 |
|
Merge multiple data sources with SQL Server Integration Services
|
By: Greg Robidoux | Read Comments (26) | Related Tips: More > Integration Services Data Flow Transformations |
Problem
When loading data into SQL Server you have the option of using SQL Server Integration Services to handle more complex loading and data transforms then just doing a straight load such as using BCP. One problem that you may be faced with is that data is given to you in multiple files such as sales and sales orders, but the loading process requires you to join these flat files during the load instead of doing a preload and then later merging the data. What options exist and how can this be done?
Solution
SQL Server Integration Services (SSIS) offers a lot more features and options then DTS offered. One of these new options is the MERGE JOIN task. With this task you can merge multiple input files into one process and handle this source data as if it was from one source.
Let's take a look at an example of how to use this.
Here we have two source files an OrderHeader and an OrderDetail. We want to merge this data and load into one table in SQL Server called Orders.
OrderHeader source file.
OrderDetail source file
Orders table
Building the SSIS Package
First create a new SSIS package and create the three Connections that we will need.
- Flat File Source 1 - OrderHeader
- Flat File Source 2 - OrderDetail
- OLE DB Destination - SQLServer
Then add a DATA FLOW task.
Next we need to build our load from these two flat file sources and then use the MERGE JOIN task to merge the data. So the Data Flow steps would look something like this.
At this point if you try to edit the MERGE JOIN task you will get the below error. The reason for this is because the data needs to be sorted for the MERGE JOIN task to work. We will look at two options for handling this sorting need.
Option #1 - Data is presorted prior to loading the data.
Let's assume that are data is sorted prior to loading. We therefore need to tell SSIS this is the case as well as show which column the data is sorted on. First if you right click on "Flat File Source" and select the "Show Advanced Editor". On the Input and Output Properties tab you need to change the "IsSorted" to True for both of the Flat File Sources.
Next you need to let SSIS know which column is the SortKey. Here we are specifying the OrderID column. This also needs to be done for both of the flat file sources.
Once this is complete you will be able to move on with the setup and select the input process as shown below.
From here you can select the columns that you want to have for output as well as determine what type of join you want to employ between these two files.
Lastly you would need to add your OLE Destination, select the table and map the columns to finish the process.
Option #2 - Source data is not sorted
With this load process, let's assume the source data is not sorted first, so we need to use the SORT task to sort the data prior to using the MERGE JOIN task. The following shows our Flat File sources and then a SORT task after each one of these and then lastly our MERGE JOIN task.
If you right click the Sort task and select Edit you will get a screen such as following. Here you need to select which column the data should be sorted on. This needs to be done for both of the flat source files.
After this is done you can move on and finish the load process. The MERGE JOIN works just like it was stated above as well as the OLE DB Destination.
Next Steps
- As you can see this process is not all that complicated, but not knowing where and how to set the IsSorted parameter could get frustrating.
- This gives you two different options that you can use to employ the MERGE JOIN task. The ideal process would be to presort the data, so this extra step can be eliminated when loading your data. Also, if the files are large the Sort task could be a time consuming task.
Last Update: 9/5/2007
About the author
Print
Comments and Feedback:
| Friday, March 19, 2010 - 10:32:21 AM - darck1 | Read The Tip |
|
Is there a way I can merge join without it first looking at every record in each datasource? If I have over 1,000,000 records in the left table and 34 in the right table but the join is the only method of selection then I don't want to have to process a million records. |
|
| Thursday, September 09, 2010 - 4:01:13 PM - Netty | Read The Tip |
|
EXTREMELY well written! Thanks for sharing! :) |
|
| Wednesday, February 16, 2011 - 10:07:50 AM - swingnchad | Read The Tip |
|
Nice intro to BIDS and SSIS! This is a very good place to start. |
|
| Friday, March 11, 2011 - 5:46:41 PM - Anonymous | Read The Tip |
| Sunday, April 15, 2012 - 7:42:11 PM - Shahid | Read The Tip |
|
Very well written. Screenshots are really very helpful to easily understand. Very well done (Y) |
|
| Saturday, April 28, 2012 - 4:14:48 PM - td | Read The Tip |
|
The sorting stage seems to take forever on 4 million rows on each file. Is that sound normal? Is there a better way to do this? How about using for loop or the multi flat.file container? thoughts? |
