Qlik Community

QlikView App Development

Discussion Board for collaboration related to QlikView App Development.

Announcements
BARC’s The BI Survey 19 makes it official. BI users love Qlik. GET REPORT
Partner
Partner

[Perfect Key]?

Hi all,

Please see this screenshot from my data model:

PerfectKey.gif

Can anyone please explain what "Perfect Key" means?  Not every key has it.  I do like to think my model is perfect ...

Thanks,

Jason

1 Solution

Accepted Solutions

Re: [Perfect Key]?

Hi Jason,


That means that the field is populated with 100% different unique values. That's not always the case because the key value can be more than once in the same field (for example, in a 1:n relation i. e.: invoice header - invoice details).

Hope that makes sense.

Miguel

View solution in original post

15 Replies

Re: [Perfect Key]?

Hi Jason,


That means that the field is populated with 100% different unique values. That's not always the case because the key value can be more than once in the same field (for example, in a 1:n relation i. e.: invoice header - invoice details).

Hope that makes sense.

Miguel

View solution in original post

Partner
Partner

Re: [Perfect Key]?

Ah - that would make sense!

Thanks Miguel.

Jason

francescopuppin
New Contributor III

Re: [Perfect Key]?

Hi Miguel,

I have not found these definitions on the reference manual, therfore I have decided to find the answer on my own, but it seems to be different from your answer.

I believe that the definition that you have given is actually the definition of [Primary Key].

The [Perfect Key] is a bit more: it is a [Primary Key] (unique) with no "orphans".

Let me clarify this with 2 examples:

Example 1)

InvoiceDetails:

LOAD *

INLINE [

InvoiceID, InvoiceLine, ProductID, Quantity, Amount

Inv001,1,PR01,10,50

Inv001,2,PR02,10,40

Inv002,1,PR01,30,150

Inv002,2,PR03,10,800

];

Invoices:

LOAD *

INLINE [

InvoiceID, ClientID

Inv001,CL0001

Inv002,CL0002

];

In this example, InvoiceID is [Key] in InvoiceDetails and [Perfect Key] in Invoices: it is unique and it has no orphans.

Example 2)

InvoiceDetails:

LOAD *

INLINE [

InvoiceID, InvoiceLine, ProductID, Quantity, Amount

Inv001,1,PR01,10,50

Inv001,2,PR02,10,40

Inv002,1,PR01,30,150

Inv002,2,PR03,10,800

Inv003,1,PR01,10,50

];

Invoices:

LOAD *

INLINE [

InvoiceID, ClientID

Inv001,CL0001

Inv002,CL0002

];

In this example, InvoiceID is [Key] in InvoiceDetails and [Primary Key] in Invoices. There is a value in InvoiceDetails, Inv003, that does not find a corresponding value in the Invoices table. This is called a "orphan". For this reason InvoiceID is not anymore perfect, but it is still primary, because it is unique.

If you copy-paste the example you can verify it

I hope that it all makes sense to you.

Jason, if you both agree, I would kindly ask you to mark this post as correct answer. this would perhaps help other developers. I have been helped a lot so far, and i would be very happy to begin to give my small contribution.

Regards

Francesco

Re: [Perfect Key]?

Hi Francesco,

Thanks for chipping in and for sharing your example.

In QlikView, once the tables have been loaded, key fields are actually ONE field (one listbox), not two fields as it happens in a relational model. In your example 2, the field InvoiceID is not 100% populated with unique values, because it has repeated values and there is no full correspondence between both tables, therefore the primary key and the key.

That's why I mentioned the 1:n relationship. Or to put my own post in different words: if the field is 100% populated with unique values, and these values have a relationship of 1:1 (in your example 1, InvoiceID), then is a perfect key, otherwise is a primary key or a key.

I don't see any differences here between what we both say, in your example 1, InvoiceID is a perfect key because all values in that field alone are unique, and the field is populated 100% in the model, isn't it?

If values are not unique, then there is not a 1:1 relationship, that means, if values are 1:n then there is no possible "unique" values, because they are not unique but repeated!

Hope this makes sense as well.

Miguel

P.S.: I'd strongly recommend you to create a Document where you explain with as much detail as possible the findings on keys, even attaching a small sample QVW, so people can download it or print it as PDF.

francescopuppin
New Contributor III

Re: [Perfect Key]?

Hi Miguel,

your definition of perfect key and primary key is different from the definition of Qlikview.

Please follow these steps:

1) Copy this text

InvoiceDetails:

LOAD *

INLINE [

InvoiceID, InvoiceLine, ProductID, Quantity, Amount

Inv001,1,PR01,10,50

Inv001,2,PR02,10,40

Inv002,1,PR01,30,150

Inv002,2,PR03,10,800

];

Invoices:

LOAD *

INLINE [

InvoiceID, ClientID

Inv001,CL0001

Inv002,CL0002

];

2) Paste it in a new Qlkview document, save it and run it

3) Go to the Table Editor and put the mouse on the field invoice ID of the table Invoices

4) You will notice the evidence: Qlikview says that it si a perfect key

5) Read again your phrase if the field is 100% populated with unique values, and these values have a relationship of 1:1 (in your example 1, InvoiceID), then is a perfect key, otherwise is a primary key or a key.

6) Observe the data: it is a 1:n relationship.

Now, the field invoice ID of the table Invoices is 100% populated with unique values, but it is on a 1:n relationship. Hence, according to your definition, it is a Primary Key. But according to Qliktech it is a Perfect key. Now we have just one choice: should we trust you or Qliktech?

