Path: csiph.com!x330-a1.tempe.blueboxinc.net!usenet.pasdenom.info!news.albasani.net!weretis.net!feeder1.news.weretis.net!news.solani.org!.POSTED!not-for-mail
From: Peter Otten <__peter__@web.de>
Newsgroups: comp.lang.python
Subject: Re: groupby - summing multiple columns in a list of lists
Followup-To: comp.lang.python
Date: Tue, 17 May 2011 20:24:14 +0200
Organization: None
Lines: 74
Message-ID: <iquedu$pur$1@solani.org>
References: <2f60e6c7-1ce2-4094-a777-e544293e8843@hg8g2000vbb.googlegroups.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 7Bit
X-Trace: solani.org 1305656574 26587 eJwFwQkBACAIA8BKoI4nDoLrH8E7bFNrPwY7IDg5b4uglrODqpeVAlkyQLBVY+6khWMV430gZBE1 (17 May 2011 18:22:54 GMT)
X-Complaints-To: abuse@news.solani.org
NNTP-Posting-Date: Tue, 17 May 2011 18:22:54 +0000 (UTC)
X-User-ID: eJwNxcEBwCAIA8CVAhKUcQBl/xHa+xyXi/c2pxuHI9wtZkQXaHgRXQyoyFVpHp2Xx3TdpK//xFTC32mtnMAHKlIU3w==
Cancel-Lock: sha1:6wfDWjNF9ypza+BMPM0ju04/Dvs=
X-NNTP-Posting-Host: eJwNyMEBwCAIA8CVgpCo46DC/iO09zy6THeGqGCz7WwQlidW+AmSVZxohBJ5Xz3R2+zGmNrpKv/fOApDQAt9a72dLxGM/gCrjhlq
Xref: x330-a1.tempe.blueboxinc.net comp.lang.python:5587

Jackson wrote:

> I'm currently using a function pasted in below. This allows me to sum
> a column (index) in a list of lists.
> 
> So if mylist = [[1, 2, 3], [1, 3, 4], [2, 3, 4], [2, 4, 5]]
> group_results(mylist,[0],1)
> 
> Returns:
> [(1, 5), (2, 7)]
> 
> What I would like to do is allow a tuple/list of index values, rather
> than a single index value to be summed up, so you could say
> group_results(mylist,[0],[1,2]) would return [(1, 5,7), (2, 7,9)] but
> I'm struggling to do so, any thoughts? Cheers
> 
> from itertools import groupby as gb
> from operator import itemgetter as ig
> 
> def group_results(table,keys,value):
>     res = []
>     nkey = ig(*keys)
>     value = ig(value)
>     for k, group in gb(sorted(table,key=ig(*keys)),nkey):
>         res.append((k,sum(value(row) for row in group)))
>     return res

You could write a version of sum() that can cope with tuples:

from itertools import groupby, imap

def itemgetter(keys, rowtype=tuple):
    def getitem(value):
        return rowtype(value[key] for key in keys)
    return getitem

def sum_all(rows):
    rows = iter(rows)
    sigma = next(rows)
    rowtype = type(sigma)
    sigma = list(sigma)
    for row in rows:
        for i, x in enumerate(row):
            sigma[i] += x
    return rowtype(sigma)

def group_results(table, key, value):
    get_key = itemgetter(key)
    get_value = itemgetter(value)
    table = sorted(table, key=get_key)
    for keyvalue, group in groupby(table, get_key):
        yield keyvalue + sum_all(imap(get_value, group))

but I'd probably use a dict-based approach:

def group_results(table, key, value):
    get_key = itemgetter(key)
    get_value = itemgetter(value)
    grouped = {}
    for row in table:
        key = get_key(row)
        value = get_value(row)
        if key in grouped:
            grouped[key] = tuple(a + b for a, b in zip(grouped[key], value))
        else:
            grouped[key] = value
    return [k + v for k, v in sorted(grouped.iteritems())]

if __name__ == "__main__":
    items = [(1, 2, 3), (1, 3, 4), (2, 3, 4), (2, 4, 5)]
    print list(group_results(items, [0], [1, 2]))

Note that the function built with my version of itemgetter() will always 
return a tuple.