Hi, We are thinking on having some User Data Filters for our managers, but started to consider performance impact when using filter expression

IN

with a huge collection of User IDs, the performance would usually degrade in this scenario but we want to better understand it, in case there are any internal mechanism to improve this already in place, as the data filter only allows a MAQL string. Also, can we consider that this IN clause is directly passed to the SQL statement?

Hi Jhonatan, I can definitely confirm that the UDF condition is already involved in the SQL querying your data source. You can easily verify that by navigating to “Analyze” tab, create some report and in the URL, change “edit” for “debug”. It will download a file containing the SQL as well and you can compare non-UDF vs UDF report.

Hi Julius, do you know if a MAQL can directly access GD resources like user_group I need to do something like {label/user_group.id} IN ("group1", "group2") so that one user would only be able to see data for all the users in a certain user_group

Not in the MAQL. UDFs doesn’t work like that. The user or user_group is specified when you are setting the UDF in the workspace. Example payload:

{
  "data": {
    "type": "userDataFilter",
    "id": "<data-filter-id>",
    "attributes": {
      "maql": "<maql-expression>",
      "title": "<data-filter-name>"
    },
    "relationships": {
      "user": {
        "data": {
          "id": "<user-id>",
          "type": "user"
        }
      }
    }
  }
}

You can either use “type”: “user” or “type”: “userGroup”

problem is even if I assign an UDF to a user_group, I need to filter the data by only the users in that group manager is in GROUP 1, he should only see the data for users in GROUP 1, so somehow I need pass the user_ids to the UDF

maybe I misunderstood, so users are actual columns in your data?

We have a User dataset/table with all Users

Apologies, I obviously didn’t exactly understand your question before. So to clarify further —

WHERE

clause isn’t supported in UDF MAQL in this context because UDFs only allow flat, row-level filters. You can use

AND

,

OR

, or

IN (...)

with static values like:

"maql": "{label/user.id} IN (\"user1\", \"user2\", \"user3\")"

But expressions like:

{label/user.id} IN ({label/user_manager.user_id} WHERE {label/user_manager.manager_id} = "mgr123")

won’t work, since UDFs can’t evaluate filtered sets or subqueries. To achieve that logic, you’d need to precompute the list of users for each manager and apply it statically in the UDF.

yeah we are aware of the first approach, the main concern is performance, have you guys not had any customer complaining about performance when having like 1k users in a filter like that?

The limit indeed looks to be 1,2k it doesn’t allow me to create more. I am not aware of any customer who would reach out with this problem and I’m afraid that UDFs were to designed for similar purpose (of such high attribute element volumes).

I understand what you mean, but it's not that we want to see the manager user data, but we want the manager to only see data from all users they manage which is why it would have to be something like: {label/user.id} IN ({label/user_manager.user_id} WHERE {label/user_manager.manager_id} = "mgr123") as

label/user_manager.user_id

would contain all the users associated with their managers

Hey Jhonatan, sry I was not online yet. Yes indeed, combining multiple UDFs or even UDFs and WDFs is always an intersection so the condition is actually AND. This is nicely explained in our documentation:

Hi @Julius Kos thank you - please do submit a product feedback on our behalf 🙏

🎉 New note created.