A custom CSV reader implementation with direct file access

The default builtin Python csv lib uses an 8KB read-ahead buffer on the file pointer, making fp.tell() yield inaccurate results. This library addresses that head on, explicitly passing back the file pointer position with each row, as well as allowing for direct seeking.



pip install csv-position-reader


>>> import csv_position_reader

>>> with open('tests/data/basic.csv', 'r') as fp:
...     reader = csv_position_reader.DictReader(fp)
...     position, row = reader.next()
...     print "position: %s" % position
...     print "row: %s" % row
...     reader.seek(position)
...     position_new, row_new = reader.next()
...     assert position == position_new
...     assert row == row_new
position: 26
row: {'city': 'Atlanta', 'favorite_color': 'black', 'name': 'Sam'}

Why? / Who Cares?

Because after poring through a CSV one time, you can now build a dictionary/cache of where each row lives for future O(1) access! You're now a stone's throw away from a CSV-driven database!