Mailing-List: contact dev-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hbase.apache.org
From: Andrew Purtell <andrew.purtell@gmail.com>
Content-Type: text/plain;
	charset=us-ascii
Content-Transfer-Encoding: quoted-printable
Mime-Version: 1.0 (1.0)
Subject: Re: Consesus on removing a public method from IA.Private class whose
 instance is returned from IA.public class.
Message-Id: <9C535495-0C4E-47BC-B885-F6A8D922077D@gmail.com>
Date: Fri, 4 Dec 2015 13:21:43 -0800
References: 
 <8E858C3A39F0D046B420FBA6F75448C066784DC4@szxeml513-mbs.china.huawei.com>
 <CALte62wy71-XDBJ7LEdNqpMb05Hb1SR4K2SeNcaTxWWJzGEXDA@mail.gmail.com>
 <CAAjhxrovmK8AYQBA9YJJYBEgTZamav4nOtzrcWsdUiisX69qMA@mail.gmail.com>
 <CADcMMgHSnJAwn7paeh+1n2oJNQBxV1k6CEPuk-Z2rzKmM1SW_Q@mail.gmail.com>
 <6171871B-E98B-49BD-8A2D-1A98D0DD588A@gmail.com>
 <CAAjhxrqDMPir-9j66=CsSu5i+abh4o20zeDSKFbGmqFpbdX0bg@mail.gmail.com>
 <26C55F65-D445-4E58-8BB1-9CAAB4894B1E@gmail.com>
 <CAAjhxrqL7xAgBPKBPXF-Py07juu0A9Sn__WYGR+uyWA4yyNMUg@mail.gmail.com>
In-Reply-To: 
 <CAAjhxrqL7xAgBPKBPXF-Py07juu0A9Sn__WYGR+uyWA4yyNMUg@mail.gmail.com>
To: dev@hbase.apache.org

I'm fine with either option. Deprecating methods of private classes isn't us=
eful IMHO bit I suppose extra 'nice' for users doing things they shouldn't.=20=


> On Dec 4, 2015, at 12:50 PM, Apekshit Sharma <appy@cloudera.com> wrote:
>=20
> Ahh.. if I had mentioned it earlier, there wouldn't have been any confusio=
n.
>=20
> So the thing is, earlier patches (up to v6) are deleting the functions whi=
ch
> take byte[] or String as argument.
> These are like 15 or so 'useless' functions and I believe it's okay to
> delete them.
>=20
> Later I updated the patch (v7-v8) to mark them deprecated (instead of
> deleting) to decouple that jira from discussion [1]
>=20
> @Andrew: which patch did you look at? Can you please confirm your take on
> this.
>=20
> Thanks.
>=20
> [1]
> https://issues.apache.org/jira/browse/HBASE-14769?focusedCommentId=3D15032=
734&page=3Dcom.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#=
comment-15032734
>=20
> On Wed, Dec 2, 2015 at 5:32 PM, Andrew Purtell <andrew.purtell@gmail.com>
> wrote:
>=20
>> Ok, so no annotation change then. I looked at HBASE-14769. Change there
>> seems fine.
>>=20
>>=20
>>> On Dec 2, 2015, at 5:03 PM, Apekshit Sharma <appy@cloudera.com> wrote:
>>>=20
>>> In this particular case, we are deleting functions in HBaseAdmin which
>> take
>>> byte[] or String as argument.
>>> HBASE-14769
>>>=20
>>> On Wed, Dec 2, 2015 at 4:56 PM, Andrew Purtell <andrew.purtell@gmail.com=

>>>=20
>>> wrote:
>>>=20
>>>> We also might want to fix the annotation. Like Stack said it depends on=

>>>> the particulars. I suggest filing a JIRA.
>>>>=20
>>>>> On Dec 2, 2015, at 4:36 PM, Stack <stack@duboce.net> wrote:
>>>>>=20
>>>>> Appy makes a pretty good argument.
>>>>>=20
>>>>> What in particular are we discussing Ashish?
>>>>>=20
>>>>> Thanks,
>>>>> St.Ack
>>>>>=20
>>>>>> On Wed, Dec 2, 2015 at 3:49 PM, Apekshit Sharma <appy@cloudera.com>
>>>> wrote:
>>>>>>=20
>>>>>> I think it should be okay to remove them. When I imagine myself on
>> other
>>>>>> side, a dev using client api of Foo project, and I see a class
>> returning
>>>>>> reference to an internal class, I would not trust that function. If I=

>>>>>> really need something from returned ref, I'd probably use it, and if
>> it
>>>>>> goes away tomorrow, i'd probably curse it too, but I can't complain
>>>> since
>>>>>> it's not anything I trusted to being with.
>>>>>>=20
>>>>>> Now thinking as dev:
>>>>>> Given that documentation explicitly states that functions can
>> disappear
>>>>>> from private class, and nothing about transitiveness, I believe we ca=
n
>>>>>> remove functions from private class A.
>>>>>>=20
>>>>>>=20
>>>>>>> On Thu, Nov 26, 2015 at 7:10 AM, Ted Yu <yuzhihong@gmail.com> wrote:=

>>>>>>>=20
>>>>>>> bq. we should not remove it directly
>>>>>>>=20
>>>>>>> My understanding is the same.
>>>>>>>=20
>>>>>>> Cheers
>>>>>>>=20
>>>>>>> On Wed, Nov 25, 2015 at 10:22 PM, ashish singhi <
>>>>>> ashish.singhi@huawei.com>
>>>>>>> wrote:
>>>>>>>=20
>>>>>>>> Hello to all.
>>>>>>>>=20
>>>>>>>> I know that we can safely remove any APIs from a
>>>>>>> InterfaceAudience.Private
>>>>>>>> class and the same is described in the HBase book.
>>>>>>>>=20
>>>>>>>> Now consider a case (I did not find this mentioned in our semvar),
>>>>>>>> There is a class 'A' with InterfaceAudience.Private and class 'B'
>> with
>>>>>>>> InterfaceAudience.Public and then there is a public API in class 'B=
'
>>>>>>> which
>>>>>>>> returns an instance of class 'A'. (We missed to mark the public
>> method
>>>>>> in
>>>>>>>> class 'B' as deprecated when we marked the class 'A' as
>>>>>>>> InterfaceAudience.Private).
>>>>>>>>=20
>>>>>>>> Now the question is, can we remove the public APIs from class 'A'
>>>>>> without
>>>>>>>> marking them deprecated ?
>>>>>>>>=20
>>>>>>>> I think we should not remove it directly, it will break the users
>>>> using
>>>>>>>> these(removed) public methods from class 'A'.
>>>>>>>> P.S: This point came up  in HBASE-14769.
>>>>>>>>=20
>>>>>>>> Regards,
>>>>>>>> Ashish Singhi
>>>>>>=20
>>>>>>=20
>>>>>>=20
>>>>>> --
>>>>>>=20
>>>>>> Regards
>>>>>>=20
>>>>>> Apekshit Sharma | Software Engineer, Cloudera | Palo Alto, California=

>> |
>>>>>> 650-963-6311
>>>=20
>>>=20
>>>=20
>>> --
>>>=20
>>> Regards
>>>=20
>>> Apekshit Sharma | Software Engineer, Cloudera | Palo Alto, California |
>>> 650-963-6311
>=20
>=20
>=20
> --=20
>=20
> Regards
>=20
> Apekshit Sharma | Software Engineer, Cloudera | Palo Alto, California |
> 650-963-6311