Return-Path: X-Original-To: apmail-crunch-dev-archive@www.apache.org Delivered-To: apmail-crunch-dev-archive@www.apache.org Received: from mail.apache.org (hermes.apache.org [140.211.11.3]) by minotaur.apache.org (Postfix) with SMTP id 2FB2E1075A for ; Sat, 19 Oct 2013 20:38:17 +0000 (UTC) Received: (qmail 17437 invoked by uid 500); 19 Oct 2013 20:38:16 -0000 Delivered-To: apmail-crunch-dev-archive@crunch.apache.org Received: (qmail 17353 invoked by uid 500); 19 Oct 2013 20:38:09 -0000 Mailing-List: contact dev-help@crunch.apache.org; run by ezmlm Precedence: bulk List-Help: List-Unsubscribe: List-Post: List-Id: Reply-To: dev@crunch.apache.org Delivered-To: mailing list dev@crunch.apache.org Received: (qmail 17343 invoked by uid 99); 19 Oct 2013 20:38:07 -0000 Received: from athena.apache.org (HELO athena.apache.org) (140.211.11.136) by apache.org (qpsmtpd/0.29) with ESMTP; Sat, 19 Oct 2013 20:38:07 +0000 X-ASF-Spam-Status: No, hits=1.5 required=5.0 tests=HTML_MESSAGE,RCVD_IN_DNSWL_LOW,SPF_PASS X-Spam-Check-By: apache.org Received-SPF: pass (athena.apache.org: domain of mkwhit@gmail.com designates 209.85.223.173 as permitted sender) Received: from [209.85.223.173] (HELO mail-ie0-f173.google.com) (209.85.223.173) by apache.org (qpsmtpd/0.29) with ESMTP; Sat, 19 Oct 2013 20:38:03 +0000 Received: by mail-ie0-f173.google.com with SMTP id u16so8838832iet.18 for ; Sat, 19 Oct 2013 13:37:42 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:in-reply-to:references:from:date:message-id:subject:to :content-type; bh=OK21vGXj/rTsmjPg6OQcMMKbM5lmTmDLcDff3t6Q1z4=; b=VF0sqnESIOKqyaPwOjWiZ5fE/K35hLfJR0stn35ysTOdaNKi3Ml7+4YXdqaesQCNiP sJ6EUti70/YL58bjnwy45BB5g3wbLc5d93DHjz9tlcHikjVn0Q1arnzoXuBbU2VrIsuC 6Y54SJi//Lfnhbnqf1Zb/rLzaYTxhS5a7rN7dMcg1mgeTeed1B7J9/ZP3hBubFaEGNIu H+IZUSx86YAi55eIvsKSiitGqYrJ5E/lXAWlZXZHTiyoLCgYK4XOcl3nTwSi/l+sxUtJ 2NfmfpuHyQpLwPD5PVQEPyxiW+k4ZB3JBJV56EwEvW5jQqhNmSR/+2dxnR2Q44AFTk8J cdng== X-Received: by 10.50.119.70 with SMTP id ks6mr4288009igb.22.1382215061974; Sat, 19 Oct 2013 13:37:41 -0700 (PDT) MIME-Version: 1.0 Received: by 10.64.87.198 with HTTP; Sat, 19 Oct 2013 13:37:21 -0700 (PDT) In-Reply-To: References: From: Micah Whitacre Date: Sat, 19 Oct 2013 15:37:21 -0500 Message-ID: Subject: Re: Process of CombineFn returns ? To: dev@crunch.apache.org Content-Type: multipart/alternative; boundary=089e0111d9b8c5b67e04e91e0522 X-Virus-Checked: Checked by ClamAV on apache.org --089e0111d9b8c5b67e04e91e0522 Content-Type: text/plain; charset=ISO-8859-1 Sorry for any confusion in my example. When I stated DoFn, I didn't mean a DoFn as they exist currently but instead the "uber" CombineFn which would do the transformation and the aggregation. I meant to demonstrate that the same amount of code exists it would just be split across two Fns instead located in one. So development effort shouldn't be too significant. I am curious about how doing this transformation would affect the planned pipeline since the CombineFn processing before a GBK is an optimization. Specifically in the following example... PTable t1... PGT pgt1 = t1.gbk(); PCollection c1 = pgt1.parallelDo(CombineFn); pipeline.write(c1); pipeline.write(pgt1); Would the plan in this case equal two separate reduces? Or would we not execute the CombineFn in the map phase at all? I'm trying