Skip to main content
Announcements
Have questions about Qlik Connect? Join us live on April 10th, at 11 AM ET: SIGN UP NOW
ArturoMuñoz
Employee
Employee

Are you stuck with a complex data model? Do you need some tips to resolve the linkage of different fact tables? Does your model have data knots, circular references or synthetic keys?

If you have ever faced any of these situations you've probably heard about Link Tables and Concatenate as methods that can help you to put some light in your data model. It’s hard to find a general rule to determine when you should Concatenate rather than joining tables by a Link Table so today we will highlight the main differences of both methods by reviewing usage case scenarios.

Lets start with some basic examples:

Concatenate

“This statement forces concatenation with an existing named table or the latest previously created Logical Table. A concatenation is in principle the same as the SQL UNION statement, but with two differences: first that Concatenate prefix can be used no matter if the tables have identical field names or not; and secondly that no removals of identical records are made”

Source: QlikView help

Most basic example of Concatenate usage could be when you need to merge two or more tables that have identical structures, let say you have a historical data warehouse with sales from 2005 to 2012 and then another table with 2013 sales live in you transactional database. To create a visualization of sales trends over the years you will want to have everything normalized in one fact table, Sales table, with the data from 2005 to 2013.

concatenate.png

Link Table(s)

“(…) a junction table is a database table that contains common fields from two or more other database tables within the same database. It is on the many side of a one-to-many relationship with each of the other tables. Junction tables are known under many names, among them cross-reference table, bridge table, join table, map table, intersection table, linking table, many-to-many resolver, Link Table, pairing table, pivot table, transition table, or association table. (…).”

Source: Wikipedia

This definition is mostly about relational SQL databases so if we adapt it for QlikView it could be something like the following.

Link Table: It´s a table that contains common fields from two or more tables (within the same database or not). Easy, right?

The most common scenario for using Link Tables will be to replace synthetic keys and to avoid circular references by joining two or more fact tables against a common set of dimensions.

linktable.png

See much more in the attached files.

AMZ

Enjoy Qliking!

23 Comments
Not applicable

Very nice, thanks specially for the attached document.

0 Likes
11,426 Views
beck_bakytbek
Master
Master

Thanks for sharing

11,426 Views
william_ericksson
Contributor III
Contributor III

Dear Arturo,

Is there a rule of thumb regarding total qvd file size, number of rows, or number of columns that would suggest it would be preferable to implement a concatenated fact table rather than a link table for a data model with multiple fact types and conformed dimensions?

Thanks,
Will

11,373 Views
paulyeo11
Master
Master

Hi William

File size is depend on the project. You can have many row & column also depend on the project .

Better to have link table with many fact table.

Dimension also can have many depend on your project.

But if your GUI have too many sheet , it may affect loading time when open QV.

Paul Yeo

DIrector

TDS Technology (S) P/L

Whatsapp +65 9326 1804

www.tdstech.com<http://www.tdstech.com>

0 Likes
11,373 Views
ArturoMuñoz
Employee
Employee

Hi William,

Not really there's no hard limits here. As stated in the pdf performance gains can be observed when concatenating large tables. In complex data models hybrid solutions are often the way to go.

I particularly always start my data modeling with concatenation because it's simpler and most of the times does the job. If the business use case (extensive use of set analysis is an issue) or if the data model requires it (data at different levels of granularity) I work on link tables as well.

11,373 Views
renjithpl
Specialist
Specialist

link_fact_1.jpg

My both fact tables link to the Link table using Key1, In this case is it good to concatenate it in one table,

as both facts have huge records and different fields.

With current data model i am facing a lot of frontend performance issue.

0 Likes
11,373 Views
trishita
Creator III
Creator III

SYNTHETIC KEY.PNGI want that there is only one link between calender and generalinformation table which would be survey_date

11,373 Views
datanibbler
Champion
Champion

Hi Arturo,

I find your post very interesting indeed. We have some reports already finished here and all of them have a link_table in the DataModel. The datamodels are very complex, there are a lot of tables linked to the linktable, each one with a different key, some tables even have several different keys depending on the type of document ... because of that, we don't always have a 1:n relationship. I'm maybe going to amend that when I have the time - data is only aggregated on the GUI, so it doesn't really matter (as of now) whether a key links to 1 or to 3 records, but it just looks a lot more orderly when the key is unique everywhere ... That will mean I'll have to split some tables into several parts.

11,373 Views
RickWild64
Partner - Creator
Partner - Creator
Hi Trishita,

If you remove SURVEY_DATE_NUM from the table General_Information, the synthetic key will disappear, and the tables will associate using SURVEY_DATE only..

Note I would use SURVEY_DATE_NUM as the key, as I have sometimes had trouble matching date formats between tables.

So, have the field SURVEY_DATE_NUM in General_Information and in Calendar, and have the field SURVEY_DATE only in the Calendar.

0 Likes
11,243 Views
trishita
Creator III
Creator III

rickwild‌ When I tried to remove the survey_date_num from the load script of general informations and tried to load the script it failed to load and showed an error

Field not found error

Field 'SURVEY_DATE_NUM' not found.

The problem is I cannot  keep SURVEY_DATE only in the Calendar.This is because I am using SURVEY_DATE IN General Informations and SoundingCorrections table to join them on the basis of interval Match(SURVEY_DATE,IMO_NO)

0 Likes
11,243 Views