June 21, 2008 at 11:12 am
Comments posted to this topic are about the item T-SQL Data Processing
June 23, 2008 at 7:15 am
I may be missing something here, but why use a While loop for this kind of thing? Do a set-based merge of all the update data, then upsert it into the master table. Two steps, very simple, very clean, no While loop.
- Gus "GSquared", RSVP, OODA, MAP, NMVP, FAQ, SAT, SQL, DNA, RNA, UOI, IOU, AM, PM, AD, BC, BCE, USA, UN, CF, ROFL, LOL, ETC
Property of The Thread
"Nobody knows the age of the human race, but everyone agrees it's old enough to know better." - Anon
June 23, 2008 at 8:28 am
I don't like dedupe, let me tell you that 🙂
But I can't imagine how this WHILE loop would work on millions of rows of data comparison
I had issues with comparing between 400K records vs 8000 inserts using SET operations (still acceptable if I separate by country, etc..)
started looking to SSIS Fuzzy Matching and Fuzzy Lookup
Fuzzy Matching - It's pretty cool, dedupe within say the Master table without any T-SQL work, takes a while but it even gives confidence score
Fuzzy Lookup - I am still working on it, supposedly I can lookup those 8000 inserts in the 400K Master table
June 23, 2008 at 9:33 am
I'm not a fan of dedup either, and didn't expect a harty round of approbation from this article...
I've found that you should ALWAYS consider your alternatives when it comes to SQL programming. Things frequently don't "work" as we think they "should".
I use this code to dedup databases of over 1 billion (yep, the B word) rows, a quarter of which (~250,000,000) are duplicates of some kind (exact or based on some "fuzzy" logic). One database processes in about 3 days, the other in about 18 hours (one has a larger record size than the other).
This represented a significant reduction in processing times for both these databases over the previous methodology written using a hybrid of external and internal (SQL w/cursors) coding; and its all done in SQL.
I am constantly looking for betterprocessing techniques that perform the required functions AND run quicker than existing procedures. So far, this is it.
As for SSIS, I will look into it, having not used it before.
June 23, 2008 at 7:36 pm
Small suggestion: maybe I missed a reference to this, but wouldn't the Bit datatype be perfectly suited to the Data Source column (one bit column per source)? No need for explicit bitwise operations, acheives the same thing as you're doing against a TinyInt under the covers and the resulting code would be a little more human-readable...?
Regards,
Jacob
June 23, 2008 at 11:19 pm
This process was designed on a SQL 2000 system and ported to SQL 2005 with minimal changes. When I refactor it for SQL 2005 that would be the way to go.
June 25, 2008 at 9:16 am
Select
PartyId,
FirstName,
LastName,
Case when Source&1<>0 then 1 else 0 End[Source1],
Case when Source&2<>0 then 2 else 0 End[Source2],
Case when Source&4<>0 then 4 else 0 End[Source3],
AddDate,
ModDate
from dbo.ExistingMaster
Will you please let me know what Case when Source&1<>0 then 1 else 0 for in this code ?
Thx.
June 25, 2008 at 10:22 am
Our current database represents data source as a TinyInt column containing bit values to indicate source type; 1=Source1, 2=Source2, 4=Source3, etc... where the decimal numbers 1, 2, & 4 represent binary bits 00000001, 00000010, 00000100, respectively.
In order to process these values within then constraints of SQL (SQL not having an aggregate OR function) I break the TinyInt into 3 separate columns Source1, Source2 & Source3 (as seen in the code). The 3 case statements, you've identified, serve this purpose.
June 25, 2008 at 12:01 pm
Source is the column name. I still do not understand Source&1 or Source&2 here.
June 25, 2008 at 2:03 pm
Source is the name of a TinyInt column in the table.
Source&4<>0 is a method of determining a binary bit's value, in this case the 3rd bit from the right.
Source&4 is a boolean AND operaration using 4 as the mask, so if:
Source = 00000111 (decimal 7)
opcode = &
(mask) = 00000100 (decimal 4)
--------------------------------
yields 00000100
which is not equal to zero therefore the CASE statement would return 1.
July 7, 2008 at 9:00 am
Frances L (6/25/2008)
Source is the column name. I still do not understand Source&1 or Source&2 here.
For enlightenment, read this article and then wade through the discussion:
http://www.sqlservercentral.com/articles/Miscellaneous/2748/
John
July 30, 2008 at 2:02 pm
Great article! very cool!
Viewing 12 posts - 1 through 11 (of 11 total)
You must be logged in to reply to this topic. Login to reply