So, I hope that it is clear now.

The definition of Perfect key is the following: A Perfect Key is a key populated with unique values, and it must have ALL the values that are in the associated fields. Or, to say the same concept in a Qlikview terminology, the Perfect key is a Primary Key with Subset Ratio = 100%.

If yo uhave time, please run also the example 2: you will notice that it's downgraded to primary key, because it does not have the value Inv003, therefore now the subset ratio is 67% (which means 2 values out of the 3 existing)

I really hope that you will correct your definitions now

Have a nice day

F.


francescopuppin
New Contributor III

Re: [Perfect Key]?

Oh my God, I have a terrible suspect.

Miguel, please let's clarify. The relation between the 2 tables, in the example 1 (here below, again), in your opinion, is it 1:1  or is it 1:n?

Because.. if you are saying that it is 1:1, then we have a big problem 😞

InvoiceDetails:

LOAD *

INLINE [

InvoiceID, InvoiceLine, ProductID, Quantity, Amount

Inv001,1,PR01,10,50

Inv001,2,PR02,10,40

Inv002,1,PR01,30,150

Inv002,2,PR03,10,800

];

Invoices:

LOAD *

INLINE [

InvoiceID, ClientID

Inv001,CL0001

Inv002,CL0002

];

Not applicable

Re: [Perfect Key]?

Hi Francesco,

Thanks for your answer.

Please  find the my load script,

Invoices:

LOAD *

INLINE [

InvoiceID, ClientID

Inv001,CL0001

Inv002,CL0002

         ,CL0003   -> Here InvoiceID is Null

         ,CL0004

];

But, here showing Invoice is Perfect Key and Information density 100% and Subset ration 100%.

Invoice.bmp

I have doubt, perfect key always Unique Id and Not null(Unique + Not Null) values right?

And,

InvoiceDetails:

LOAD *

INLINE [

InvoiceID, InvoiceLine, ProductID, Quantity, Amount

Inv001,1,PR01,10,50

Inv001,2,PR02,10,40

Inv002,1,PR01,30,150

Inv002,2,PR03,10,800

          ,4,PR04,20,700

];

Invoice_details.bmp

Here InvoiceID is 5th row is null, still showing Key and Information density 100% and Subset ration 100%.

i am bit confusion,

1.Perfect Key always Primary key,Unique Id and not null values Properly?

2.Subset ratio 100% means always having values Properly? .

if i understand wrong plz excuse.

Regards,

raja.

francescopuppin
New Contributor III

Re: [Perfect Key]?

Hi,

in this case Qlikview is definitely wrong.

I suppose that you are using an old version of Qlikview (maybe 9?). I have heard that this bug of Qlikview has been fixed later.

I just put the code in QV 11 Personal, and it shows [Key] on both sides, as expected.

Also the density is wrong, and it appears wrong also in my version. This is probably because the spaces in the inline are seen as data.. But if you amend the code as follows, the density will change

Invoices:

LOAD

LEN(TRIM(InvoiceID)) AS LEN,

IF(LEN(TRIM(InvoiceID))=0,Null(),InvoiceID) AS InvoiceID,

ClientID

;

LOAD *

INLINE [

InvoiceID, ClientID

Inv001,CL0001

Inv002,CL0002

          ,CL0003

          ,CL0004

];

InvoiceDetails:

LOAD *

INLINE [

InvoiceID, InvoiceLine, ProductID, Quantity, Amount

Inv001,1,PR01,10,50

Inv001,2,PR02,10,40

Inv002,1,PR01,30,150

Inv002,2,PR03,10,800

          ,4,PR04,20,700

];

Here is how I see it:

It has density 50% (2 NULLs out of 4 records) and subset rati 67% (2 values out of 3, because the space is counted as a value).

If you change the code even more, you will have the most correct picture:

Invoices:

LOAD

LEN(TRIM(InvoiceID)) AS LEN,

IF(LEN(TRIM(InvoiceID))=0,Null(),InvoiceID) AS InvoiceID,

ClientID

;

LOAD *

INLINE [

InvoiceID, ClientID

Inv001,CL0001

Inv002,CL0002

          ,CL0003

          ,CL0004

];

InvoiceDetails:

LOAD

IF(LEN(TRIM(InvoiceID))=0,Null(),InvoiceID) AS InvoiceID,

InvoiceLine, ProductID, Quantity, Amount

;

LOAD *

INLINE [

InvoiceID, InvoiceLine, ProductID, Quantity, Amount

Inv001,1,PR01,10,50

Inv001,2,PR02,10,40

Inv002,1,PR01,30,150

Inv002,2,PR03,10,800

          ,4,PR04,20,700

];

Of course the data set is for test.. but for a correct data set you should never have a missing invoice ID!

I hope it helps

Francesco

Not applicable

Re: [Perfect Key]?

From the Book QlikView 11 for Developers:

"The name of the field. Optionally, if the field is a key field, a qualifier is shown enclosed in square brackets. This qualifier indicates the following levels of key quality:

[Perfect Key] indicates that every row contains a key value, and that all of these key values are unique. At the same time, the field's subset ratio is 100 percent. This qualifier should be seen in dimension tables, where every key should uniquely identify a single record.

[Primary Key] indicates that all key values are unique, but not every row contains a key value or the field's subset ratio is less  than 100 percent

[Key] indicates that the key is not unique. This qualifier is usually seen in fact tables, where the same dimension value may be associated with many different facts."