[sqlalchemy] Re: oracle speed problem and arraysize hack

2008-05-30 Thread Michael Bayer
the arraysize=50 default is set up in r4827 on the 0.4 branch and r4828 on the trunk. I was surprised to find the rules for BLOB no longer accesssible are different than what I had assumed they were, so binary results and all work just fine here. On May 27, 2008, at 5:34 PM, gniquil

[sqlalchemy] Re: oracle speed problem and arraysize hack

2008-05-27 Thread Waldemar Osuch
On May 26, 5:09 pm, Michael Bayer [EMAIL PROTECTED] wrote: On May 26, 2008, at 6:20 PM, Waldemar Osuch wrote: Reading the docs for arraysize, its not clear if this only applies to fetchone() and a non-arg fetchmany(), or to all fetches (I'm hoping you have better detail on this.). If

[sqlalchemy] Re: oracle speed problem and arraysize hack

2008-05-27 Thread Michael Bayer
On May 27, 2008, at 3:21 PM, Waldemar Osuch wrote: Quote: Up to this point the default arraysize is 1 meaning that a single row is internally fetched at a time. This has nothing to do with fetchone(), fetchmany() or fetchall(). Regardless of which of those methods is used, internally

[sqlalchemy] Re: oracle speed problem and arraysize hack

2008-05-27 Thread gniquil
Hey, i did a test included below with this output: $ python cxotest.py Setting arraysize to 1 before execution yields 25.8921508181 Setting arraysize to 500 before execution yields 0.26524348765 Setting arraysize to 1 after execution yields 25.8829982582 Setting arraysize to 500 after execution

[sqlalchemy] Re: oracle speed problem and arraysize hack

2008-05-27 Thread Michael Bayer
So, I think i'll probably implement the patch I posted earlier and actually default arraysize to 50, although I have a feeling it might mess around with some of the BLOB-oriented functionality (since BLOBs must be fetched at the point of receiving the row, else the cursor moves on)..if