db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Peter Ondruška <peter.ondru...@gmail.com>
Subject Re: How would you implement a record-level versioning system with Derby and Java?
Date Thu, 09 Jul 2009 15:14:20 GMT
I would recommend also looking at some data warehousing articles as
versioning is related to that. Though I do not have any concrete to
recommend :-(. Peter

2009/7/9, Alessandro Bottoni <alexbottoni@yahoo.it>:
> Donald McLean ha scritto:
>> Actually, this is a question that I have previously considered for a
>> project that I am starting.
>>
>> My thought was to have an identity column (value assigned by the DBMS)
>> and a separate ID column. From here, you can do several different
>> things:
>>
>> 1. Have a "version" column. Use the max function to find the current
>> version, or make the current version -1 or something else easy to
>> find.
>> 2. Use the max function on the identity column to find the latest version.
>
> I thought about this mechanism but I'm not completely convinced it would
> be the right one. I mean, there are two different "user views" to take
> into account:
>
> 1) The user view onto the whole database while browsing/searching for a
> specific record.
> 2) The single-record view used to examine/edit a specific record.
>
> While it is easy to use a SELECT with MAX to get the latest version of a
> specific record, it is not clear how I could get only the most recent
> version of every single record to build up the "browse" view. It looks
> like I should use a nested SELECT:  first, select all matching records
> and, second, select the latest version of  each of them. A little bit
> messy...
>
>> 3. Have "previous" and "next" columns with the identity values of the
>> appropriate objects in the chain. The current version would have a
>> null "next".
>
> Nice approach! Actually, I did not think about this possibility. Many
> thanks for this suggestion.
>
> This would simplify a lot the SELECT operation used to generate the
> "browse" view (just select all of the records that have a NULL
> "next-version" field) while maintaining the possibility to go back and
> forth between versions (and build up a list of them, if needed).
>
> I think I will experiment with this one.
>
>> There are probably other approaches that I haven't thought of.
>
> I thought about triggers/procedures, as well, but I had the feeling they
> would have the same problem as the SELECT + MAX approach.
>
>> I have used and recommend Hibernate. It simplifies things immensely
>> IMO, though it does have its quirks (some of which are not properly
>> explained by the documentation). It also works well with a number of
>> other DBMS (if changing is something that might happen later).
>
> Thanks. I'm new to Hibernate, as well, so I'm going to ask you (all)
> which book is best for learning Hibernate. I browsed a little bit on
> Amazon's but I did not find anything to be enthusiastic about. What
> would you suggest me as a good starting point/introductory book?
>
> Any good web article/tutorial to signal, as well?
>
> Thanks again.
> --
>
> Alessandro Bottoni
> Website: http://www.alessandrobottoni.it/
>
> "In mathematics you don't understand things. You just get used to them."
>      -- John von Neumann
>

Mime
View raw message