to figure out if we supported CombineFn, would it really be different or result in the same planned job. On Sat, Oct 19, 2013 at 12:00 AM, Chandan Biswas wrote: > Thanks Gabriel for clarifying it :) > > > On Fri, Oct 18, 2013 at 11:28 PM, Gabriel Reid >wrote: > > > Hi Chandan, > > > > Inlined below. > > > > On Sat, Oct 19, 2013 at 3:31 AM, Chandan Biswas > > wrote: > > > Please correct me if I am wrong. I want to understand more how crunch > > > create map reduce jobs as pointed out by Micah in earlier mail. > > > Suppose I am doing some steps of operation as follows: > > > I have a PTable table. > > > PGroupedTable grpedTable1=table.groupByKey(); > > > Now I am applying CombineFn on grpedTable1 and getting table2 > > > PTable table2=grpedTable1.parallelDo(..,CombineFn,..); > > > PGrpoupedTable grpedTable2=table2.groupByKey(); > > > PTable table3=grpedTable2.parallelDo(..,DoFn,...); > > > > > > So, which type of grpedTable2 or grpdTable1 will be used for reducers? > My > > > understanding is type of grpedTable2 will be used for reducers and type > > of > > > grpedTable1 will be used for shuffle/sorting at map side. Otherwise, > > there > > > will be no way send the Iterable data to reducers. > > > If that is the case, then the point of not changing the type by > CombineFn > > > doesn't hold. Otherwise, not changing the type by CombineFn makes > > complete > > > sense. > > > > > > > In this example, there would be two MapReduce jobs kicked off. The > > first one would read in table, and then use a Combiner (based on the > > CombineFn) before the reducer (i.e. before the groupByKey), and then > > the same CombineFn within the reducer, to create table2. > > > > Going from table2 would be another MapReduce job that would do nothing > > in the mapper, and execute the supplied DoFn in the reducer. > > > > > It will be awesome to have such functionality like Spark as Josh > pointed > > > out to overcome it in Crunch. > > > > Just to be clear, adding the "Aggregatable" functionality in Crunch > > won't actually add anything that isn't possible right now -- instead, > > it will just wrap current functionality into a more readable unit (at > > least that's how I see it). > > > > - Gabriel > > > > > > > Thanks, > > > Chandan > > > > > > > > > > > > On Fri, Oct 18, 2013 at 7:34 PM, Josh Wills > wrote: > > > > > >> I'm certainly not opposed to having something like this. Spark makes > > this > > >> distinction via Accumulable vs. Accumulator: > > >> > > >> > > >> > > > http://spark.incubator.apache.org/docs/0.8.0/api/core/index.html#org.apache.spark.Accumulable > > >> > > >> > > > http://spark.incubator.apache.org/docs/0.8.0/api/core/index.html#org.apache.spark.Accumulator > > >> > > >> Maybe we want something like "Aggregatable" to go along with our > > >> Aggregator (which could extend Aggregatable)? > > >> > > >> > > >> > > >> On Fri, Oct 18, 2013 at 1:36 PM, Gabriel Reid > >> >wrote: > > >> > > >> > This use case (map/combine to ) seems to come up > > >> > repeatedly. The solution (map to > and then > > >> > combine) works but is also pretty unintuitive. > > >> > > > >> > Any thoughts on adding a util in Crunch to do this? It would > basically > > >> > just need to be a static util method that takes a MapFn<> > > >> > and a CombineFn and would take care