Ultimate Long TM1 quiz

[Adam]

I'm noticing that participants became less active. 2 months left, pull yourselves together guys!

This is fun for me and we should keep going even beyond the two month mark. I we all want to play nice in the sandbox (who am I kidding Version dimension) so we’re not all immediately jumping in.

Question #41:
The cube and rule are shown on screenshots below. B1 is not fed. What will be in C1? Will it take numbers from not-fed calculated cells of B1?

I remember to this day when I was learning TM1, I made a post-it note of how rules work as I was figuring it all out. One of the bullets was DB statement will always calculate correctly unless you’re referencing a consolidated cell that has unfed leaves. (Even then it’s correct, not summing its unfed leaves. It’s only “not correct” if the unfed leaf was unintentionally unfed.)

Therefore, you would see 1, 2, 3 in row C1.

[Elessar]

Correct answer:
0, 1, 2, 3

Winner of this round is Adam!

Question #42:
You have TM1 server with sensitive data. You have a well-done security model in TM1: access to sensitive data is granted only to right users. But windows administrator can take DATA folder and launch it on his own TM1 server with his }Clientgrups cube file where he is admin. And thus get access to sensitive data. Can we do anything with this?

[Adam]

Question #42:
You have TM1 server with sensitive data. You have a well-done security model in TM1: access to sensitive data is granted only to right users. But windows administrator can take DATA folder and launch it on his own TM1 server with his }Clientgrups cube file where he is admin. And thus get access to sensitive data. Can we do anything with this?

One approach could be to encrypt the contents of the data folder using the out-of-the-box encryption. Then, ensure the server admin doesn't have access to the keystore.

[Elessar]

Which out-of-the-box encryption do you mean?

[Elessar]

Correct answer:
Yes, you can encrypt DATA folder using TM1Crypt utility: https://www.ibm.com/docs/en/planning-an ... encryption

Winner of this round is Adam!

Question #43:
Optimize this rule: ['Y'] = IF(['X'] < 0, 0, ['X']);

[ascheevel]

Code: Select all

['Y'] = MAX(0, ['X']);

[MarenC]

['Y'] = MAX(0, ['X']);

It is clever but is that really optimised?

I will say this rule cannot be optimised but will throw this into the mix:

['Y'] = IF(['X'] < 0, STET, ['X']);

Maren

[ascheevel]

It is clever but is that really optimised?

I'll defer to Elessar's judgement on whether or not my suggestion was the optimization they were looking for, but I do believe Max/Min functions can be quite a bit faster than IF statement processing. I'm sure I've read a comment before by lotsa saying exactly this, but my memory isn't as good as it once was. This could of course be tested using }StatsByRule cube, but I haven't the time or energy this week to work up and execute a test scenario. It's also my opinion that my suggestion is easier to read than the original IF statement, but we always think our own solutions are the best and easiest to read. If it isn't just my opinion in this case, then I've optimized for both performance and readability.

[Emixam]

Code: Select all

['Dim Name':'Y'] = MAX(0, ['Dim Name':'X']);

[ascheevel]

Nice!

[Elessar]

Agree, nice catch!

I doubt somebody can make anything better, so here is a side exercise while we are waiting for Friday: can we rewrite such rule without IF and MAX/MIN?

[Adam]

What about adding Hier Name and N: with C: only if relevant?

Code: Select all

['Dim Name':'Hier Name':'Y'] = N: MAX(0, ['Dim Name':'Hier Name':'X']);

can we rewrite such rule without IF and MAX/MIN?

For sake of readability I’m going to leave out the DimName/HierName/N:/C:

Code: Select all

['Y'] = (['X'] + ABS(['X'])) / 2;

I wound NOT use this in production because MAX approach is much more legible.

[MarenC]

How optimised are these alternative solutions, what does the empirical evidence show?

[Emixam]

How optimised are these alternative solutions, what does the empirical evidence show?

As per Cubewise, including the dimension reference is faster because this constrains the element search to a single dimension instead of searching over all dimensions.

https://code.cubewise.com/blog/7-tips-t ... ics-rules/

[gtonkin]

I would go one step further and fully qualify everything i.e. dim:hier:member

[Elessar]

Correct answer:
It's hard to decide!
Emixam's answer was a really good catch. This dim:element optimization was measured and described by Cubewise. I have not measured if dim:hier:element is faster than dim:element. So, Adam, if you can prove this, you will also get a point also.

For the "MAX vs IF": I've realized recently that I've measured this long ago in 9.5.2, and MAX was much faster than IF expression in this case. I cannot guarantee this in PA, so it would be nice if somebody measures this. Anyway, you should avoid IFs in TM1 in any case.

Winner of this round is Emixam!

Question #44:
What are the differences between this methods of data clear?

• ViewZeroOut (view covering the whole cube)
• CubeClearData
• CubeDestroy+CubeCreate

[MarenC]

Anyway, you should avoid IFs in TM1 in any case.

This is a joke right?

[Adam]

This is a joke right?

I didn’t take it as a joke. When it’s possible to construct your TM1 DB so it doesn’t use/uses less IF statements, it will be faster. Sometimes it’s not possible and you deal with the consequences. Often times it’s trivial, but it might add up. Not something that I typically worry about myself.

[MarenC]

I would like to bet there isn't a model out there that doesn't use an if somewhere, be it a TI or a rule.

[Elessar]

Of course, you cannot avoid all IFs. But you should minimize them.

As an example:

Code: Select all

['A'] = N: IF(!Country @= 'Klingon Empire', ['B'], ['C']);

vs

Code: Select all

['A', 'Country':'Klingon Empire'] = N: ['B'];
['A'] = N: ['C'];