class CSVFormat(FileFormat):
_name = "csv"
_description = _("CSV File")
_args = [
TextDelimiterArgument("fieldSeparator", "Field Separator"),
TextDelimiterArgument("recordSeparator", "Record Separator"),
TextDelimiterArgument("quoteChar", "Quote Character"),
CheckboxArgument("hasHeader", "Has Header")
]
_extensions = ["csv", "tsv"]
def __init__(self,
delimiter=',',
line_terminator='\n',
quote_char='"',
has_header=False,
sample="",
fields=None):
self._delimiter = delimiter
self._line_terminator = line_terminator
self._quote_char = quote_char
self._has_header = has_header
# sniffer insists on \r\n even when \n. This is safer and good enough for a preview
self._line_terminator = self._line_terminator.replace("\r\n", "\n")
self._sample_rows = self._get_sample_rows(sample)
self._num_columns = self._guess_num_columns(self._sample_rows)
self._fields = fields if fields else self._guess_fields(sample)
super(CSVFormat, self).__init__()
@staticmethod
def format_character(string):
# Morphline supports only one char representation
string = string.replace('"', '\\"')
string = string.replace('\t', '\\t')
string = string.replace('\n', '\\n')
string = string.replace('\u0001', '\\u0001')
string = string.replace('\x00', '\\u0000')
string = string.replace('\x01', '\\u0001')
string = string.replace('\x02', '\\u0002')
string = string.replace('\x03', '\\u0003')
return string
@classmethod
def _valid_character(self, char):
return isinstance(
char, basestring) and len(char) == 1 or char.startswith('\\')
@classmethod
def _guess_dialect(cls, sample):
sniffer = csv.Sniffer()
dialect = sniffer.sniff(sample)
has_header = sniffer.has_header(sample)
return dialect, has_header
@classmethod
def valid_format(cls, format_):
valid = super(CSVFormat, cls).valid_format(format_)
valid = valid and cls._valid_character(
cls.format_character(format_["fieldSeparator"]))
valid = valid and cls._valid_character(
cls.format_character(format_["recordSeparator"]))
valid = valid and cls._valid_character(
cls.format_character(format_["quoteChar"]))
valid = valid and isinstance(format_["hasHeader"], bool)
return valid
@classmethod
def _get_sample(cls, file_stream):
encoding = i18n.get_site_encoding()
for reader in [TextFileReader, GzipFileReader]:
file_stream.seek(0)
sample_data, sample_lines = reader.readlines(file_stream, encoding)
file_stream.seek(0)
if sample_data is not None:
yield sample_data, sample_lines
@classmethod
def _guess_from_file_stream(cls, file_stream):
for sample_data, sample_lines in cls._get_sample(file_stream):
try:
dialect, has_header = cls._guess_dialect(sample_data)
delimiter = dialect.delimiter
line_terminator = dialect.lineterminator
quote_char = dialect.quotechar
return cls(
**{
"delimiter": delimiter,
"line_terminator": line_terminator,
"quote_char": quote_char,
"has_header": has_header,
"sample": sample_data
})
except Exception:
LOG.exception('Warning, cannot read the file format.')
# Guess dialect failed, fall back to defaults:
return cls()
@classmethod
def _from_format(cls, file_stream, format_):
for sample_data, sample_lines in cls._get_sample(file_stream):
try:
delimiter = format_["fieldSeparator"].encode('utf-8')
line_terminator = format_["recordSeparator"].encode('utf-8')
quote_char = format_["quoteChar"].encode('utf-8')
has_header = format_["hasHeader"]
return cls(
**{
"delimiter": delimiter,
"line_terminator": line_terminator,
"quote_char": quote_char,
"has_header": has_header,
"sample": sample_data
})
except Exception:
LOG.exception('Warning, cannot read the file format.')
