_{(source: citypages.com)}

I was trying to find an image in google, that I knew I had attached to a meta post a while ago. I utterly failed to find it using a keyword search, but I was able to find it on the cloudapp sandbox by searching with the query select Id, ParentId, Body from Posts where OwnerUserId = 132636 and Body like '%img%';

(For the curious, the post was here -- now deleted, so searching on meta itself also failed to find it.)

It choked on a simple select * from votes query.

Here's an old version of the procedure I used to run queries on StackQL:

CREATE PROCEDURE WebQuery 
@QueryText nvarchar(1000),
@IPHash binary(16)
AS
BEGIN

   --Log query
   INSERT INTO Logs (QueryText, IPHash, QueryHash) VALUES (@QueryText, @IPHash, HashBytes('md5', LTrim(RTrim(Upper(@QueryText))))

    -- no writing allowed, so no need to lock on select
   SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED 

   -- throttles
   SET ROWCOUNT 500 
   SET QUERY_GOVERNOR_COST_LIMIT 500

   exec (@QueryText)

END

Note the code for throttling (rowcount and cost_limit). I was never happy with this:

• the cost limit control never really worked right - I either had to turn it too high to be at all effective or it was so low normal queries would keep bumping into it.
• it's pretty easy to get past the throttling by including your own set commands in the query, but at least then I still have a log so I could check and see who is breaking my controls and why.
• I played with the row limits, and anything > ~1500 performance problems

The query hash is a naive attempt to make it easier to spot identical queries, so I could do things like group by the hash and show counts. New version is instead making more of an emphasis on re-using old queries, so that you can plug in parameters without breaking the hash and I can update a count in the log directly.

The CSV export seems to broken, its giving out part CSV, part something(JSON!?) output.

Sample output:

User Link,Reputation,Days,RepPerDays
{
  "title": "John T",
  "id": 1931
},56421,288,195
{
  "title": "Wil",
  "id": 4386
},35724,270,132
{
  "title": "unknown",
  "id": 8672
},23661,246,96
{
  "title": "nik",
  "id": 263
},20317,289,70
{
  "title": "quack quixote",
  "id": 12786
},13409,212,63

Link to query: http://odata.stackexchange.com/superuser/q/946

One word: STEXDEX.

Very nice one.

Here is the one I have tested for users pages, limit 35 on each result:

DECLARE @start INT
DECLARE @limit INT
DECLARE @page INT

SET @page = 1
SET @start = 1+35*(@page-1)
SET @limit = 35

SELECT * FROM
(
    SELECT row_number() OVER (ORDER BY Reputation DESC) AS i, *
    FROM users
) AS T
WHERE T.i BETWEEN (@start) AND (@start + @limit) 

And I've tried to log in with my google account but I am getting the following error:

Sorry, an error occurred while processing your request.

The current result display is fairly clean and generally OK, but it doesn't play well with Chrome's web page search. It looks like it hides the text that has scrolled off the screen -- good for some optimisations, but it means you can't search within results.

Not a problem that I can see, but IE8 is reporting this at the moment:

Webpage error details

User Agent: Mozilla/4.0 (compatible; MSIE 8.0; Windows NT 6.0; Trident/4.0; SLCC1; .NET CLR 2.0.50727; MDDC; .NET CLR 3.5.30729; .NET CLR 3.0.30729) Timestamp: Sun, 30 May 2010 11:09:10 UTC

Message: Invalid argument. Line: 210 Char: 11 Code: 0 URI: http://odata.stackexchange.com/Scripts/codemirror/select.js

NOTE It occurred while I was editing and executing a query, but rerunning the query for meta and then for SO again and I can't repro it :-(