From a conversation. "I want a better CSV that knows what my columns mean".

ACSV is "Annotated CSV" format. Annotations are provided in a "preamble". As far as possible, annotation semantics are left up to the consuming code. The only exception to this (in this POC version) is that there must be a list of "columns", and the metadata for each column must include a "fieldname" property. This is passed to csv.DictReader, and will presumably be made optional in some way.

Annotations come in two flavors: Global annotations apply to the document as a whole, and might include title, description, that sort of jazz.

Column annotations are listed in order. Column annotations currently must include a "fieldname" for each column - this is passed to csv.DictReader.

Usage:

from acsv import DictReader f = open("test.acsv") dr = DictReader(f) dr.annotations {'global': {'title': 'Test CSV', 'description': 'An annotated CSV document for testing ACSV code'}, 'columns': [{'fieldname': 'Time', 'units': 'seconds', 'type': 'int'}, {'fieldname': 'Distance', 'units': 'meters', 'type': 'float'}]} [row for row in dr] [{'0': '1'}, {'0': '2'}, {'0': '3.1'}, {'0': '3.9'}, {'0': '4.8'}]

Status: This first pass is purely a POC, using JSON for the preamble. Next step is to develop a better format for the preamble and implement that.