Mailing-List: contact user-help@couchdb.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@couchdb.apache.org
Received-SPF: pass (athena.apache.org: domain of jchris@gmail.com designates
 74.125.83.180 as permitted sender)
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=gmail.com; s=gamma;
        h=content-type:mime-version:subject:from:in-reply-to:date
         :content-transfer-encoding:message-id:references:to:x-mailer;
        b=GxHrqR3nqZY1Z7C+U5/HyJMaIwCpXT8wBmVxW50SPODxIXjxaiJELgFXPhA/5ivMiH
         R4GS3QHaNtYMn15Vcd+fPpGbcIS4xhPPn2IsjHbVq9HYEhjXqN3OZ2t3/mKTkXyMluEj
         8JLLQ8+cV1aIDqgCj4ANsPfPLjuURGiRO1lsg=
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0 (Apple Message framework v1078)
Subject: Re: Reduce function results
From: J Chris Anderson <jchris@gmail.com>
In-Reply-To: <4BF6C4CA.2030405@gmail.com>
Date: Fri, 21 May 2010 10:53:57 -0700
Content-Transfer-Encoding: quoted-printable
Message-Id: <2FEF50BE-2500-4300-818D-E0246DBD4004@gmail.com>
References: <4BF51F4F.4070706@gmail.com>
	 <AANLkTim_cr-a4x2C4JY3jUjWQTMWVFFad-jBzGHBzf4B@mail.gmail.com>
	 <4BF5CC48.7010604@gmail.com>
 <AANLkTikJ3om40gQA-zSw5bDykVxyz-H8SvK1G_CHi07O@mail.gmail.com>
 <4BF63DC9.7060304@gmail.com> <4BF6428F.9010802@gmail.com>
 <4DB9C42A-4B93-4634-A1FD-DB9540B4EA18@gmail.com> <4BF6C4CA.2030405@gmail.com>
To: user@couchdb.apache.org


On May 21, 2010, at 10:37 AM, Tom Sante wrote:

> Because the data was stored like that, but indeed for future database =
this will be stored as a JSON number. Still doesn't explain why the =
different behaviour of Number() in couch compared to 'js'.
>=20

The only factor I can think of is the inclusion of json2.js in the =
main.js source code... You might have a look through share/server for =
clues.

> Side question:
> What would result in the smallest document byte size? storing it as a =
string like "0.345" (only need precision of 0.001) or as a JSON number?

You'd have to experiment to know for sure. I'm gonna guess that numbers =
will store smaller than strings (they will definitely sort faster).

Chris

> For a few docs this will not matter but in my case a dataset size of =
10 million docs (seperated in different DB's), this could have a =
significant impact.
>=20
> On 21/05/10 18:41, J Chris Anderson wrote:
>> It's this line that looked odd to me:
>>=20
>>>>>> var v =3D new Number(values[val][v_n]);
>>>>=20
>>=20
>>=20
>>=20
>> I don't usually see the new Number() construct. Why not just store as =
a JSON number and then use values[val][v_n] directly?
>>=20
>> Chris
>>=20
>> On May 21, 2010, at 1:21 AM, Tom Sante wrote:
>>=20
>>> Ok the log() helped to narrow things down:
>>> res[v_n]=3D{'mi':v,'ma':v,'c':1,'t':v};
>>> in this context v isn't returned as a Number but as a {}
>>> In number context, like in a Math.min(v,othernumber); it works fine.
>>> Same thing happens with log:
>>> log(v) returns {}
>>> log(''+v) returns -0.453
>>>=20
>>> Any reason why this would be the case since this is not the =
behaviour of a Number in normal js (JavaScript-C 1.7.0 2007-10-03) ?
>>>=20
>>> I worked around it by forcing v to become a string here
>>> res[v_n]=3D{'mi':''+v,'ma':''+v,'c':1,'t':''+v};
>>> But when i later do a sum like res[v_n].t +=3D v;
>>> it just concats the v to it instead of adding its values
>>> So there i have to convert it back to a Number() before adding.
>>>=20
>>> On 21/05/10 10:01, Tom Sante wrote:
>>>> Yeah I'll try that, see if it gives me any clues.
>>>> Is there a way to run your code manually for debugging via the =
command
>>>> line with the JS engine the way couch calls it to build the view? =
(like
>>>> an interactive mode of couchjs)
>>>>=20
>>>> On 21/05/10 04:59, Zachary Zolton wrote:
>>>>> Reduce functions can be tricky. My best advice: try log()'ing all =
your
>>>>> input and intermediary results; the answer usually jumps out to =
ya.
>>>>>=20
>>>>> On Thursday, May 20, 2010, Tom Sante<tom.sante@gmail.com>  wrote:
>>>>>> On 20/05/10 23:52, David Goodlad wrote:
>>>>>>=20
>>>>>> On Thu, May 20, 2010 at 9:38 PM, Tom Sante<tom.sante@gmail.com>  =
wrote:
>>>>>>=20
>>>>>> Hi,
>>>>>>=20
>>>>>> I have a reduce function like this:
>>>>>> [snip]
>>>>>> So for some reason if there is only 1 element to be reduced, on =
the
>>>>>> line of
>>>>>> res[v_n]=3D{'mi':v,'ma':v,'c':1,'t':v};
>>>>>> v becomes v=3D{} and not the actual number. Any ideas why?
>>>>>>=20
>>>>>> I tested the reduce function with the command line 'js' and send =
it the
>>>>>> above test keys manually as function argument and then it does =
result
>>>>>> in the
>>>>>> correct return:
>>>>>> "01:00:000074" ->  {raw: {mi: -0.213, ma: -0.213, c: 1, t: =
-0.213}}
>>>>>> "01:00:000084" ->  {raw: {mi: -0.213, ma: -0.213, c: 1, t: =
-0.213}}
>>>>>>=20
>>>>>>=20
>>>>>> Sounds like you need to specify group_level; Try playing around =
in
>>>>>> Futon, you'll see a group_level dropdown.
>>>>>>=20
>>>>>> Dave
>>>>>>=20
>>>>>>=20
>>>>>> Why would I need a group_level parameter?
>>>>>> My key is a simple string "01:00:000074".
>>>>>> Its my values don't get correctly generated by couchjs, since it =
does
>>>>>> out but proper values in normal commandline js.
>>>>>>=20
>>>>>> var v =3D new Number(values[val][v_n]);
>>>>>> if(res.hasOwnProperty(v_n)){
>>>>>> res[v_n].mi =3D Math.min(v,res[v_n].mi);
>>>>>> res[v_n].ma =3D Math.max(v,res[v_n].ma);
>>>>>> res[v_n].c++;
>>>>>> res[v_n].t +=3D v;
>>>>>> }else{
>>>>>> res[v_n]=3D{'mi':v,'ma':v,'c':1,'t':v};
>>>>>> }
>>>>>>=20
>>>>>> It seems that in the above part in the 'else' case the v isn't a
>>>>>> number but an empty object {}
>>>>>>=20
>>>>=20
>>>=20
>>=20
>>=20
>=20