|
| Thursday, June 28, 2012 - 3:27:04 AM - Meinathan | Read The Tip |
|
Very good.... Thanks for sharing. |
|
| Tuesday, July 10, 2012 - 5:04:20 AM - Robert Bosco JOHN PETER | Read The Tip |
|
I want to merge multiple flat file data and load into multiple table in SQL Server. Could you please advice me. Have a nice day |
|
| Tuesday, July 10, 2012 - 8:33:04 AM - Greg Robidoux | Read The Tip |
|
Robert Bosco - you should be able to follow the process above to do what you are asking. |
|
| Tuesday, July 10, 2012 - 9:48:40 AM - Robert Bosco JOHN PETER | Read The Tip |
|
Greg Robidoux - I understood the concept that you explained above and I had also followed. In OLE DB Destination part, I can have selected either only one table or one view but my question is that I ought to be use multiple destination tables. |
|
| Tuesday, July 10, 2012 - 11:15:34 AM - Greg Robidoux | Read The Tip |
|
Robert Bosco - see if you can use a Multicast task after the Merge Join task. |
|
| Thursday, July 12, 2012 - 1:34:06 PM - AJ | Read The Tip |
|
How can I transfer only the data are in flat source but not in flat source 1? I mean if I had a row in OrderHeader like this : order id order date ... 5 7/12/12 ... that this order id (5) is not in Order detail table. but I want transfer only this record (order id = 5) to detination. |
|
| Monday, July 23, 2012 - 1:29:42 AM - vijay | Read The Tip |
|
Very Nice artical for basic learners.... easy to understand |
|
| Friday, August 10, 2012 - 4:07:43 PM - Jason | Read The Tip |
|
Excellent tip. |
|
| Friday, November 23, 2012 - 4:05:30 PM - mita | Read The Tip |
|
I am trying to create a single row from two selects using merge join in SSIS and the join key is SSN from two selects. Can it be done using this example ? |
|
| Monday, November 26, 2012 - 7:17:36 AM - Greg Robidoux | Read The Tip |
|
@mita - yes you should be able to join two sets of data based on SSN. Just make sure the data is sorted by SSN. |
|
| Saturday, March 09, 2013 - 8:11:32 PM - kerany | Read The Tip |
|
@Robert Bosco, you can use the conditional split component to load your data into different tables with condition |
|
| Thursday, March 21, 2013 - 3:53:03 AM - Nagaraj G | Read The Tip | ||
|
|||
| Friday, April 12, 2013 - 2:58:50 AM - Arun | Read The Tip |
|
Hi,
I am new to SSIS, is it possible to copy data from 4 source tables to one destination table using SSIS? I can see from 2 to 1 is possible.
Thanks. |
|
| Friday, April 12, 2013 - 9:49:47 AM - Greg Robidoux | Read The Tip |
|
@arun - you should be able to do multiple merges to do more than 2 sources. |
|
| Friday, April 12, 2013 - 2:56:47 PM - Arun | Read The Tip |
|
@greg - thanks for your response. Could you elaborate on that please? I am trying to use merge join, and for that sorting all the data in all the tables and their resultant output, consequently, it eats up a lot of time and ends up complaining about buffer. The tables that I read from for instance have all got half a million records. |
|
| Friday, April 12, 2013 - 4:19:29 PM - Greg Robidoux | Read The Tip |
|
@arun - are these tables all SQL Server tables or are they from some other source? If they are all SQL Server tables you could just write a SQL statement joining the tables and not use the MERGE task. |
|
| Tuesday, April 23, 2013 - 6:10:57 AM - Ramesh | Read The Tip |
|
Hi Friends, Am New to SSIS i have a doubt in merge join we have one source(like emp table which is oledb source) and one transformation(like Dept table which is a lookup transformation) and i have one destination (which is oledb destination). now my Requirement is can i use MERGE JOIN in place of Transformation. How can i achieve this. |
|
| Tuesday, April 23, 2013 - 10:06:06 AM - Greg Robidoux | Read The Tip |
|
@Ramesh - if both of these tables are in the same database then you can just do a join instead of using an SSIS task. It is hard to tell exactly what you need to do from your description. |
|
| Wednesday, April 24, 2013 - 6:37:20 AM - Ramesh | Read The Tip |
|
Thanks for that suggestion, but in my lookup transformation haveing reference data of Dept table, Am New to SSIS i have a doubt in merge join we have one source(like emp table which is oledb source) and one transformation(like Dept table which is a lookup transformation this transformation having reference data of dept) and i have one destination (which is oledb destination). now my Requirement is can i use MERGE JOIN in place of Transformation. different databases How can i achieve this. |
|
| Wednesday, April 24, 2013 - 7:20:33 AM - Greg Robidoux | Read The Tip |
|
@Ramesh - yes you can use a merge join if you want. Just follow the example above and substitute your Dept table for the OrderHeader and your other table for OrderDetail. |
|
|
privacy | disclaimer | copyright | advertise | about authors | contribute | feedback | giveaways | user groups Some names and products listed are the registered trademarks of their respective owners. Edgewood Solutions LLC | MSSharePointTips.com | MSSQLTips.com |