csv.py: thin wrapper over _csv

csv.py is almost entirely a thin shim. The heavy lifting (parsing, quoting, dialect validation) lives in the _csv C extension. The Python file re-exports that module's public names, then adds DictReader, DictWriter, and a handful of convenience helpers on top.

Map

Line range	Symbol	Role
1-8	from _csv import *	re-exports all C-level names
9-14	__version__, __doc__	module metadata
15-60	DictReader	wraps reader iterator, yields dicts
61-120	DictWriter	wraps writer, maps fieldnames to rows
121-155	Dialect	base class; register_dialect delegates to _csv
156-175	excel, excel_tab, unix_dialect	built-in dialect subclasses
176-200	Sniffer	heuristic dialect detector

Reading

Re-export pattern

CPython keeps the fast path entirely in C and uses the Python file only for the public API surface. The star-import at the top of the file pulls in reader, writer, register_dialect, unregister_dialect, get_dialect, list_dialects, field_size_limit, and all the QUOTE_* constants.

# Lib/csv.py:6
from _csv import *
# _csv exposes: reader, writer, register_dialect, unregister_dialect,
# get_dialect, list_dialects, field_size_limit, Error, QUOTE_ALL,
# QUOTE_MINIMAL, QUOTE_NONNUMERIC, QUOTE_NONE, Dialect (C type)

Any gopy port must implement _csv as a built-in module first. The Python layer then compiles without changes.

DictReader

DictReader wraps the low-level reader object and yields dict rows keyed by fieldnames. The fieldname list is read lazily from the first row when fieldnames is not supplied at construction time.

# Lib/csv.py:82-95  (DictReader.__next__)
def __next__(self):
    if self.line_num == 0:
        self.fieldnames  # side-effect: read header row if needed
    row = next(self.reader)
    self.line_num = self.reader.line_num
    while row == []:
        row = next(self.reader)
    d = dict(zip(self.fieldnames, row))
    lf = len(self.fieldnames)
    lr = len(row)
    if lf < lr:
        d[self.restkey] = row[lf:]
    elif lf > lr:
        for key in self.fieldnames[lr:]:
            d[key] = self.restval
    return d

Dialect registration

User-defined dialects must subclass csv.Dialect and set class-level attributes. register_dialect validates the attributes through _csv's C validator before storing the dialect name.

# Lib/csv.py:152-155  (register_dialect)
def register_dialect(name, dialect=None, **fmtparams):
    # Dialect() call merges fmtparams over the base dialect,
    # then _csv.register_dialect validates field types.
    if dialect is not None:
        dialect = type(str(name), (dialect,), fmtparams)
    _csv.register_dialect(name, dialect)

gopy notes

• _csv must be a built-in module registered before csv is imported. The Sniffer class is pure Python and can be ported after _csv.
• DictReader.fieldnames uses a @property that mutates self._fieldnames on first access. The Go port needs to handle that lazy-init path carefully.
• excel and excel_tab are registered via register_dialect at module init time (bottom of csv.py). Ensure the built-in module init calls register_dialect before user code runs.