openjpa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Milosz Tylenda (JIRA)" <j...@apache.org>
Subject [jira] Commented: (OPENJPA-1304) Refactor naming policy of database schema elements
Date Tue, 22 Sep 2009 19:08:16 GMT

    [ https://issues.apache.org/jira/browse/OPENJPA-1304?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12758373#action_12758373
] 

Milosz Tylenda commented on OPENJPA-1304:
-----------------------------------------

I expect many users would like quite a lot the possibility to provide their own naming policies.
The JPA defaults probably look awkward to many DBAs.

At a first glance I find the issue quite risky. How about targetting this to the 2.1 release?

I find DBDictionary overfatted and it's getting fatter every month. Maybe it could be desirable
and possible to move as much code as we can to that new DatabaseElementNamingPolicy. Probably
a uni/bi-directional link would need to exist between DBDictionary and DatabaseElementNamingPolicy.
Another idea would be to make DatabaseElementNamingPolicy database-aware (possibly supplying
it with a live connection). That would further lessen the burden from DBDicitonary but I am
not sure it justifies combining policy with database specifics.

Would using the policy need a live connection?

If user specifies a table or column name in anno/xml that is incorrect for the database, do
we modify the name or complain?

The change could also address OPENJPA-84 - use quoting instead of appending digits.

As for identifier lengths we could try getting them from drivers by default and keep the possibility
of configuring them thru properties as a hack for crazy drivers. But that won't work if identifier
lengths need to be known without a connection (offline).


> Refactor naming policy of database schema elements
> --------------------------------------------------
>
>                 Key: OPENJPA-1304
>                 URL: https://issues.apache.org/jira/browse/OPENJPA-1304
>             Project: OpenJPA
>          Issue Type: Improvement
>          Components: jdbc, jpa, kernel, query, sql
>    Affects Versions: 2.0.0-M3, 2.0.0-M4, 2.0.0
>            Reporter: Pinaki Poddar
>            Assignee: Pinaki Poddar
>             Fix For: 2.0.0
>
>
> The issue of naming database elements (Table/Column/Schema/Sequence/Constraints) is getting
trickier with several aspects
>   a) The logic/protocol of naming of database elements is somewhat well-spread at different
parts
>   b) few assumptions are implicit (like full name of a table is a concatenation of schema.table
-- but MySQL, for example, will not like that)
>   c) other 'container' things often cache these elements by their names but these names
may get shortened/modified because of database length restrictions/keyword clash
>   d) database variations impose few other twists (such as case-sensitivity)
> -- all these make naming a complex issue. One can see when methods like MappingDefaults.correctName()
appear -- as if we know we have done mistakes elsewhere :)
>   e) the proverbial straw on the camel's back is the new JPA 2.0 requirement of these
names be optionally quoted/delimited with default/platform specific quote characters.
>    
> The question is should we attempt refactoring name-related operations into a DatabaseElementNamingPolicy?
> Is it too risky?
>  If we do how the policy works with DBDictionary?
> How can we ensure backward compatibility? 
> How can we leverage it by adding new features that follow some rules to map name persistent
Java element names to database element names saving the user to specify column names everywhere?
> Comments/Thoughts?

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message