of the singleton collection > > >> > mapping stuff internally. On the one hand I'm thinking that this > could > > >> > be pretty useful, but I'm not sure if it would make things more > > >> > intuitive or possibly have the reverse effect. > > >> > > > >> > Any opinions? I'm up for putting it together if people think it's > > worth > > >> it. > > >> > > > >> > - Gabriel > > >> > > > >> > > > >> > On Fri, Oct 18, 2013 at 4:14 PM, Micah Whitacre > > >> wrote: > > >> > > Thinking about the technical issues at first glance you could say > > the > > >> > > restriction is just the way the java generics are written for the > > >> > CombineFn > > >> > > class but if you think about what is actually happening it would > be > > >> > awkward > > >> > > to support changing types in the CombineFn especially when it is > > paired > > >> > > with a GroupByKey. As I showed in the example the CombineFn > > >> essentially > > >> > > bookends the GBK operation by performing processing on the types > > before > > >> > and > > >> > > after the sorting. The GBK's types describe the output of the map > > >> phase > > >> > > and input to the reduce. If the CombineFn changed the types then > > the > > >> > > output wouldn't match the types describe by the GBK. I'm guessing > > this > > >> > > could lead to a number of problems trying to compute the types and > > plan > > >> > for > > >> > > the job. > > >> > > > > >> > > > > >> > > On Fri, Oct 18, 2013 at 8:55 AM, Micah Whitacre > > > >> > wrote: > > >> > > > > >> > >> I'm not sure I follow how there is extra effort involved. Are > you > > >> > talking > > >> > >> development effort or processing effort? From a development > > effort in > > >> > both > > >> > >> cases you need to write code that translates T to U and combines > > the > > >> > >> values. The difference is whether that logic exists inside of a > > >> single > > >> > >> DoFn or is split into a MapFn + CombineFn. So the development > > effort > > >> > >> should be the same. > > >> > >> > > >> > >> > > >> > >> On Fri, Oct 18, 2013 at 8:11 AM, Chandan Biswas < > > >> cbiswas1983@gmail.com > > >> > >wrote: > > >> > >> > > >> > >>> yeah.. i see what you are talking about. But it will take extra > > >> effort > > >> > to > > >> > >>> convert to U type. So, is there any specific reason the way > > CombineFn > > >> > >>> created initially that CombineFn will not allow other return > type. > > >> Was > > >> > >>> there any constraints (design / complexity) to restrict to this > > >> > behavior? > > >> > >>> Thanks, > > >> > >>> > > >> > >>> > > >> > >>> On Thu, Oct 17, 2013 at 8:47 PM, Micah Whitacre < > mkwhit@gmail.com > > > > > >> > wrote: > > >> > >>> > > >> > >>> > Chandan, > > >> > >>> > So let's apply your situation to the types and conversion > > that > > >> is > > >> > >>> > proposed and break it down where logic will be applied. Say > we > > >> have > > >> > a > > >> > >>> > PCollection that is like the following: > > >> > >>> > > > >> > >>> > Mapper 1: > > >> > >>> > > > >> > >>> > > > >> > >>> > > > >> > >>> > > > >> > >>> > Mapper 2 > > >> > >>> > > > >> > >>> > > > >> > >>> > This will be represented by the PTable. We > > then > > >> > apply > > >> > >>> a > > >> > >>> > MapFn to transform it into PTable and we'd get > the > > >> > >>> following > > >> > >>> > in our PCollection: > > >> > >>> > > > >> > >>> > Mapper 1 > > >> > >>> > > > > >> > >>> > > > > >> > >>> > > > > >> > >>> > > > >> > >>> > Mapper 2 > > >> > >>> > > > > >> > >>> > > > >> > >>> > Then if we were to use the GBK + CombineFn, the output of the > > map > > >> > phase > > >> > >>> > would be.. > > >> > >>> > > > >> > >>> > Mapper 1 > > >> > >>> > > > > >> > >>> > > > > >> > >>> > > > >> > >>> > Mapper 2 > > >> > >>> > > > > >> > >>> > > > >> > >>> > Notice Mapper 1 would only be emitting 2 items instead of 3 > and > > >> > >>> therefore > > >> > >>> > less data is sent over the wire and has to be sorted. Also in > > the > > >> > >>> reducer > > >> > >>> > after the GBK is completed the CombineFn would finish its work > > and > > >> > you'd > > >> > >>> > get the following: > > >> > >>> > > > >> > >>> > Reducer 1 > > >> > >>> > > > > >> > >>> > > > > >> > >>> > > > >> > >>> > The only case where this would not improve performance is if > you > > >> > never > > >> > >>> emit > > >> > >>> > data for the same key from the same mapper or your mapper > > doesn't > > >> > reduce > > >> > >>> > the size of the data. > > >> > >>> > > > >> > >>> > > > >> > >>> > On Thu, Oct 17, 2013 at 8:18 PM, Chandan Biswas < > > >> > cbiswas1983@gmail.com > > >> > >>> > >wrote: > > >> > >>> > > > >> > >>> > > I have PTable. and getting after reduce > > >> > PTable > >> > >>> > > Book> > > >> > >>> > > > > >> > >>> > > T--> Comment{ String comment, String author}, U--> > Book{String > > >> id, > > >> > >>> String > > >> > >>> > > lengthiestComment, int noOfComments} > > >> > >>> > > > > >> > >>> > > But wanted to some aggregations in the map side based on > some > > >> logic > > >> > >>> > instead > > >> > >>> > > of all aggregations at reduce side. > > >> > >>> > > Yes in worst case, data flow over the n/w will remain same, > > but > > >> > >>> sorting > > >> > >>> > > will be improved. > > >> > >>> > > > > >> > >>> > > Thanks, > > >> > >>> > > Chandan > > >> > >>> > > > > >> > >>> > > > > >> > >>> > > On Thu, Oct 17, 2013 at 6:46 PM, Josh Wills < > > jwills@cloudera.com > > >> > > > >> > >>> wrote: > > >> > >>> > > > > >> > >>> > > > On Thu, Oct 17, 2013 at 4:41 PM, Chandan Biswas < > > >> > >>> cbiswas1983@gmail.com > > >> > >>> > > > >wrote: > > >> > >>> > > > > > >> > >>> > > > > Yeah, I agree with Micah that it will not eliminate the > > >> reduce > > >> > >>> phase > > >> > >>> > > > > entirely. But the dummy object of U suggested by Josh > (or > > >> > >>> converting > > >> > >>> > > to U > > >> > >>> > > > > type in map for every record) will not improve > > performance > > >> > >>> because > > >> > >>> > > same > > >> > >>> > > > > amounts of records will be sorted and aggregated in the > > >> reduce > > >> > >>> phase. > > >> > >>> > > > > > >> > >>> > > > > > >> > >>> > > > I don't think that's true-- the records of type U will be > > >> > combined > > >> > >>> on > > >> > >>> > the > > >> > >>> > > > map-side, which would reduce the amount of data that is > > pushed > > >> > over > > >> > >>> the > > >> > >>> > > > network and improve performance. > > >> > >>> > > > > > >> > >>> > > > Can you give any additional details about what T and U are > > in > > >> > this > > >> > >>> > > > scenario? :) > > >> > >>> > > > > > >> > >>> > > > > > >> > >>> > > > > > >> > >>> > > > > But > > >> > >>> > > > > my point is, can we improve it by applying a combiner > > where > > >> the > > >> > >>> > > combineFn > > >> > >>> > > > > provides output as different type. If we have same type, > > we > > >> can > > >> > >>> use > > >> > >>> > the > > >> > >>> > > > > combiner to do some aggregation in map side which > improves > > >> > >>> > performance. > > >> > >>> > > > > But, can we have some mechanism by which the same > > advantage > > >> > can be > > >> > >>> > > > achieved > > >> > >>> > > > > when combineFn emits different type. I think, emitting > > same > > >> > type > > >> > >>> by > > >> > >>> > > > > CombineFn has restricted its use. Can we have new > > CombineFn > > >> > that > > >> > >>> > allows > > >> > >>> > > > us > > >> > >>> > > > > to output different type not only same type as input? > > >> > >>> > > > > > > >> > >>> > > > > > > >> > >>> > > > > On Thu, Oct 17, 2013 at 5:05 PM, Josh Wills < > > >> > jwills@cloudera.com> > > >> > >>> > > wrote: > > >> > >>> > > > > > > >> > >>> > > > > > Yeah, my experience in these kinds of situations is > that > > >> you > > >> > >>> need > > >> > >>> > to > > >> > >>> > > > come > > >> > >>> > > > > > up with a "dummy" or singleton version of U for the > case > > >> > where > > >> > >>> > there > > >> > >>> > > is > > >> > >>> > > > > > only a single T and do that conversion on the map side > > of > > >> the > > >> > >>> job, > > >> > >>> > > > before > > >> > >>> > > > > > the combiner runs. I think Chao had an issue like this > > >> awhile > > >> > >>> ago, > > >> > >>> > > > where > > >> > >>> > > > > he > > >> > >>> > > > > > had a PTable and wanted to write a > > combiner > > >> > that > > >> > >>> > > would > > >> > >>> > > > > > return a PTable>. The > > solution > > >> > was to > > >> > >>> > > > convert > > >> > >>> > > > > > the map-side object to a PTable > >> Collection>, > > >> > >>> where > > >> > >>> > > the > > >> > >>> > > > > > value on the map-side was a singleton list containing > > just > > >> > that > > >> > >>> > > double > > >> > >>> > > > > > value. Does that sort of trick work here? > > >> > >>> > > > > > > > >> > >>> > > > > > > > >> > >>> > > > > > On Thu, Oct 17, 2013 at 2:57 PM, Micah Whitacre < > > >> > >>> mkwhit@gmail.com> > > >> > >>> > > > > wrote: > > >> > >>> > > > > > > > >> > >>> > > > > > > Ok so the feature you are trying to achieve is the > > >> > proactive > > >> > >>> > > > > combination > > >> > >>> > > > > > of > > >> > >>> > > > > > > data before performing the GBK like the javadoc > > >> describes. > > >> > >>> > > > Essentially > > >> > >>> > > > > > in > > >> > >>> > > > > > > that situation the CombineFn is being used as a > > >> > Combiner[1] to > > >> > >>> > > > combine > > >> > >>> > > > > > the > > >> > >>> > > > > > > data local to that mapper before doing the GBK and > > then > > >> > >>> further > > >> > >>> > > > > combining > > >> > >>> > > > > > > the data in the reduce operation. It will not > > >> necessarily > > >> > >>> > > eliminate > > >> > >>> > > > > the > > >> > >>> > > > > > > need for all processing in the reduce. > > >> > >>> > > > > > > > > >> > >>> > > > > > > If you want to use this functionality you will need > > to do > > >> > the > > >> > >>> > > > > following: > > >> > >>> > > > > > > > > >> > >>> > > > > > > PTable map to PTable > > >> > >>> > > > > > > PTable gbk to PGT > > >> > >>> > > > > > > PGT combine PTable > > >> > >>> > > > > > > > > >> > >>> > > > > > > This will take advantage of any optimization > provided > > by > > >> > the > > >> > >>> > > > CombineFn. > > >> > >>> > > > > > > > > >> > >>> > > > > > > [1] - http://wiki.apache.org/hadoop/HadoopMapReduce > > >> > >>> > > > > > > > > >> > >>> > > > > > > > > >> > >>> > > > > > > > > >> > >>> > > > > > > On Thu, Oct 17, 2013 at 4:30 PM, Chandan Biswas < > > >> > >>> > > > cbiswas1983@gmail.com > > >> > >>> > > > > > > >wrote: > > >> > >>> > > > > > > > > >> > >>> > > > > > > > Hello Micah, > > >> > >>> > > > > > > > Yes we are using MapFn now. That aggregation and > > >> > >>> computation is > > >> > >>> > > > being > > >> > >>> > > > > > > done > > >> > >>> > > > > > > > in reduce phase. As CombineFn after GBK runs into > > map > > >> > side, > > >> > >>> > then > > >> > >>> > > > > those > > >> > >>> > > > > > > most > > >> > >>> > > > > > > > computations can be done in map side which are now > > >> > running > > >> > >>> in > > >> > >>> > > > reduce > > >> > >>> > > > > > > phase. > > >> > >>> > > > > > > > Some smaller aggregations and computations can be > > done > > >> on > > >> > >>> > reduce > > >> > >>> > > > > phase. > > >> > >>> > > > > > > > My point was to do some aggregation (and create a > > new > > >> > >>> object) > > >> > >>> > in > > >> > >>> > > > map > > >> > >>> > > > > > > phase > > >> > >>> > > > > > > > instead of in reduce phase. > > >> > >>> > > > > > > > > > >> > >>> > > > > > > > Thanks, > > >> > >>> > > > > > > > Chandan > > >> > >>> > > > > > > > > > >> > >>> > > > > > > > > > >> > >>> > > > > > > > On Thu, Oct 17, 2013 at 3:48 PM, Micah Whitacre < > > >> > >>> > > mkwhit@gmail.com> > > >> > >>> > > > > > > wrote: > > >> > >>> > > > > > > > > > >> > >>> > > > > > > > > Chandan, > > >> > >>> > > > > > > > > I think what you are wanting will just be a > > simple > > >> > >>> MapFn > > >> > >>> > > > instead > > >> > >>> > > > > > of > > >> > >>> > > > > > > a > > >> > >>> > > > > > > > > CombineFn. The doc of the CombineFn[1] sounds > > like > > >> > what > > >> > >>> you > > >> > >>> > > want > > >> > >>> > > > > > with > > >> > >>> > > > > > > > the > > >> > >>> > > > > > > > > statement "A special > > >> > >>> > > > > > > > > DoFn< > > >> > >>> > > > > > > > > >> > >>> > > > http://crunch.apache.org/apidocs/0.7.0/org/apache/crunch/DoFn.html > > >> > >>> > > > > > >> > >>> > > > > > > > > implementation > > >> > >>> > > > > > > > > that converts an > > >> > >>> > > > > > > > > Iterable< > > >> > >>> > > > > > > > > > > >> > >>> > > > > > > > > > >> > >>> > > > > > > > > >> > >>> > > > > > > > >> > >>> > > > > > > >> > >>> > > > > > >> > >>> > > > > >> > >>> > > > >> > >>> > > >> > > > >> > > > http://download.oracle.com/javase/6/docs/api/java/lang/Iterable.html?is-external=true > > >> > >>> > > > > > > > > > > > >> > >>> > > > > > > > > of > > >> > >>> > > > > > > > > values into a single value" but it is expecting > > the > > >> > value > > >> > >>> to > > >> > >>> > be > > >> > >>> > > > of > > >> > >>> > > > > > the > > >> > >>> > > > > > > > same > > >> > >>> > > > > > > > > time. Since you are wanting