| 1 |
1388
|
aaronmk
|
# CSV I/O
|
| 2 |
|
|
|
| 3 |
|
|
import csv
|
| 4 |
|
|
import StringIO
|
| 5 |
|
|
|
| 6 |
1623
|
aaronmk
|
import strings
|
| 7 |
1444
|
aaronmk
|
import util
|
| 8 |
|
|
|
| 9 |
4211
|
aaronmk
|
delims = ',;\t`'
|
| 10 |
1623
|
aaronmk
|
tsv_delim = '\t'
|
| 11 |
|
|
escape = '\\'
|
| 12 |
1442
|
aaronmk
|
|
| 13 |
1623
|
aaronmk
|
ending_placeholder = r'\n'
|
| 14 |
|
|
|
| 15 |
|
|
def is_tsv(dialect): return dialect.delimiter == tsv_delim
|
| 16 |
|
|
|
| 17 |
1621
|
aaronmk
|
def sniff(line):
|
| 18 |
|
|
'''Automatically detects the dialect'''
|
| 19 |
1623
|
aaronmk
|
dialect = csv.Sniffer().sniff(line, delims)
|
| 20 |
1621
|
aaronmk
|
# TSVs usually don't quote fields (nor doublequote embedded quotes)
|
| 21 |
1623
|
aaronmk
|
if is_tsv(dialect): dialect.quoting = csv.QUOTE_NONE
|
| 22 |
1621
|
aaronmk
|
else: dialect.doublequote = True # Sniffer doesn't turn this on by default
|
| 23 |
|
|
return dialect
|
| 24 |
|
|
|
| 25 |
1923
|
aaronmk
|
def stream_info(stream, parse_header=False):
|
| 26 |
1444
|
aaronmk
|
'''Automatically detects the dialect based on the header line
|
| 27 |
1923
|
aaronmk
|
@return NamedTuple {header_line, header, dialect}'''
|
| 28 |
1444
|
aaronmk
|
info = util.NamedTuple()
|
| 29 |
|
|
info.header_line = stream.readline()
|
| 30 |
1923
|
aaronmk
|
info.header = None
|
| 31 |
|
|
if info.header_line != '':
|
| 32 |
|
|
info.dialect = sniff(info.header_line)
|
| 33 |
|
|
if parse_header:
|
| 34 |
|
|
info.header = reader_class(info.dialect)(
|
| 35 |
|
|
StringIO.StringIO(info.header_line), info.dialect).next()
|
| 36 |
1660
|
aaronmk
|
else: info.dialect = None # line of '' indicates EOF = empty stream
|
| 37 |
1444
|
aaronmk
|
return info
|
| 38 |
|
|
|
| 39 |
5146
|
aaronmk
|
tsv_encode_map = [
|
| 40 |
|
|
('\t', r'\t'),
|
| 41 |
|
|
('\n', '\\\n'),
|
| 42 |
|
|
]
|
| 43 |
|
|
|
| 44 |
1623
|
aaronmk
|
class TsvReader:
|
| 45 |
|
|
'''Unlike csv.reader, for TSVs, interprets \ as escaping a line ending but
|
| 46 |
5145
|
aaronmk
|
ignores it before everything else (e.g. \N for NULL).
|
| 47 |
|
|
Also interprets '\n' as a newline and '\t' as a tab.
