."
% (self.infile.name))
# The first column is the source ID and the second is the destination ID.
self.start_id = 0
self.end_id = 1
self.start_namespace = None
self.end_namespace = None
for idx, field in enumerate(header[2:]):
self.column_names[idx + 2] = field