Igor Tandetnik wrote:
Thomas DILIGENT <[EMAIL PROTECTED]> wrote:
What about the performance issue ?
What about it?
You need to provide some context. Your questions, standing alone,
don't make much sense, no offence intended. What problem are you
trying to solve, and what specifically are you concerned about?
The program I am working on performs the following steps:
- transform events from a trace into persistent (C++ and java) objects
that are stored to and load from a sqlite database.
- query objects for analysis.
Only one client accesses this database in the same time, i.e. we do not
have any synchronisation issue.
As queries are performed on object's load and store, the sequence of
queries depends on the usage of these objects, i.e. the incoming events.
In this context, I try to improve speed of database accesses :
Instead of using autocommit, I would like to start and end transactions
which would contain a fixed number (as an example) of transactions.
First question : Will transactions increase speed ?
In the first step, objects are stored and may be queried to retrieve
some information.
As a result, since objects load and store are performed regarding
incoming events, queries are performed in an undeterministic way.
Thus I have to define a strategy that hooks object load and store in
order to perform begin and commit at the judicious time.
Second question : When to perform begin and commit ?
In this context, is it realistic and judicious to perform begin at the
beginning of the first step and commit at the end of the first step ?
Hope this is clear enough ...
Thomas
Igor Tandetnik
-----------------------------------------------------------------------------
To unsubscribe, send email to [EMAIL PROTECTED]
-----------------------------------------------------------------------------
-----------------------------------------------------------------------------
To unsubscribe, send email to [EMAIL PROTECTED]
-----------------------------------------------------------------------------
