Mailing-List: contact issues-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Date: Wed, 27 Dec 2017 08:02:00 +0000 (UTC)
From: "Zheng Hu (JIRA)" <jira@apache.org>
To: issues@hbase.apache.org
Message-ID: <JIRA.13088147.1500447206000.535388.1514361720245@Atlassian.JIRA>
In-Reply-To: <JIRA.13088147.1500447206000@Atlassian.JIRA>
References: <JIRA.13088147.1500447206000@Atlassian.JIRA> <JIRA.13088147.1500447206746@jira-lw-us.apache.org>
Subject: [jira] [Updated] (HBASE-18410) FilterList  Improvement.
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
archived-at: Wed, 27 Dec 2017 08:02:05 -0000


     [ https://issues.apache.org/jira/browse/HBASE-18410?page=3Dcom.atlassi=
an.jira.plugin.system.issuetabpanels:all-tabpanel ]

Zheng Hu updated HBASE-18410:
-----------------------------
    Release Note:=20
In this task, we fixed all existing bugs in FilterList, and did the code re=
factor which ensured interface compatibility . =20

The primary bug  fixes are :=20
1. For sub-filter in FilterList with MUST_PASS_ONE, if previous filterKeyVa=
lue() of sub-filter returns NEXT_COL, we cannot make sure that the next cel=
l will be the first cell in next column, because FilterList choose the mini=
mal forward step among sub-filters, and it may return a SKIP. so here we ad=
d an extra check to ensure that the next cell will match preivous return co=
de for sub-filters.=20
2. Previous logic about transforming cell of FilterList is incorrect, we sh=
ould set the previous transform result (rather than the given cell in quest=
ion) as the initial vaule of transform cell before call filterKeyValue() of=
 FilterList.
3. Handle the ReturnCodes which the previous code did not handle.=20

About code refactor, we divided the FilterList into two separated sub-class=
es: FilterListWithOR and FilterListWithAND,  The FilterListWithOR has been =
optimised to choose the next minimal step to seek cell rather than SKIP cel=
l one by one, and the FilterListWithAND  has been optimised to choose the n=
ext maximal key to seek among sub-filters in filter list. All in all, The c=
ode in FilterList is clean and easier to follow now.

Note that ReturnCode NEXT_ROW has been redefined as skipping to next row in=
 current family,   not to next row in all family. it=E2=80=99s more reasona=
ble, because ReturnCode is a concept in store level, not in region level.

Another bug that needs attention is: filterAllRemaining() in FilterList wit=
h MUST_PASS_ONE  will now return false if the filter list is empty whereas =
earlier it used to return true for Operator.MUST_PASS_ONE.  it's more reaso=
nable now.

  was:
In this task, we fixed all existing bugs in FilterList, and did the code re=
factor which ensured interface compatibility . =20

The primary bug  fixes are :=20
1. For sub-filter in FilterList with MUST_PASS_ONE, if previous filterKeyVa=
lue() of sub-filter returns NEXT_COL, we cannot make sure that the next cel=
l will be the first cell in next column, because FilterList choose the mini=
mal forward step among sub-filters, and it may return a SKIP. so here we ad=
d an extra check to ensure that the next cell will match preivous return co=
de for sub-filters.=20
2. Previous logic about transforming cell of FilterList is incorrect, we sh=
ould set the previous transform result (rather than the given cell in quest=
ion) as the initial vaule of transform cell before call filterKeyValue() of=
 FilterList.
3. Handle the ReturnCodes which the previous code did not handle.=20

About code refactor, we divided the FilterList into two separated sub-class=
es: FilterListWithOR and FilterListWithAND,  The FilterListWithOR has been =
optimised to choose the next minimal step to seek cell rather than SKIP cel=
l one by one, and the FilterListWithAND  has been optimised to choose the n=
ext maximal key to seek among sub-filters in filter list. All in all, The c=
ode in FilterList is clean and easier to follow now.

Note that ReturnCode NEXT_ROW has been redefined as skipping to next row in=
 current family,   not to next row in all family. it=E2=80=99s more reasona=
ble, because ReturnCode is a concept in store level, not in region level.

Another bug that needs attention is:=20


> FilterList  Improvement.=20
> -------------------------
>
>                 Key: HBASE-18410
>                 URL: https://issues.apache.org/jira/browse/HBASE-18410
>             Project: HBase
>          Issue Type: Umbrella
>          Components: Filters
>            Reporter: Zheng Hu
>            Assignee: Zheng Hu
>             Fix For: 2.0.0-alpha-4
>
>
> FilterList.java is complex now, and we have found some improvements for i=
t.  So create an issue to address it.   =20


--
This message was sent by Atlassian JIRA
(v6.4.14#64029)