Queries regarding Put and Scanner Result

[Imran M Yousuf <im...@gmail.com>]

Hi,

I need some suggestions regarding usage of Put and ResultScanner's Result.

First, the Result. Since we are using Object to row(s) we do a
conversion of Object to row and vice versa. When doing row to object I
am doing something like Result.next(int) then every result is
converted to to its object in their individual thread, i.e. conversion
of row to object takes place parallelly. Now my question is, will it
have any effect if this conversion does take place in parallel?

Second, when doing object to row, we are mapping a one to many
relation in a specific column family, e.g. cf1, now when I will update
the row I will populate cf1 with values set newly, e.g. old values are
1,2,3 and new values are 1,3,4, now I will populate cf1 to have 1,3,4.
Now my first question is what will happen to the value '2'? When I do
a scan/get on that row will cf1 return 1,3,4 or 1,2,3,4? If the answer
is 1,2,3,4, when getting cf1 I want to get 1,3,4 only as it was set
and not 2, what is the best way to achieve it?

Thank you,

-- 
Imran M Yousuf
Twitter: @imyousuf - http://twitter.com/imyousuf
Blog: http://imyousuf-tech.blogs.smartitengineering.com/
Mobile: +880-1711402557

Re: Queries regarding Put and Scanner Result

[Andrey Stepachev <oc...@gmail.com>]

2010/9/25 Imran M Yousuf <im...@gmail.com>:
> On Thu, Sep 23, 2010 at 11:28 AM, Andrey Stepachev <oc...@gmail.com> wrote:
>> 2010/9/23 Imran M Yousuf <im...@gmail.com>:
>> How you populate family cf1? 1,2,3,4 = are qualifiers? or you put values
>> under the same qualifiers?
>>
>> In case of qualifiers, you should delete qualifier+value explicitly.
>
> Hmm, this is the information I needed confirmation on.

I mean that if you put values in a row:
key1 {
  cf1:qual1:value1
  cf1:qual2:value2
  cf1:qual3:value3
}

and then put
key1 {
  cf1:qual1:value4
  cf1:qual3:value5
}

you row will be
key1 {
  cf1:qual1:value4
  cf1:qual2:value2
  cf1:qual3:value5
}

So, to remove cf1:qual2 you need to write null (hide value), or delete
all versions
of this qualifier. If you, for example delete cf1:qual3:value5, value3
will be returned
on you get for cf1:qual3. This is how versioning works in hbase.


> Thanks,
>
> --
> Imran M Yousuf
> Twitter: @imyousuf - http://twitter.com/imyousuf
> Blog: http://imyousuf-tech.blogs.smartitengineering.com/
> Mobile: +880-1711402557
>

Re: Queries regarding Put and Scanner Result

[Imran M Yousuf <im...@gmail.com>]

On Thu, Sep 23, 2010 at 11:28 AM, Andrey Stepachev <oc...@gmail.com> wrote:
> 2010/9/23 Imran M Yousuf <im...@gmail.com>:
>> <snip />
>> Second, when doing object to row, we are mapping a one to many
>> relation in a specific column family, e.g. cf1, now when I will update
>> the row I will populate cf1 with values set newly, e.g. old values are
>> 1,2,3 and new values are 1,3,4, now I will populate cf1 to have 1,3,4.
>> Now my first question is what will happen to the value '2'? When I do
>> a scan/get on that row will cf1 return 1,3,4 or 1,2,3,4? If the answer
>> is 1,2,3,4, when getting cf1 I want to get 1,3,4 only as it was set
>> and not 2, what is the best way to achieve it?
>
> How you populate family cf1? 1,2,3,4 = are qualifiers? or you put values
> under the same qualifiers?
>
> In case of qualifiers, you should delete qualifier+value explicitly.

Hmm, this is the information I needed confirmation on.

> This can be done in two ways:
> 1. delete whole cf for given key
> http://hbase.apache.org/docs/r0.20.6/api/org/apache/hadoop/hbase/client/Delete.html#deleteFamily(byte[],
> long)
>

This is the brutforce approach.

> 2. get current values, make diff with new value, and delete obsolete value.
>

This is the more elegant approach but for a lot of simultaneous Puts'
it could be expensive.

I will ponder on them and check how to implement them efficiently.

Thanks,

-- 
Imran M Yousuf
Twitter: @imyousuf - http://twitter.com/imyousuf
Blog: http://imyousuf-tech.blogs.smartitengineering.com/
Mobile: +880-1711402557

Re: Queries regarding Put and Scanner Result

[Andrey Stepachev <oc...@gmail.com>]

2010/9/23 Imran M Yousuf <im...@gmail.com>:
> Hi,
>
>
> Second, when doing object to row, we are mapping a one to many
> relation in a specific column family, e.g. cf1, now when I will update
> the row I will populate cf1 with values set newly, e.g. old values are
> 1,2,3 and new values are 1,3,4, now I will populate cf1 to have 1,3,4.
> Now my first question is what will happen to the value '2'? When I do
> a scan/get on that row will cf1 return 1,3,4 or 1,2,3,4? If the answer
> is 1,2,3,4, when getting cf1 I want to get 1,3,4 only as it was set
> and not 2, what is the best way to achieve it?

How you populate family cf1? 1,2,3,4 = are qualifiers? or you put values
under the same qualifiers?

In case of qualifiers, you should delete qualifier+value explicitly.
This can be done in two ways:
1. delete whole cf for given key
http://hbase.apache.org/docs/r0.20.6/api/org/apache/hadoop/hbase/client/Delete.html#deleteFamily(byte[],
long)

2. get current values, make diff with new value, and delete obsolete value.

>
> Thank you,
>
> --
> Imran M Yousuf
> Twitter: @imyousuf - http://twitter.com/imyousuf
> Blog: http://imyousuf-tech.blogs.smartitengineering.com/
> Mobile: +880-1711402557
>