[DBD-SQLite] collation_needed() : new implementation

[Dami Laurent (PJ)]

Hi folks,

For info, I just committed a new implementation for "collation_needed",
that no longer uses dynamic memory allocation, and therefore should no
longer leak. The extra information needed is just an additional field in
the imp_dbh_st struct (this makes sense because at any point in time, a
dbh cannot have more than one "collation_needed" handler -- I should
have realized that in the first place!).

To address Adam's argument about global registry, I first thought of a
bunch of methods "register_global_collation", "list_global_collations",
etc., that would hide the internal hash implementation. But the API
became heavy, and finally it seemed  more perlish to expose that stuff
as a regular hash, so that clients can use all Perl idioms for hashes
--- except that this is a "write-once" hash : any attempt to modify or
delete an existing entry raises an exception. Hackers who for any
obscure reason would need to override an existing entry can always use
the 'tied' API -- but then there is really no risk of doing it
accidentally.

The doc has been adapted in consequence.

I hope these solutions are satisfactory to everyone.

Greetings, Laurent Dami
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.scsys.co.uk/pipermail/dbd-sqlite/attachments/20090806/130ab3ee/attachment.htm