TM1 Forum

Posted: **Thu Jan 24, 2019 7:39 am**

Hi,

Bit of a long shot this, does anyone know of a clever way to analyse a whole cube for Underfeeding?

I have used the overfeeding trick involving the cloned cube and that works really well, but I understand Underfeeding will be much harder, as the cells in question wouldn't be readable for comparison. I realise I can just reconcile to base data or use "Check Feeders" but this is painstaking and I was looking for a more rounded approach.

I thought of the following :-
Clone the cube with one extra Dim with two elements (maybe "skip" and "non skip") and no rules, write a quick TI to export all N-level rule values to the cloned cube on the "skip" layer. Then turn off skipcheck, do the same and put it on "non skip", compare the two layers. I realise this may take FOREVER on a large cube though.

Now I write this, I am considering maybe its not worth all the effort vs the risk of something materially being missed!

Posted: **Thu Jan 24, 2019 9:36 am**

You could try this - quite similar to yours but doesn't involve changing the rules.

Create a clone cube with an extra measure dimension
Difference = Values - Fed Values
Write a TI to export N level values from your real cube to the clone cube
Run it with zero suppress on, writing to Fed Values
Run it with zero suppress off, writing to Values
Review the differences

You can almost certainly test on a subset of the content eg one year and version.

Posted: **Thu Jan 24, 2019 10:34 am**

Thanks - that's simpler. I will give it a go when I get chance and report back on the outcome!

Posted: **Sat Feb 02, 2019 12:15 pm**

LiamLawrence wrote: ↑Thu Jan 24, 2019 7:39 amBit of a long shot this, does anyone know of a clever way to analyse a whole cube for Underfeeding?

Yes. Run this TI command where you enter the cubename and inspect the generated text file:

Code: Select all

DebugUtility( 113, 0, 0, 'cubename', 'underfed cells.txt', '' );

Wim

Posted: **Sat Feb 02, 2019 8:38 pm**

Hi Wim

Do you have, or could you point me in the direction of, the full list of codes for DebugUtility() function ?

I've searched for these for a while and can't locate them.

Cheers
Ajay

Posted: **Sat Feb 02, 2019 9:13 pm**

Ajay wrote: ↑Sat Feb 02, 2019 8:38 pm Hi Wim

Do you have, or could you point me in the direction of, the full list of codes for DebugUtility() function ?

I've searched for these for a while and can't locate them.

Cheers
Ajay

Hi,

No I don't have such a list. Basically it's trial and error from my part.

For example, a loop from 1 to 200 or even 1000 is what I did, for various combinations of input values for the other arguments.

I saw that number 108 can export the data of a cube to a text file that you provide:

Code: Select all

DebugUtility( 108, 0, 0, 'cubename', 'test.txt', '' );

Number 100 can export 'calculated feeders to a text file and show the count of fed cells', loosely speaking (no big investigation from my side).
There is another one that can delete all application cubes and (respective) dimensions, except the chosen cube (and hence, its dimensions).

Number 116 seems to be not used and throws an error when used. Other numbers do not throw an error, as per my testing (numbers 1 to 1000 were tested).

It would be good if IBM staff could provide a list of at least the ones that are useful to TM1 developers / admin persons.
I have the impression that only numbers 100-135 have been used, but I can be wrong on that one.
If anyone can help discovering the numbers, please join and post your findings. It could be similar to the list of constants that are used in *.pro files.

Posted: **Mon Feb 04, 2019 10:19 am**

Thanks Wim

I've raised a ticket with IBM this morning for a full list.

"If" I get one back from them, I will post it on the forum.

Cheers
Ajay

Posted: **Mon Feb 04, 2019 10:26 am**

Hi,

I think DebugUtility is for IBM's internal use, and I doubt it's documented well. So they unlikely will share it with the TM1 Society.

There is a topic for collecting known codes: https://www.tm1forum.com/viewtopic.php?f=3&t=10475
@Wim, kindly ask you to post you investigation there

Posted: **Mon Feb 04, 2019 11:59 am**

Elessar wrote: ↑Mon Feb 04, 2019 10:26 am@Wim, kindly ask you to post you investigation there

Hello,

Sure, here's my test code for investigations:

Code: Select all

i = 0;
While( i <= 1000 );

   If( i <> 116 );

      AsciiOutput( 'test.txt', Expand( 'Start of %i%' ));
      DebugUtility( i, 0, 0, 'TM1_' | NumberToString( i ) | '.txt', '', '' );
      AsciiOutput( 'test.txt', Expand( 'End of %i%' ));

   EndIf;

   i = i + 1;

End;

and several variations, like using cubename only, or cubename and text file, and so on.

Posted: **Tue Feb 05, 2019 5:33 am**

Thanks Wim,

I meant "Please post your investigation's results there: https://www.tm1forum.com/viewtopic.php?f=3&t=10475", so that all the known DebugUtility commands would be in one topic

Sorry for misleading you

Posted: **Tue Feb 05, 2019 10:17 am**

Hello,

I would like to refrain from doing that, since it's all trial and error and guessing from my part. The chances of being wrong are rather high.
Let's wait for IBM to come up with a list, instead of this speculation.
Certainly when I noticed that there are commands that DELETE cubes and dimensions, it's dangerous and I don't want to ruin TM1 models because of some late-night speculation and guesswork. By the way, the SetMTQ command that was proposed in the other topic, does not work for me.

Posted: **Sat Apr 02, 2022 10:45 am**

Interesting to know DebugUtility and tried to run the following code

Code: Select all

DebugUtility( 114, 0, 0, 'myCube' , 'targetFolder' , '' );

Looks like it only produce a feeder file for the cube in PAL 2.0.9.4 after the first run.
The second run works with obfuscated data.

Posted: **Sat Apr 02, 2022 11:06 am**

For me it seems to work every time, generating a *.cub file.
Version 2.0.9.3

Maybe you are mixing modes 113 and 114 for DebugUtility ?

Posted: **Sat Apr 02, 2022 3:54 pm**

Think Wim has called it - 114 is to output to an obfuscated .cub file, 113 should give you information on underfed cells.

Try something like:

Code: Select all

DebugUtility( 113, 0, 0, pCube, 'pPath\cubename_underfed_cells.txt' , '[element][, element][,…]' );

e.g. DebugUtility(pFunction, 0, 0, pCube, sFilename, 'Budget,Luxemburg,T Series 4.0 L Sedan');

You could also play with:

Code: Select all

DebugUtility( 100, 0, 0, pCube, 'pPath\cubename_feeders.txt', '' );

# List all calculated feeders in a text file, including the total number, for the given cube

Code: Select all

DebugUtility( 105, 0, 1, pCube, 'pPath\cubename_overfed_cells.txt' , '' );

# Create a file with the list of overfed Cells (Cells Fed but with 0 value)

Also, try DebugUtilityEx:

Code: Select all

DebugUtilityEx('FeedersFromCell','SalesCube','Budget','United States','L Series 1.6 L Convertible','Units','Jan','FeedersFromCell.txt');

Gives a lot of information on feeders and dependencies.

As with all undocumented functions, use at your own peril.

TM1 Forum

Checking for Underfed cells

Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells

Re: Checking for Underfed cells