Todd G. Nist wrote:
> Hello all,
>
> I have a question with regards to the update method of the BaseDAO class
> found in the WebPIM application. As it is currently written it would appear
> to allow for writes of dirty data. That is to say that if two users read
> the same record, lets say company record, and user1 changes the company
> field to "XYZ" and User2 changes it to "AAAAA", if user2 commits his record
> and then user1 commits his, user2's changes will be lost. No exception is
> thrown since the update just does an update where the id is equal to the
> value of the id of the crs.getBigDecimal("id") of the current row.
>
> Am I missing something here or is this an issue/bug with the current
> implementation? If the update was being applied to the CRS and then the
> crs.acceptChanges() method applied, the CRS would handle comparing the
> contents of the CRS to the actual db prior to committing the changes, but
> since a new connection is obtained and a standard update applied, there is
> no support for concurrency. It seems that there either needs to be a
> version field on all tables which is included in the update statement of the
> prepareUpdateSQL method which does something like below:
>
> UPDATE theTable SET version=<newversion> [and other fields] WHERE
> [criteria] AND VERSION=<oldversion>
>
> where <oldversion> is the version is the version from the crs current row,
> and
> <newversion>=<oldversion>+1.
>
>
> In this way if two users update the same record of a table at roughly the
> same time, the first ones changes will be committed, but the second ones
> will generate an error which can be caught and the end user notified of for
> corrective action.
>
> Any thoughts on this?
You are correct! It should check fields and it will.
>
> I have modified the BaseDAO to support multiple tables rather then one as it
> currently does. This opens up another issue of transactional integrity. If
> a crs is created from a join between two or more tables, one needs to start
> a transaction to support the individual updates as one unit of work. I
> have added methods to support transactions to ensure that if the crs has
> data from more then one table that either all changes are committed or none
> are.
>
> Any thoughts on this?
>
You are correct! (alternative is a bean that contains a bean that each
have DAO)
If you would like to contribute source to the DAO with several other
developers, I have added it to sourceforge under basejDAO. I will post
the initial code. Just add yourself to the developers list.
> Regards,
>
> Todd G. Nist
>
>
_______________________________________________
MVC-Programmers mailing list
[EMAIL PROTECTED]
http://www.basebeans.com:8081/mailman/listinfo/mvc-programmers
