Mailing-List: contact user-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@hbase.apache.org
Received-SPF: pass (nike.apache.org: local policy)
References: 
 <CANBXuZA6xmQi6F8rAjQaxA-7y7taMugB8d8R5b=s8W93Qg-BzQ@mail.gmail.com>
	<CALte62yorU5u63bNf=ZJi6KK3-gPkm3YdaWDr1uHcdfu4HgsiA@mail.gmail.com>
	<CANBXuZB8Q-CeqojYxiLxS7w-k9M1pgXKG6Hpn9nLrbTCFsNdnQ@mail.gmail.com>
	<1372677275.83181.YahooMailNeo@web140601.mail.bf1.yahoo.com>
 <CANBXuZCE5mEVRnHxyuussr=jjK_kYvuScYGq=h7bg0T3uGNzcA@mail.gmail.com>
Message-ID: <1372705264.79838.YahooMailNeo@web140603.mail.bf1.yahoo.com>
Date: Mon, 1 Jul 2013 12:01:04 -0700 (PDT)
From: lars hofhansl <larsh@apache.org>
Reply-To: lars hofhansl <larsh@apache.org>
Subject: Re: Behavior of Filter.transform() in FilterList?
To: Christophe Taton <taton@wibidata.com>,
  "user@hbase.apache.org" <user@hbase.apache.org>
In-Reply-To: 
 <CANBXuZCE5mEVRnHxyuussr=jjK_kYvuScYGq=h7bg0T3uGNzcA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: multipart/alternative;
 boundary="-1172831624-845832186-1372705264=:79838"

---1172831624-845832186-1372705264=:79838
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable

It would make sense, but it is not immediately clear how to do so cleanly. =
We would no longer be able to call transform at the StoreScanner level (or =
evaluate the filter multiple times, or require the filters to maintain thei=
r - last - state and only apply transform selectively).=0A=0AI added transf=
orm() a while ago in order to allow a Filter *not* to transform. Before eac=
h we defensively made a copy of the key, just in case a Filter (such as Key=
OnlyFilter) would modify it, now this is a formalized, and the filter is re=
sponsible for making a copy only when needed.=0A=0A=0A-- Lars=0A=0A=0A=0A__=
______________________________=0A From: Christophe Taton <taton@wibidata.co=
m>=0ATo: user@hbase.apache.org; lars hofhansl <larsh@apache.org> =0ASent: M=
onday, July 1, 2013 10:27 AM=0ASubject: Re: Behavior of Filter.transform() =
in FilterList?=0A =0A=0A=0AOn Mon, Jul 1, 2013 at 4:14 AM, lars hofhansl <l=
arsh@apache.org> wrote:=0A=0AYou want transform to only be called on filter=
s that are "reached"?=0A>I.e. FilterA and FilterB, FilterB.transform should=
 not be called if a KV is already filtered by FilterA?=0A>=0A=0AYes, that's=
 what I naively expected, at first.=0A=0AThat's not how it works right now,=
 transform is called in a completely different code path from the actual fi=
ltering logic.=0A>=0A=0AIndeed, I just learned that.=0AI found no documenta=
tion of this behavior, did I miss it?=0AIn particular, the javadoc of the w=
orkflow of Filter doesn't mention transform() at all.=0AWould it make sense=
 to apply transform() only if the return code for filterKeyValue() includes=
 the KeyValue?=0A=0AC.=0A=0A-- Lars=0A>=0A>=0A>----- Original Message -----=
=0A>From: Christophe Taton <taton@wibidata.com>=0A>To: user@hbase.apache.or=
g=0A>Cc:=0A>Sent: Sunday, June 30, 2013 10:26 PM=0A>Subject: Re: Behavior o=
f Filter.transform() in FilterList?=0A>=0A>On Sun, Jun 30, 2013 at 10:15 PM=
, Ted Yu <yuzhihong@gmail.com> wrote:=0A>=0A>> The clause 'family=3DX and c=
olumn=3DY and KeyOnlyFilter' would be represented=0A>> by a FilterList, rig=
ht ?=0A>> (family=3DA and colymn=3DB) would be represented by another Filte=
rList.=0A>>=0A>=0A>Yes, that would be FilterList(OR, [FilterList(AND, [fami=
ly=3DX, column=3DY,=0A>KeyOnlyFilter]), FilterList(AND, [family=3DA, column=
=3DB])]).=0A>=0A>So the behavior is expected.=0A>>=0A>=0A>Could you explain=
, I'm not sure how you reach this conclusion.=0A>Are you saying it is expec=
ted, given the actual implementation=0A>FilterList.transform()?=0A>Or are t=
here some other details I missed?=0A>=0A>Thanks!=0A>C.=0A>=0A>On Mon, Jul 1=
, 2013 at 1:10 PM, Christophe Taton <taton@wibidata.com> wrote:=0A>>=0A>> >=
 Hi,=0A>> >=0A>> > From=0A>> >=0A>> >=0A>> https://github.com/apache/hbase/=
blob/0.95/hbase-client/src/main/java/org/apache/hadoop/hbase/filter/FilterL=
ist.java#L183=0A>> > ,=0A>> > it appears that Filter.transform() is invoked=
 unconditionally on all=0A>> > filters in a FilterList hierarchy.=0A>> >=0A=
>> > This is quite confusing, especially since I may construct a filter lik=
e:=0A>> >=A0 =A0 =A0(family=3DX and column=3DY and KeyOnlyFilter) or (famil=
y=3DA and colymn=3DB)=0A>> > The KeyOnlyFilter will remove all values from =
the KeyValues in A:B as=0A>> well.=0A>> >=0A>> > Is my understanding correc=
t? Is this an expected/intended behavior?=0A>> >=0A>> > Thanks,=0A>> > C.=
=0A>> >=0A>>=0A>=0A>
---1172831624-845832186-1372705264=:79838--