to combine the > values > > >> > into a > > >> > >>> > > > different > > >> > >>> > > > > > > form > > >> > >>> > > > > > > > it > > >> > >>> > > > > > > > > should be fairly trivial to write a MapFn that > > >> converts > > >> > >>> the > > >> > >>> > > > > > Iterable > > >> > >>> > > > > > > > -> > > >> > >>> > > > > > > > > U. > > >> > >>> > > > > > > > > > > >> > >>> > > > > > > > > [1] - > > >> > >>> > > > > > > > > > > >> > >>> > > > > > > > > >> > >>> > > > > > > >> > >>> > > > > >> > >>> > > >> > > > http://crunch.apache.org/apidocs/0.7.0/org/apache/crunch/CombineFn.html > > >> > >>> > > > > > > > > > > >> > >>> > > > > > > > > > > >> > >>> > > > > > > > > On Thu, Oct 17, 2013 at 3:30 PM, Chandan Biswas > < > > >> > >>> > > > > > cbiswas1983@gmail.com > > >> > >>> > > > > > > > > >wrote: > > >> > >>> > > > > > > > > > > >> > >>> > > > > > > > > > I was trying to refactoring some stuffs and > > trying > > >> to > > >> > >>> use > > >> > >>> > > > > > combineFn. > > >> > >>> > > > > > > > > > But when I went into deeper, found that I > can't > > do > > >> > it as > > >> > >>> > > Crunch > > >> > >>> > > > > > > doesn't > > >> > >>> > > > > > > > > > allow it the functionality I needed. For > > example, I > > >> > >>> have a > > >> > >>> > > > > > > > > > PGroupedTable. I wanted to apply > > >> CombineFn > > >> > on > > >> > >>> it > > >> > >>> > > and > > >> > >>> > > > > > wanted > > >> > >>> > > > > > > > to > > >> > >>> > > > > > > > > > get PCollection instead of T. Right now, > > >> > CombineFn > > >> > >>> > > allows > > >> > >>> > > > > only > > >> > >>> > > > > > > > same > > >> > >>> > > > > > > > > > type as return value. The use case of this > need > > is > > >> > that > > >> > >>> > there > > >> > >>> > > > > will > > >> > >>> > > > > > be > > >> > >>> > > > > > > > > some > > >> > >>> > > > > > > > > > time saving in sorting. It's natural that when > > >> > >>> aggregating > > >> > >>> > > some > > >> > >>> > > > > > > objects > > >> > >>> > > > > > > > > at > > >> > >>> > > > > > > > > > map side can create a new different type > object. > > >> > >>> > > > > > > > > > > > >> > >>> > > > > > > > > > Any thought on it? Am I missing any thing? If > > this > > >> > can > > >> > >>> be > > >> > >>> > > > written > > >> > >>> > > > > > in > > >> > >>> > > > > > > > > > different way using existing way please let me > > >> know. > > >> > >>> > > > > > > > > > > > >> > >>> > > > > > > > > > Thanks > > >> > >>> > > > > > > > > > Chandan > > >> > >>> > > > > > > > > > > > >> > >>> > > > > > > > > > > >> > >>> > > > > > > > > > >> > >>> > > > > > > > > >> > >>> > > > > > > > >> > >>> > > > > > > > >> > >>> > > > > > > > >> > >>> > > > > > -- > > >> > >>> > > > > > Director of Data Science > > >> > >>> > > > > > Cloudera > > >> > >>> > > > > > Twitter: @josh_wills > > >> > >>> > > > > > > > >> > >>> > > > > > > >> > >>> > > > > > >> > >>> > > > > > >> > >>> > > > > > >> > >>> > > > -- > > >> > >>> > > > Director of Data Science > > >> > >>> > > > Cloudera > > >> > >>> > > > Twitter: @josh_wills > > >> > >>> > > > > > >> > >>> > > > > >> > >>> > > > >> > >>> > > >> > >> > > >> > >> > > >> > > > >> > > >> > > >> > > >> -- > > >> Director of Data Science > > >> Cloudera > > >> Twitter: @josh_wills > > >> > > > --089e0111d9b8c5b67e04e91e0522--