@classmethod
def get_instance(cls, file_stream, format_):
if cls.valid_format(format_):
return cls._from_format(file_stream, format_)
else:
return cls._guess_from_file_stream(file_stream)
@property
def sample(self):
return self._sample_rows
@property
def fields(self):
return self._fields
@property
def delimiter(self):
return self._delimiter
@property
def line_terminator(self):
return self._line_terminator
@property
def quote_char(self):
return self._quote_char
def get_format(self):
format_ = super(CSVFormat, self).get_format()
specific_format = {
"fieldSeparator": self.delimiter,
"recordSeparator": self.line_terminator,
"quoteChar": self.quote_char,
"hasHeader": self._has_header
}
format_.update(specific_format)
return format_
def _guess_num_columns(self, sample_rows):
counts = {}
for row in sample_rows:
num_columns = len(row)
if num_columns not in counts:
counts[num_columns] = 0
counts[num_columns] += 1
if counts:
num_columns_guess = max(counts.iteritems(),
key=operator.itemgetter(1))[0]
else:
num_columns_guess = 0
return num_columns_guess
def _guess_field_types(self, sample_rows):
field_type_guesses = []
num_columns = self._num_columns
for col in range(num_columns):
column_samples = [
sample_row[col] for sample_row in sample_rows
if len(sample_row) > col
]
field_type_guess = guess_field_type_from_samples(column_samples)
field_type_guesses.append(field_type_guess)
return field_type_guesses
def _get_sample_reader(self, sample):
if self.line_terminator != '\n':
sample = sample.replace('\n', '\\n')
return csv.reader(sample.split(self.line_terminator),
delimiter=self.delimiter,
quotechar=self.quote_char)
else:
return csv.reader(StringIO.StringIO(sample),
delimiter=self.delimiter,
quotechar=self.quote_char)
def _guess_field_names(self, sample):
reader = self._get_sample_reader(sample)
first_row = reader.next()
if self._has_header:
header = []
for i, field in enumerate(first_row):
header.append(field if field not in header else '%s_%s' %
(field, i))
else:
header = ["field_%d" % (i + 1) for i in range(self._num_columns)]
return header
def _get_sample_rows(self, sample):
NUM_SAMPLES = 5
header_offset = 1 if self._has_header else 0
reader = itertools.islice(self._get_sample_reader(sample),
header_offset, NUM_SAMPLES + 1)
sample_rows = list(reader)
return sample_rows
def _guess_fields(self, sample):
header = self._guess_field_names(sample)
types = self._guess_field_types(self._sample_rows)
if len(header) == len(types):
# create the fields
fields = [Field(header[i], types[i]) for i in range(len(header))]
else:
# likely failed to guess correctly
LOG.warn(
"Guess field types failed - number of headers didn't match number of predicted types."
)
fields = []
return fields
class CSVFormat(FileFormat):
_name = "csv"
_description = _("CSV File")
_args = [
TextDelimiterArgument("fieldSeparator", "Field Separator"),
TextDelimiterArgument("recordSeparator", "Record Separator"),
TextDelimiterArgument("quoteChar", "Quote Character"),
CheckboxArgument("hasHeader", "Has Header")
]
_extensions = ["csv", "tsv"]
def __init__(self,
delimiter=',',
line_terminator='\n',
quote_char='"',
has_header=False,
sample="",
fields=None):
self._delimiter = delimiter
self._line_terminator = line_terminator
self._quote_char = quote_char
self._has_header = has_header
# sniffer insists on \r\n even when \n. This is safer and good enough for a preview
self._line_terminator = self._line_terminator.replace("\r\n", "\n")
self._sample_rows = self._get_sample_rows(sample)
self._num_columns = self._guess_num_columns(self._sample_rows)
self._fields = fields if fields else self._guess_fields(sample)
super(CSVFormat, self).__init__()
@staticmethod
def format_character(string):
# Morphline supports only one char representation
string = string.replace('"', '\\"')
string = string.replace('\t', '\\t')
string = string.replace('\n', '\\n')
string = string.replace('\u0001', '\\u0001')
string = string.replace('\x00', '\\u0000')
string = string.replace('\x01', '\\u0001')
string = string.replace('\x02', '\\u0002')
string = string.replace('\x03', '\\u0003')
return string
@classmethod
def _valid_character(self, char):
return isinstance(
char, basestring) and len(char) == 1 or char.startswith('\\')
@classmethod
def _guess_dialect(cls, sample):
sniffer = csv.Sniffer()
dialect = sniffer.sniff(sample)
has_header = cls._hasHeader(sniffer, sample, dialect)
return dialect, has_header
# Copied from python2.7/csv.py with small modification to 1st line
# Results in large performance gain from not having to reprocess the file if dialect is known.
@classmethod
def _hasHeader(self, sniffer, sample, dialect):
# ******Changed from********
# rdr = reader(StringIO(sample), self.sniff(sample))
from _csv import reader
rdr = reader(string_io(sample), dialect)
header = next(rdr) # assume first row is header
columns = len(header)
columnTypes = {}
for i in range(columns):
columnTypes[i] = None
checked = 0
for row in rdr:
# arbitrary number of rows to check, to keep it sane
if checked > 20:
break
checked += 1
if len(row) != columns:
continue # skip rows that have irregular number of columns
for col in list(columnTypes.keys()):
for thisType in [int, long, float, complex]:
try:
thisType(row[col])
break
except (ValueError, OverflowError):
pass
else:
# fallback to length of string
thisType = len(row[col])
# treat longs as ints
if thisType == long:
thisType = int
if thisType != columnTypes[col]:
if columnTypes[col] is None: # add new column type
columnTypes[col] = thisType
else:
# type is inconsistent, remove column from
# consideration
del columnTypes[col]
# finally, compare results against first row and "vote"
# on whether it's a header
hasHeader = 0
for col, colType in list(columnTypes.items()):
if type(colType) == type(0): # it's a length
if len(header[col]) != colType:
hasHeader += 1
else:
hasHeader -= 1
else: # attempt typecast
try:
colType(header[col])
except (ValueError, TypeError):
hasHeader += 1
else:
hasHeader -= 1
return hasHeader > 0
@classmethod
def valid_format(cls, format_):
valid = super(CSVFormat, cls).valid_format(format_)
valid = valid and cls._valid_character(
cls.format_character(format_["fieldSeparator"]))
valid = valid and cls._valid_character(
cls.format_character(format_["recordSeparator"]))
valid = valid and cls._valid_character(
cls.format_character(format_["quoteChar"]))
valid = valid and isinstance(format_["hasHeader"], bool)
return valid
@classmethod
def _get_sample(cls, file_stream):
encoding = i18n.get_site_encoding()
for reader in [TextFileReader, GzipFileReader]:
file_stream.seek(0)
sample_data, sample_lines = reader.readlines(file_stream, encoding)
file_stream.seek(0)
if sample_data is not None:
yield sample_data, sample_lines
@classmethod
def _guess_from_file_stream(cls, file_stream):
for sample_data, sample_lines in cls._get_sample(file_stream):
try:
lines = itertools.islice(string_io(sample_data),
IMPORT_PEEK_NLINES)
sample_data_lines = ''
for line in lines:
sample_data_lines += line
dialect, has_header = cls._guess_dialect(
sample_data_lines
) # Only use first few lines for guessing. Greatly improves performance of CSV library.
