hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael Dagaev" <michael.dag...@gmail.com>
Subject Re: Question on Multimap
Date Tue, 11 Nov 2008 19:04:02 GMT
Jonathan

    Let's consider the "Person" example. There is table "Person" with
"PersonalDetails" column family.
The column family contains many different key-value pairs (name,
address, email, etc.) and I do not know
the keys upfront. Each table row contains data about a single person
and has a key -- person ID.
We store and fetch the "PersonalDetails" data from the table by person
ID as Map<String, String>

   So far so good.

   Now I have discovered that the "PersonalData" data is Map<String,
List<String>> rather than Map<String, String>,
i.e. there are a few values for each key.

   I would not like to use versions for that. I am thinking about a
separate table "PersonalDetails" with row key = person ID + key + md5
of value. In order to read all values of a given key I would try to
use the Scanner API.

Thank you for your cooperation,
M.


On Tue, Nov 11, 2008 at 8:18 PM, Jonathan Gray <jlist@streamy.com> wrote:
> Michael,
>
> Can you be more specific with what you would like to have?
>
> In what way would you like to be able to store the results, and then fetch
> them?
>
> Would these primitives suffice?
>
> add_name(user, "New Name")
> del_name(user, "Name to Del")
> get_names(user) -> ["Name1", "Name2", ...]
>
> Seems you could do that with cell versions, unless you need versions for
> something else.
>
> Another option would be to create a separate family for names but that would
> be too much if you need to support multi-value for many fields.
>
> I will think more about this... Seems there should be an easy enough
> solution.
>
> JG
>
> -----Original Message-----
> From: rathi [mailto:riteshrathi@gmail.com]
> Sent: Tuesday, November 11, 2008 2:22 AM
> To: hbase-user@hadoop.apache.org
> Subject: Re: Question on Multimap
>
> You can use the label support in hbase where you can add data like
> name:label1 name:label2 ....etc. Label can be added to any column family at
> anytime while adding data without pre-announcing it. In you case it can be
> name:first, name:second etc.
>
> Hope it helps.
>
> rathi
>
> On Tue, Nov 11, 2008 at 3:25 PM, <michael.dagaev@gmail.com> wrote:
>
>> Hi, all
>>
>> Let there is a "People" table and its column family "Personal Details"
>> contains name, address, email etc. For instance, "Personal Details:name"
>> column contains value "Michael". Now I have to support multiple names for
>> the same person. For instance, I want the "Personal Details:name" column
> to
>> contain a list of values: "Michael", "Mike", "Misha", etc. I would not
> like
>> to use timestamps/versions for that. I would not like to implement a list
> as
>> a single string with delimiters (eg "Michael,Mike,Misha" ) either.
>>
>> How would you suggest implementing that?
>>
>> Thank you for your cooperation,
>> M.
>>
>
>

Mime
View raw message