Remove Duplicates & Relationsship bug?

[eniX]

Hi guys,

I have two tables t1 and t2. They are related through column r. In t2 the values in column r are distinct, in t1 they are not. I applied the function "Remove Duplicates" (aka Table.Distinct()) on t1.r (not on the whole table), but I still can't establish a 1:1 relationship between these tables: "The cardinality you selected isn't valid for this relationship". I counted the rows and the data shrinked. I also tried to group the table by r and count the group rows - there is no group with more than 1 row...Finally I tried to recreate this scenario with dummy data and it works like it should. 

Is it possible that Power BI doesn't update the meta data and thus thinks the values in t1 are still not distinct?

My data is coming from a SQL Server (import mode).

[WolfBiber]

Do you have NULL and Empty Values in this column? 

[eniX]

No sir.

In the meantime I loaded the data into jupyter notebook and preprocessed it there. I loaded transformed data into Power BI and tried to establish a relationship - it's still n:1...So the problem seems not to arise from using the "Remove Duplicate" function. Still have no clue 😕

[WolfBiber]

Hi,

can you share some example Data as PBIX?

[eniX]

Sure, if you tell me how do I do this?

[WolfBiber]

just upload the file to some cloud storage like onedrive and share the link

[WolfBiber]

ok, as you can see you have more data rows in one table than in the other 

and you have ArtDLNR in one table which isnt present in the other, and vice versa. 

and you have many rows with Empty  ArtDLNr (not a unique constraint),

 

thats not a 1:1.

You have to cleanse your data or keap it as 1:*

Ich wünsche Ihnen viel Erfolg

Viele Grüsse aus HH

[eniX]

Well, first of all, thank you for taking time to help me.

I'm sorry, but I can't really understand your answer.

1) One-to-one relation means, there is at most one entity in the tables (defined by a id/join key). It doesn't have to be exactly one. Consider my example:

Table1 und Table2 do not have exactly same join keys. We have values in Table1, which we do not have in Table2 and vice versa. The relationship is still 1:1 because we don't have more than one same value in those tables.

Then cosndier Table3, it has duplicate join key. That's why it's one-to-many, not one-to-one.

So the only "cleanse" activity I have to do is to dump duplicates, which I did in my data.

2) What you describe in your answer - to guaranteed have a value in both tables - is called "referential integrity". This is definitely not required for one-to-one-relationship.

3) Could you tell me where in my data are null values? I'm pretty sure I have none.

[WolfBiber]

Good morning,

it seems that you are more expiriented than others. I just tried to explain some problems that can occur with 1:1 relationsships.

1.) and 2.) thats true, and if this is your desired result everything is fine. 

3.) the main mistakes made for a 1:1 is that the undestanding that empy values "" and Null values are treated same. 

4.) did you cleanse duplicates and then combined 2 Columns? Maybe thats the reason you have duplicates in your Key column in HREF

But I can't use the Query Editor cause I don't have the source files. So I have only Dax Methods to check your data

[WolfBiber]

Hey,

you have duplicates e.g.: "17 X 1075_64" in HREF_preprocessed

[eniX]

---

@WolfBiber wrote:

Hey,

you have duplicates e.g.: "17 X 1075_64" in HREF_preprocessed

---

^^

---

@WolfBiber wrote:

Good morning,

it seems that you are more expiriented than others. I just tried to explain some problems that can occur with 1:1 relationsships.

1.) and 2.) thats true, and if this is your desired result everything is fine. 

3.) the main mistakes made for a 1:1 is that the undestanding that empy values "" and Null values are treated same. 

4.) did you cleanse duplicates and then combined 2 Columns? Maybe thats the reason you have duplicates in your Key column in HREF

But I can't use the Query Editor cause I don't have the source files. So I have only Dax Methods to check your data

---

But neither I have null values in my data, nor duplicates. That's why the topic name contains "bug?"...

4) --> I created new calculated column, concatenating two strings. I removed duplicates after the creation of the column. I also tried both to transformate my data directly in Power BI and outside of it, with Jupyter Notebooks. Same result.

Would it help you if I gave you the data as well?

[eniX]

BTW I can create your "TT" calculated table without an error.

[WolfBiber]

In the example data you gave me you have duplicates:

[eniX · Accepted solution]

Thank you for your patience WolfBiber,

With help of your example I finally found the problem...

• Power Query --> case sensitive
• Power Pivot / Power View --> case insensitive

That's why I can't find "duplicates" in Power Query, but on the canvas you see duplicate values... (it even transformed small x to capital letter). just WOW Microsoft 😞

For our example:

• HArtNr "03005.00101" is assigned to ArtDLNr "17 X 1075_64"
• HArtNr "03005.00203" is assigned to ArtDLNr "17 x 1075_64"

Possible solutions/work arounds and some stuff to read:

http://www.thebiccountant.com/2015/08/17/create-a-dimension-table-with-power-query-avoid-the-bug/

http://www.thebiccountant.com/2016/10/27/tame-case-sensitivity-power-query-powerbi/

https://community.powerbi.com/t5/Desktop/Identifying-Duplicates-by-not-taking-Case-Sensitivity-into/...

https://community.powerbi.com/t5/Desktop/quot-Remove-Duplicate-quot-doesn-t-remove-all-duplicate/td-...

Still didn't figure out how I can fix my problem. I could remove duplicates in case-insensitive manner with

= Table.Distinct(aTable, { "aColumn", Comparer.OrdinalIgnoreCase } )

but I don't want to lose this distinction. ATM I'm wondering how I can get around this. But if necessary, I will open another thread for this.

[6yearoldbug]

Absolutely incredible that a 6 year old bug with thousands of view and an incredible amount of threads created about it still has not been fixed.