|
| 48 |
|
|
'''
|
| 49 |
1623
|
aaronmk
|
def __init__(self, stream, dialect):
|
| 50 |
|
|
assert is_tsv(dialect)
|
| 51 |
|
|
self.stream = stream
|
| 52 |
|
|
self.dialect = dialect
|
| 53 |
|
|
|
| 54 |
|
|
def __iter__(self): return self
|
| 55 |
|
|
|
| 56 |
|
|
def next(self):
|
| 57 |
|
|
record = ''
|
| 58 |
|
|
ending = None
|
| 59 |
|
|
while True:
|
| 60 |
|
|
line = self.stream.readline()
|
| 61 |
|
|
if line == '': raise StopIteration
|
| 62 |
|
|
|
| 63 |
|
|
raw_contents, ending = strings.extract_line_ending(line)
|
| 64 |
|
|
contents = strings.remove_suffix(escape, raw_contents)
|
| 65 |
|
|
record += contents
|
| 66 |
|
|
if len(contents) == len(raw_contents): break # no line continuation
|
| 67 |
|
|
record += ending_placeholder
|
| 68 |
3055
|
aaronmk
|
|
| 69 |
|
|
# Prevent "new-line character seen in unquoted field" errors
|
| 70 |
|
|
record = record.replace('\r', ending_placeholder)
|
| 71 |
|
|
|
| 72 |
1623
|
aaronmk
|
row = csv.reader(StringIO.StringIO(record), self.dialect).next()
|
| 73 |
5145
|
aaronmk
|
return [v.replace(r'\n', '\n').replace(r'\t', '\t') for v in row]
|
| 74 |
1623
|
aaronmk
|
|
| 75 |
|
|
def reader_class(dialect):
|
| 76 |
|
|
if is_tsv(dialect): return TsvReader
|
| 77 |
|
|
else: return csv.reader
|
| 78 |
|
|
|
| 79 |
|
|
def make_reader(stream, dialect): return reader_class(dialect)(stream, dialect)
|
| 80 |
|
|
|
| 81 |
1388
|
aaronmk
|
def reader_and_header(stream):
|
| 82 |
|
|
'''Automatically detects the dialect based on the header line
|
| 83 |
|
|
@return tuple (reader, header)'''
|
| 84 |
1923
|
aaronmk
|
info = stream_info(stream, parse_header=True)
|
| 85 |
1958
|
aaronmk
|
return (make_reader(stream, info.dialect), info.header)
|
| 86 |
1446
|
aaronmk
|
|
| 87 |
|
|
##### csv modifications
|
| 88 |
|
|
|
| 89 |
|
|
# Note that these methods only work on *instances* of Dialect classes
|
| 90 |
|
|
csv.Dialect.__eq__ = lambda self, other: self.__dict__ == other.__dict__
|
| 91 |
|
|
csv.Dialect.__ne__ = lambda self, other: not (self == other)
|
| 92 |
2114
|
aaronmk
|
|
| 93 |
|
|
##### Row filters
|
| 94 |
|
|
|
| 95 |
|
|
class Filter:
|
| 96 |
|
|
'''Wraps a reader, filtering each row'''
|
| 97 |
|
|
def __init__(self, filter_, reader):
|
| 98 |
|
|
self.reader = reader
|
| 99 |
|
|
self.filter = filter_
|
| 100 |
|
|
|
| 101 |
|
|
def __iter__(self): return self
|
| 102 |
|
|
|
| 103 |
|
|
def next(self): return self.filter(self.reader.next())
|
| 104 |
|
|
|
| 105 |
|
|
std_nulls = [r'\N']
|
| 106 |
|
|
empty_nulls = [''] + std_nulls
|
| 107 |
|
|
|
| 108 |
|
|
class NullFilter(Filter):
|
| 109 |
|
|
'''Translates special string values to None'''
|
| 110 |
|
|
def __init__(self, reader, nulls=std_nulls):
|
| 111 |
|
|
map_ = dict.fromkeys(nulls, None)
|
| 112 |
|
|
def filter_(row): return [map_.get(v, v) for v in row]
|
| 113 |
|
|
Filter.__init__(self, filter_, reader)
|
| 114 |
|
|
|
| 115 |
|
|
class StripFilter(Filter):
|
| 116 |
|
|
'''Strips whitespace'''
|
| 117 |
|
|
def __init__(self, reader):
|
| 118 |
|
|
def filter_(row): return [v.strip() for v in row]
|
| 119 |
|
|
Filter.__init__(self, filter_, reader)
|
