On 2009-05-07 23:48:43 -0500, Chris Rebert <c...@rebertia.com> said:
On Thu, May 7, 2009 at 2:23 PM, Kevin D. Smith <kevin.sm...@sas.com> wrote:
I need the behavior of heapq.merge to merge a bunch of results from a
database. I was doing this with sorted(itertools.chain(...), key=
...), but
I would prefer to do this with generators. My issue is that I need
the key
argument to sort on the correct field in the database. heapq.merge
doesn't
have this argument and I don't understand the code enough to know if it's
possible to add it. Is this enhancement possible without drasticall
y
changing the current code?
I think so. Completely untested code:
def key(ob):
#code here
class Keyed(object):
def __init__(self, obj):
self.obj = obj
def __cmp__(self, other):
return cmp(key(self.obj), key(other.obj))
def keyify(gen):
for item in gen:
yield Keyed(item)
def stripify(gen):
for keyed in gen:
yield keyed.obj
merged = stripify(merge(keyify(A), keyify(B), keyify(C))) #A,B,C being
the iterables
Ah, that's not a bad idea. I think it could be simplified by letting
Python do the comparison work as follows (also untested).
def keyify(gen, key=lamda x:x):
for item in gen:
yield (key(item), item)
def stripify(gen):
for item in gen:
yield item[1]
After looking at the heapq.merge code, it seems like something like
this could easily be added to that code. If the next() calls were
wrapped with the tuple creating code above and the yield simply
returned the item. It would, of course, have to assume that the
iterables were sorted using the same key, but that's better than not
having the key option at all.
--
Kevin D. Smith
--
http://mail.python.org/mailman/listinfo/python-list
