DSXchange

Can anyone please offer suggestions for implementing the following as an IA data rule, ideally without creating view in source databases?

The requirement is that the total of a customer's account balances (which are possibly stored in the same or different tables) must not exceed a threshold read from a reference table.

Conceptually the need is to sum the account balances for each particular customer in the five possible tables in which the account(s) may occur, then sum those five results, then compare against the threshold value.

I don't believe that summing is possible within IA data rule definitions, even though there exist other "whole of table" expression elements.

"They" have indicated a strong preference not to have to create views, nor run aggregating queries, against the source data. The whole thing (including using IA) should be simple enough (which I interpret to mean that they want a drag-and-drop interface) for non-technical users to accomplish.

Thank you for your time.

Oh, they have another requirement, namely that the output from one data rule (for example only the rows that pass) can become the input to the next data rule. I shall be pushing for the Data Rules stage in this case, but would appreciate any alternative approaches that you may be able to offer.

I am going to take a guess here, in that you could possibly provide data rules to meet their needs, if you would first develop new preprocessing DataStage jobs that gently extract the data, and perform the aggregations using the stage (vs. using the database queries), and any other complex logic, then output the results somewhere, which would not necessarily have to be back into the same database. Then the less technical users can run "simple" pass/fail data rules against your aggregated, processed, summarized results.

"No tech skills: I want a business user to be able to do it"
Sounds like they want self-service reporting.

This isn't a great fit for InfoSphere (a suite of dev tools, regardless of things like IA rules "wizards"), unless you can sell them on the idea of using it to create some sort of logical view they can use for self-service.

Something like a Tableau or QlikView might be s better option, as the tool resolves the complexity (it's designed the answer this sort of question).
Not what a team of DS devs wants to hear though...


As for the second part (providing this hammer is the only tool you've got), you can create a rule that outputs to a named table and set IADB as an extra source, then use that named table as the source for subsequent rules.
A better option than a Data Rules stage. They still need a DS dev after all.
Still not terribly "business friendly".

In the end, technical complexity needs to be reduced somewhere between data and keyboard.
Where it gets done and how much needs to get done will determine the solution.

Or just have the user ask Watson for the answer.

Thank you for your ideas. We will give them consideration.