list comprehension (searching for onliners)

[Gerardo Herzig]

Hi all: I have this list thing as a result of a db.query: (short version)
result = [{'service_id' : 1, 'value': 10},
{'service_id': 2, 'value': 5},
{'service_id': 1, 'value': 15},
{'service_id': 2, 'value': 15},
]

and so on...what i need to do is some list comprehension that returns me
something like

result = [
{
'service_id' : 1, 'values': [ {'value': 10},
{'value': 15}]
},
{
'service_id' : 2, 'values': [ {'value': 5}, {'value': 15}]
}

My problem now is i cant avoid have "repeteated" entries, lets say, in
this particular case, 2 entries for "service_id = 1", and other 2 for
"service_id =2".
Ill keeping blew off my hair and drinking more cofee while searching for
this damn onliner im looking for.

Thanks dudes.
Gerardo

[Max Erickson]

Gerardo Herzig <ghe...@fmed.uba.ar> wrote:

Is three lines ok?

>>> output=dict()
>>> for record in result:
output.setdefault(record['service_id'], list()).append(record
['value'])

>>> output
{1: [10, 15], 2: [5, 15]}
>>>

Creating the more verbose output that you specified should be
pretty straighforward from there.

max

[Tim Chase]

> result = [{'service_id' : 1, 'value': 10},
> {'service_id': 2, 'value': 5},
> {'service_id': 1, 'value': 15},
> {'service_id': 2, 'value': 15},
> ]
>
> and so on...what i need to do is some list comprehension that returns me
> something like
>
> result = [
> {
> 'service_id' : 1, 'values': [ {'value': 10},
> {'value': 15}]
> },
> {
> 'service_id' : 2, 'values': [ {'value': 5}, {'value': 15}]
> }
>
>
> My problem now is i cant avoid have "repeteated" entries, lets say, in
> this particular case, 2 entries for "service_id = 1", and other 2 for
> "service_id =2".

Okay...while I'm not sure the opacity of a one-liner is actually
productive, it *can* be done. Whether it should, I leave that to
your discernment. :)

>>> [{'service_id': i, 'values':[{'value':d2['value']} for d2 in
result if d2['service_id'] == i ]} for i in set(d['service_id']
for d in result)]

[{'service_id': 1, 'values': [{'value': 10}, {'value': 15}]},
{'service_id': 2, 'values': [{'value': 5}, {'value': 15}]}]

There's no claiming it's efficient, as it looks like there may be
some O(N^2) logic going on under the hood (or possibly O(N*M)
where N is the size of the result-set and M is the count of
unique service_id values), as it's iterating over the result-set
in two dimensions...once to create the set of top-level indices,
and once for each result.

If you didn't have to have all those dictionaries around, it
might come out more cleanly to just have some result-set that
came out to be

{1: [10,15], 2: [5,15]}

Just a few thoughts...

-tkc

[Jon Clements]

Gerardo Herzig wrote:

If you import itertools and have your DB query return in order of
service_id (or sort the list after retrieving result), then...

[ [dict(service_id=key),[dict(value=n['value']) for n in value]] for
key,value in itertools.groupby(result,lambda x: x['service_id']) ]

... is more or less what you want.

Jon.

[Gerardo Herzig]

You are the man, you are the man!! Yes, i need those dicts in order to
assign them to a nested <TMPL_LOOP> for the htmltmpl templating engine.
Thats why i cant use the solution provided by Max, and thanks to Jon too!!
Thanks a lot dudes, i hope someday ill turn myself into some of you guys
who can actually answer questions ;)

Gerardo

>> result = [{'service_id' : 1, 'value': 10},
>> {'service_id': 2, 'value': 5},
>> {'service_id': 1, 'value': 15},
>> {'service_id': 2, 'value': 15},
>> ]
>>
>> and so on...what i need to do is some list comprehension that returns
>> me something like
>>
>> result = [
>> {
>> 'service_id' : 1, 'values': [ {'value': 10},
>> {'value': 15}]
>> },
>> {
>> 'service_id' : 2, 'values': [ {'value': 5}, {'value': 15}]
>> }
>>
>> My problem now is i cant avoid have "repeteated" entries, lets say,
>> in this particular case, 2 entries for "service_id = 1", and other 2
>> for "service_id =2".
>
>

[Paul Rubin]

Gerardo Herzig <ghe...@fmed.uba.ar> writes:
> You are the man, you are the man!! Yes, i need those dicts in order to
> assign them to a nested <TMPL_LOOP> for the htmltmpl templating
> engine. Thats why i cant use the solution provided by Max, and thanks
> to Jon too!!

I'm confused, can't you write a multi-line function and pass that?

[Fredrik Lundh]

Gerardo Herzig wrote:

> Thanks a lot dudes, i hope someday ill turn myself into some of you guys
> who can actually answer questions ;)

if you want to become a good Python programmer, you really need to get
over that "I need a oneliner" idea.

</F>

[Roberto Bonvallet]

Gerardo Herzig wrote:
[...]

> and so on...what i need to do is some list comprehension that returns me
> something like

[...]

You don't _need_ a list comprehension, you just _want_ one :)

[...]

> Ill keeping blew off my hair and drinking more cofee while searching for
> this damn onliner im looking for.

I know, trying to put complex logic in one line makes you do all that.

Go for the multiliner!
--
Roberto Bonvallet

[Carsten Haese]

On Fri, 2006-10-20 at 10:53, Fredrik Lundh wrote:
> if you want to become a good Python programmer, you really need to get
> over that "I need a oneliner" idea.

+1 QOTW

-Carsten