1557353 - Consider replacing string list metric type with a string set metric type

Status: RESOLVED INCOMPLETE

(Data Platform and Tools :: Glean: SDK, task, P4)

Description

[Georg Fritzsche [:gfritzsche]]

The current use-cases we're aware of for string lists don't require non-unique values.
E.g.: "What is the list of names for the accessibility services available on the device?"

These kind of use-cases could be covered more clearly by a string set metric type, that only submits unique values.
If we can't identify use-cases that require non-uniques, we should consider making this string list type a string set type.

Comment 1

[Georg Fritzsche [:gfritzsche]]

Nvm, i'll keep these now and we can remove them later.

Comment 2

[Jan-Erik Rediger [:janerik]]

We should clarify if that is needed, at best before string list is in use, so we can easily replace it.

Comment 3

[Jan-Erik Rediger [:janerik]]

We're replacing string list with string set.

Comment 4

[Alessio Placitelli [:Dexter]]

Attachment: glean_parsers PR

Comment 5

[Alessio Placitelli [:Dexter]]

Attachment: schema change PR

Comment 6

[Alessio Placitelli [:Dexter]]

Attachment: glean-ac PR

Comment 7

[Alessio Placitelli [:Dexter]]

(In reply to Jan-Erik Rediger [:janerik] from comment #3)

We're replacing string list with string set.

Darn, I noticed while I was down with the implementation that this is already being used by Fenix. We should probably keep both types for the time being, and deprecate the list one. Or, maybe, live with the fact that we have a list.

Comment 8

[Alessio Placitelli [:Dexter]]

After discussing this with Frank, we decided to hold it back for now, until other changes will require baking a new version of the ping schema. See here.

I'm leaving the bug open, but in the backlog, and adding a [schema-v2] tag to remind us that we should consider this in the future.

Comment 9

[Alessio Placitelli [:Dexter]]

Attachment: reverting the glean_parser change PR