Ex4pm.Core.OLAP (ex4pm v26.9.9)

Copy Markdown View Source

OLAP-style multi-dimensional operations over the canonical Ex4pm.EventLog IR (Ex4pm.OCEL.normalize/1 output). Operates on real %Ex4pm.EventLog{} structs (events with :activity, :timestamp, :attributes, :object_ids) — no separate OCEL2 struct exists in this codebase, so this module builds directly on the existing canonical IR from apps/ex4pm_core/lib/ex4pm/ocel.ex.

A "dimension" is one of:

  • :activity — the event's activity name
  • :day — the YYYY-MM-DD prefix of the event's ISO-8601 timestamp
  • :month — the YYYY-MM prefix of the event's ISO-8601 timestamp
  • {:attribute, key} — an arbitrary event attribute, looked up by string or atom key in event.attributes
  • {:object_type, type} — not a value-extraction dimension; used only via slice/3/dice/2 to filter events that reference at least one object of type (requires log.objects)

:day and :month form a two-level time hierarchy: roll_up/3 from :day to :month coarsens; drill_down/3 from :month to :day re-expands.

Summary

Functions

dice(log, filters)

Filters log to only the events matching every {dimension, value} pair in filters simultaneously (logical AND across dimensions).

drill_down(log, dimension, finer_dimension)

Expands an aggregated dimension back to a finer_dimension. Groups log's events first by dimension (the coarse grouping), then by finer_dimension within each coarse group, returning a nested map

roll_up(log, dimension, aggregate_fn)

Aggregates log's events up a dimension hierarchy. Groups events by their dimension value and applies aggregate_fn (a function of the list of events in that group) to each group.

slice(log, dimension, value)

Filters log to only the events whose dimension value equals value. Objects and object_relationships are narrowed to what the surviving events still reference, so the result is itself a coherent %EventLog{}.

Functions

dice(log, filters)

@spec dice(Ex4pm.EventLog.t(), [{atom() | {atom(), term()}, term()}]) ::
  Ex4pm.EventLog.t()

Filters log to only the events matching every {dimension, value} pair in filters simultaneously (logical AND across dimensions).

drill_down(log, dimension, finer_dimension)

@spec drill_down(
  Ex4pm.EventLog.t(),
  atom() | {atom(), term()},
  atom() | {atom(), term()}
) :: %{
  optional(term()) => %{optional(term()) => [Ex4pm.Event.t()]}
}

Expands an aggregated dimension back to a finer_dimension. Groups log's events first by dimension (the coarse grouping), then by finer_dimension within each coarse group, returning a nested map:

%{coarse_value => %{finer_value => [events]}}

This is the structural inverse of roll_up/3: rolling log up to dimension with &length/1 and summing the inner group sizes of drill_down/3's result for the same coarse key recovers the same count.

Examples

iex> Ex4pm.Core.OLAP.drill_down(log, :month, :day)
%{"2026-08" => %{"2026-08-30" => [...], "2026-08-31" => [...]}}

roll_up(log, dimension, aggregate_fn)

@spec roll_up(Ex4pm.EventLog.t(), atom() | {atom(), term()}, ([Ex4pm.Event.t()] ->
                                                          term())) :: %{
  optional(term()) => term()
}

Aggregates log's events up a dimension hierarchy. Groups events by their dimension value and applies aggregate_fn (a function of the list of events in that group) to each group.

Returns a map of dimension_value => aggregate_fn.(events_in_group), sorted by key for deterministic output.

Examples

iex> Ex4pm.Core.OLAP.roll_up(log, :day, &length/1)
%{"2026-08-30" => 2, "2026-08-31" => 3}

iex> Ex4pm.Core.OLAP.roll_up(log, :month, &length/1)
%{"2026-08" => 5}

slice(log, dimension, value)

@spec slice(Ex4pm.EventLog.t(), atom() | {atom(), term()}, term()) ::
  Ex4pm.EventLog.t()

Filters log to only the events whose dimension value equals value. Objects and object_relationships are narrowed to what the surviving events still reference, so the result is itself a coherent %EventLog{}.

Examples

iex> Ex4pm.Core.OLAP.slice(log, :activity, "ship")
%Ex4pm.EventLog{events: [...]}