delimiter = dialect.delimiter
line_terminator = dialect.lineterminator
quote_char = dialect.quotechar
return cls(
**{
"delimiter": delimiter,
"line_terminator": line_terminator,
"quote_char": quote_char,
"has_header": has_header,
"sample": sample_data
})
except Exception:
LOG.exception('Warning, cannot read the file format.')
# Guess dialect failed, fall back to defaults:
return cls()
@classmethod
def _from_format(cls, file_stream, format_):
for sample_data, sample_lines in cls._get_sample(file_stream):
try:
delimiter = format_["fieldSeparator"].encode('utf-8')
line_terminator = format_["recordSeparator"].encode('utf-8')
quote_char = format_["quoteChar"].encode('utf-8')
has_header = format_["hasHeader"]
return cls(
**{
"delimiter": delimiter,
"line_terminator": line_terminator,
"quote_char": quote_char,
"has_header": has_header,
"sample": sample_data
})
except Exception:
LOG.exception('Warning, cannot read the file format.')
@classmethod
def get_instance(cls, file_stream, format_):
if cls.valid_format(format_):
return cls._from_format(file_stream, format_)
else:
return cls._guess_from_file_stream(file_stream)
@property
def sample(self):
return self._sample_rows
@property
def fields(self):
return self._fields
@property
def delimiter(self):
return self._delimiter
@property
def line_terminator(self):
return self._line_terminator
@property
def quote_char(self):
return self._quote_char
def get_format(self):
format_ = super(CSVFormat, self).get_format()
specific_format = {
"fieldSeparator": self.delimiter,
"recordSeparator": self.line_terminator,
"quoteChar": self.quote_char,
"hasHeader": self._has_header
}
format_.update(specific_format)
return format_
def _guess_num_columns(self, sample_rows):
counts = {}
for row in sample_rows:
num_columns = len(row)
if num_columns not in counts:
counts[num_columns] = 0
counts[num_columns] += 1
if counts:
num_columns_guess = max(iter(counts.items()),
key=operator.itemgetter(1))[0]
else:
num_columns_guess = 0
return num_columns_guess
def _guess_field_types(self, sample_rows):
field_type_guesses = []
num_columns = self._num_columns
for col in range(num_columns):
column_samples = [
sample_row[col] for sample_row in sample_rows
if len(sample_row) > col
]
field_type_guess = guess_field_type_from_samples(column_samples)
field_type_guesses.append(field_type_guess)
return field_type_guesses
def _get_sample_reader(self, sample):
if self.line_terminator != '\n':
sample = sample.replace('\n', '\\n')
return csv.reader(sample.split(self.line_terminator),
delimiter=self.delimiter,
quotechar=self.quote_char)
else:
return csv.reader(string_io(sample),
delimiter=self.delimiter,
quotechar=self.quote_char)
def _guess_field_names(self, sample):
reader = self._get_sample_reader(sample)
first_row = next(reader)
if self._has_header:
header = []
for i, field in enumerate(first_row):
header.append(field if field not in header else '%s_%s' %
(field, i))
else:
header = ["field_%d" % (i + 1) for i in range(self._num_columns)]
return header
def _get_sample_rows(self, sample):
NUM_SAMPLES = 5
header_offset = 1 if self._has_header else 0
reader = itertools.islice(self._get_sample_reader(sample),
header_offset, NUM_SAMPLES + 1)
sample_rows = list(reader)
return sample_rows
def _guess_fields(self, sample):
header = self._guess_field_names(sample)
types = self._guess_field_types(self._sample_rows)
if len(header) == len(types):
# create the fields
fields = [Field(header[i], types[i]) for i in range(len(header))]
else:
# likely failed to guess correctly
LOG.warn(
"Guess field types failed - number of headers didn't match number of predicted types."
)
fields = []
return fields