Re: Really large fetches

From: Andrus (andru..bjectstyle.org)
Date: Wed Jun 19 2002 - 00:47:45 EDT

  • Next message: Andrus: "Re: Really large fetches"

    At 11:36 AM 6/18/2002 -0700, Nikhil Budhiraja wrote:
    >yup , you are right Andrei, ...this is really a common
    >problem and there is no standard design pattern to
    >resolve this.
    >
    >Although, i came across a few frameworks which claim
    >they can resolve this issue but each of them missed
    >one thing or the other.

    I'd suggest we pursue 2 main directions (let me remind that we are
    optimizing 2 things at a time - memory consumption and response times):

    1. "ResultIterator" :

    something in line with what Robert's requirements are. This query would
    return an iterator backed by the open result set. If this is used within
    web applications, it has to have a single request scope (or at least has
    to have a controlling thread that would guarantee closing of the result set
    once the processing is done). This is ideal for report generation on the
    application side.

    2. "ResultPager" :

    this is a fully disconnected results list that would have faults for pages
    and an automated mechanism to resolve them. How it is done is an
    implementation detail. You suggestion is probably good for the data that is
    not modified too often. Otherwise if you do the same query multiple times
    to read data page by page, you will get different data every time (to a
    point that your last record marker may not be there anymore). Also Misha
    mentioned to me some Oracle-specific features that may place a marker on
    the database side. I wonder if anyone heard of something like that?

    Andrus



    This archive was generated by hypermail 2.0.0 : Wed Jun 19 2002 - 00:47:26 EDT