Get short, timely messages from Mark S. Rasmussen.

Twitter is a rich source of instantly updated information. It's easy to stay updated on an incredibly wide variety of topics. Join today and follow @improvedk.

Get updates via SMS by texting follow improvedk to 40404 in the United States
Codes for other countries

Country	Code	For customers of
Two-way (sending and receiving) short codes:
Australia	0198089488 Telstra
Canada	21212 (any)
United Kingdom	86444 Vodafone, Orange, 3, O2
Indonesia	89887 AXIS, 3, Telkomsel
Ireland	51210 O2
India	53000 Bharti Airtel, Videocon
Jordan	90903 Zain
New Zealand	8987 Vodafone, Telecom NZ
United States	40404 (any)

improvedk

@AdamMachanic Oh, and thanks by the way :) about 4 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic Sorry, I will try to reduce my usage. In this case it was truly agonizing though :) about 4 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic Would you agree in the conclusion in the last two paragraphs (no need to read rest)? bit.ly/KuGFQA about 4 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic Just tried adding to both, no change. Column is nullable on Z but not on X. about 5 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic Let me post the full query + stats on DBA so it's easier to show what I've got. 5 min. about 5 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic Makes sense yes - and Y is nullable on Z. However, I've got a WHERE Y IS NOT NULL predicate in the inner query though. about 5 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic a large row count discrepancy between the tables (500 vs 50 million). Y being the primary key of X. 2/2 about 5 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic That does the trick! Interesting. So because there's no cardinality stats for Y, it assumes the worst, even with such 1/2 about 5 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic That gives the exact same execution plan - probably being rewritten to the same internally. I'll post plans & query on DBA. about 5 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic And you can assume ~500 rows in X, about ~50 million in Z - thus those 500 scans of Z are somewhat expensive. about 5 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic Instead of doing a RBAR scan on Z for each row in X. about 5 hours ago via TweetDeck in reply to AdamMachanic
@AdamMachanic Assume no indexes on Y in Z. A scan is required. I just don't see why it wouldn't create a worktable and use that temporarily. about 5 hours ago via TweetDeck in reply to AdamMachanic
@improvedk And I of course meant a scan of Z, not Y. #sqlhelp about 5 hours ago via TweetDeck in reply to improvedk
DELETE FROM X WHERE Y NOT IN (SELECT Y FROM Z) <- Why does that cause a scan of Y for each row in X? Wouldn't a worktable do it? #sqlhelp about 5 hours ago via TweetDeck
@MladenPrajdic Just as you can package HTML/JS to a native app using PhoneGap, you can embed LUA in a native wrapper. about 5 hours ago via TweetDeck in reply to MladenPrajdic
@MladenPrajdic No, but LUA interpreters come as libraries that you can embed inside apps that run natively. Flame contained a LUA VM. about 5 hours ago via TweetDeck in reply to MladenPrajdic
@MladenPrajdic That's no different from SQL, JavaScript, Java, .NET, etc. It won't run natively but if you've got the environment, sure. about 5 hours ago via TweetDeck in reply to MladenPrajdic
@MladenPrajdic Didn't I tell you to stay away from shrooms? about 6 hours ago via TweetDeck in reply to MladenPrajdic
@tonyrogerson Ah, I see. May God Save the Queen :) about 11 hours ago via TweetDeck in reply to tonyrogerson
@tonyrogerson Wait, it isn't? about 12 hours ago via TweetDeck in reply to tonyrogerson

Name Mark S. Rasmussen
Location Aarhus, Denmark
Web http://improve.dk
Bio Has a secret love affair with SQL Server internals. Author of the OrcaMDF project. Speaker, consultant, SQL Server & C# groupie. Tech Lead @ iPaper A/S.