Brian Aker wrote:
> Hi!
>
> On Aug 3, 2008, at 10:59 PM, Mats Kindahl wrote:
>
>> Although I don't disagree, what do you consider as an alternative to the
>> THD for keeping session data around?
>
> THD is pretty much session data now (especially since it is no longer
> tied thread).
Cool, but does this mean the documentation is out of date?
/**
@class THD
For each client connection we create a separate thread with THD serving as
a thread/connection descriptor
*/
If that comment is outdated, let us please correct! :)
Also, THD currently inherits from Statement, which represents a single
query. Wouldn't a more appropriate relationship be one of THD
encapsulating a linked list of Statement instances, each representing a
particular query being executed in the client connection?
Similarly, there is a lot of replication stuff inside THD itself. Could
this stuff be refactored out of THD into a true replication API. This
would allow developers working on replication to blackbox the
replication functionality away from developers modifying/refactoring THD
down to a more reasonable size. As long as there are accessor methods
to THD's member variables that the replication engine needs to be aware
of, I don't see a need to have THD methods like
set_current_stmt_binlog_row_based() and binlog_query(). In my mind, the
replication API should have methods to register a callback on Statement
execution which examines the query and performs necessary replication
functionality if the Statement meets requirements.
I guess I am envisioning a THD that would essentially be responsible for
doing the following (simplified) execution path:
Manage/facilitate NET communication
-> Pull in all statements from client
-> Create a Statement for each statement recv'd from client, pop onto
stack
-> Execute each statement
-> Send result(s) of statement back to client
-> Update counters/state
All other functionality would be done via callbacks notified at various
state changes in the Statement execution path...
Am I out of my mind about this?
> The biggest problem I now see with it is getting it properly encapsulated.
Yep.
> Cheers,
> -Brian
>
> --
> _______________________________________________________
> Brian "Krow" Aker, brian at tangent.org
> Seattle, Washington
> http://krow.net/ <-- Me
> http://tangent.org/ <-- Software
> _______________________________________________________
> You can't grep a dead tree.
>
>
>
>
> _______________________________________________
> Mailing list: https://launchpad.net/~drizzle-discuss
> Post to : [email protected]
> Unsubscribe : https://launchpad.net/~drizzle-discuss
> More help : https://help.launchpad.net/ListHelp
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
