PAW: Enable random selection of members to be passed to a TI

[JohnO]

As per the title, our users want to be able to make a random selection of elements from a specified dimension/ hierarchy (With a very large number of elements) and for that selection to be passed directly to a TI. I'm thinking I can take a delimited string into a TI without an issue but I cannot think of a way to create a delimited string based on random selections chosen within the set editor. TM1RPTROW using ElementList seems the the obvious place to start with a concatenation of the resulting element list but I cannot think of a way for random user selections to me made and passed.

I thought of another method using a simple cube with a Yes/No flag (Selection of the elements simply made in the set editor and returned to the cube viewer) but the problem is that the function could be used by multiple people simultaneously and we would want to restrict the running of the process to the list of elements selected by the user.

If there was an elegant way of creating a temporary subset then that would work but I cannot think of an elegant approach.

Any thoughts?

[Wim Gielis]

Hello John,

What comes to mind is a separate cube with the }Clients dimension inside, and an MDX to get the name of the current user.

Then the users can either paste their desired elements in that cube and the TI retrieves them.
Either you can use picklists to solicit the selections from the users, again TI will retrieve the names.

Not very elegant but not impossible either.

[Mark RMBC]

Hi John,

You were not crystal clear but I am assuming you mean potentially multiple element selections from dimensions. For example, user might select product A, Year 2021 etc or they might select Product A, B, C and D and Year 2021 etc.

Assuming this is your issue, I would normally tackle something like this by either doing something along the lines that Wim suggested, or as an alternative to that, having a measure which is effectively treated as a boolean, i.e. enter 1 or 0.

If that cell is equal to 1 then that means the intersection is included in the TI process, if it is 0 then it is ignored.

You build a view in the TI based on the 'boolean' measure being 1, i.e. exclude zeros from the view.

To end user, they would enter a 1 against the measure into any rows they wished to process, and then run the process.
Any dimensions not on the rows would be passed directly into the TI.

The end user view being something logical!

All the above depends on knowing the specifics of course.

regards,

Mark

[gtonkin]

Possibly going with your Yes/No selection but using a sandbox to isolate each user’s selection?

Perhaps if it was a standard name you could hardcode the sandbox name into the TI to process the list.

[JohnO]

Thanks all, I think I have confirmation that there is no obvious answer to the challenge.

[burnstripe]

You could do this with MDX.
Get the person to create a private subset with the members they select.
Create a process which has a parameter to pick up the name of the private subset and then have something along these lines...

SubsetCreateByMDX('Public Subset', 'TM1SubsetToSet( [DimA].[DimA], "Private Subset", "private")');
SubsetMDXSet('Dim A', 'Public Subset', '');

Within the prolog you could create a temporary subset which grabs the members from the private subset using MDX. As long as the user owning the subset triggers the process themselves, a public subset would be created with the members which you could then loop through to do whatever you needed in the TI.

[MarenC]

Hi,

I think there is an obvious answer to this challenge, go with the flags in a cube which includes the }client, and improve communication within the organisation!

If a user selects Element A and B, I presume you would not expect some other user to select Element A or B at the same time. And if that did happen maybe get them to have a word with each other!

Maren