feat(user-account-merge): add account merge tracking and edX/MIT Learn user mapping

src/ol_dbt/models/dimensional/_bridge_tables.yml

@@ -175,3 +175,41 @@ models:
   - name: userorganization_is_manager
     description: If true, the user is a manager of this organization and can access
       organization dashboard features.
+
+- name: bridge_user_account_link
+  description: >
+    Edges between platform-specific user accounts, independent of dim_user's
+    own identity-collapsing. from_platform = to_platform is a
+    same-platform redirect from an old account to its canonical replacement
+    (derived from mitxonline audit history) -- a redirect to follow. Different
+    platforms is a cross-platform co-reference (e.g. a MITx Online account's
+    linked edX.org account) -- not a redirect. Grain: one row per
+    (from_platform, from_user_id, to_platform).
+  tests:
+  - dbt_utils.unique_combination_of_columns:
+      combination_of_columns:
+      - from_platform
+      - from_user_id
+      - to_platform
+  columns:
+  - name: from_platform
+    description: Platform identifier of the source account ('mitxonline' today).
+    tests:
+    - not_null
+  - name: from_user_id
+    description: Raw platform-specific user ID of the source account.
+    tests:
+    - not_null
+  - name: to_platform
+    description: Platform identifier of the target account. Same as from_platform
+      for a merge redirect; a different platform (e.g. 'edxorg') for a cross-platform
+      link.
+    tests:
+    - not_null
+  - name: to_user_id
+    description: Raw platform-specific user ID of the target account.
+    tests:
+    - not_null
+  - name: observed_on
+    description: Timestamp of the merge (most recent, if reversed). Null for cross-platform
+      link rows.

src/ol_dbt/models/dimensional/bridge_user_account_link.sql

@@ -0,0 +1,87 @@
+{{ config(
+    materialized='table'
+) }}
+
+-- Edges between platform accounts, independent of dim_user's own identity-
+-- collapsing (grouping by hashed email). from_platform = to_platform is a
+-- merge redirect (old account -> canonical, confirmed via mitxonline audit
+-- history); different platforms is a co-reference, not a redirect.
+-- Grain: (from_platform, from_user_id, to_platform).
+with courserunenrollment_audit as (
+    select * from {{ ref('stg__mitxonline__app__postgres__courses_courserunenrollmentaudit') }}
+)
+
+, programenrollment_audit as (
+    select * from {{ ref('stg__mitxonline__app__postgres__courses_programenrollmentaudit') }}
+)
+
+, user_changes as (
+    select
+        json_extract_scalar(enrollmentaudit_data_before, '$.user') as platform_user_id_before
+        , json_extract_scalar(enrollmentaudit_data_after, '$.user') as platform_user_id_after
+        , enrollmentaudit_created_on
+    from courserunenrollment_audit
+    where json_extract_scalar(enrollmentaudit_data_before, '$.user')
+        != json_extract_scalar(enrollmentaudit_data_after, '$.user')
+
+    union all
+
+    select
+        json_extract_scalar(enrollmentaudit_data_before, '$.user') as platform_user_id_before
+        , json_extract_scalar(enrollmentaudit_data_after, '$.user') as platform_user_id_after
+        , enrollmentaudit_created_on
+    from programenrollment_audit
+    where json_extract_scalar(enrollmentaudit_data_before, '$.user')
+        != json_extract_scalar(enrollmentaudit_data_after, '$.user')
+)
+
+, grouped as (
+    select
+        cast(platform_user_id_before as integer) as platform_user_id_before
+        , cast(platform_user_id_after as integer) as platform_user_id_after
+        , max(enrollmentaudit_created_on) as last_merged_on
+    from user_changes
+    group by platform_user_id_before, platform_user_id_after
+)
+
+-- Normalize to an unordered pair so a reversed merge (both directions
+-- present) competes against itself, keeping only the most recent direction.
+-- Note: QUALIFY is not supported by Trino; using ROW_NUMBER subquery.
+, ranked_merges as (
+    select
+        *
+        , row_number() over (
+            partition by
+                least(platform_user_id_before, platform_user_id_after)
+                , greatest(platform_user_id_before, platform_user_id_after)
+            order by last_merged_on desc
+        ) as row_num
+    from grouped
+)
+
+, account_merges as (
+    select
+        'mitxonline' as from_platform
+        , platform_user_id_before as from_user_id
+        , 'mitxonline' as to_platform
+        , platform_user_id_after as to_user_id
+        , last_merged_on as observed_on
+    from ranked_merges
+    where row_num = 1
+)
+
+, platform_links as (
+    select
+        'mitxonline' as from_platform
+        , user_mitxonline_id as from_user_id
+        , 'edxorg' as to_platform
+        , user_edxorg_id as to_user_id
+        , cast(null as varchar) as observed_on
+    from {{ ref('int__mitx__users') }}
+    where user_mitxonline_id is not null
+        and user_edxorg_id is not null
+)
+
+select * from account_merges
+union all
+select * from platform_links

src/ol_dbt/models/dimensional/bridge_user_account_merge.sql

@@ -0,0 +1,68 @@
+{{ config(
+    materialized='table'
+) }}
+
+-- Verified account-merge redirects derived from mitxonline audit history
+-- (an admin reassigning a record's owner confirms two accounts are the same
+-- person). Only mitxonline today; `platform` allows adding others later.
+-- Grain: one row per (platform, platform_user_id_before). A merge can be
+-- reversed later, so only the most recent direction per account pair is kept.
+with courserunenrollment_audit as (
+    select * from {{ ref('stg__mitxonline__app__postgres__courses_courserunenrollmentaudit') }}
+)
+
+, programenrollment_audit as (
+    select * from {{ ref('stg__mitxonline__app__postgres__courses_programenrollmentaudit') }}
+)
+
+, user_changes as (
+    select
+        json_extract_scalar(enrollmentaudit_data_before, '$.user') as platform_user_id_before
+        , json_extract_scalar(enrollmentaudit_data_after, '$.user') as platform_user_id_after
+        , enrollmentaudit_created_on
+    from courserunenrollment_audit
+    where json_extract_scalar(enrollmentaudit_data_before, '$.user')
+        != json_extract_scalar(enrollmentaudit_data_after, '$.user')
+
+    union all
+
+    select
+        json_extract_scalar(enrollmentaudit_data_before, '$.user') as platform_user_id_before
+        , json_extract_scalar(enrollmentaudit_data_after, '$.user') as platform_user_id_after
+        , enrollmentaudit_created_on
+    from programenrollment_audit
+    where json_extract_scalar(enrollmentaudit_data_before, '$.user')
+        != json_extract_scalar(enrollmentaudit_data_after, '$.user')
+)
+
+, grouped as (
+    select
+        cast(platform_user_id_before as integer) as platform_user_id_before
+        , cast(platform_user_id_after as integer) as platform_user_id_after
+        , max(enrollmentaudit_created_on) as last_merged_on
+    from user_changes
+    group by platform_user_id_before, platform_user_id_after
+)
+
+-- Normalize to an unordered pair so a reversed merge (both directions
+-- present) competes against itself, keeping only the most recent direction.
+-- Note: QUALIFY is not supported by Trino; using ROW_NUMBER subquery.
+, ranked as (
+    select
+        *
+        , row_number() over (
+            partition by
+                least(platform_user_id_before, platform_user_id_after)
+                , greatest(platform_user_id_before, platform_user_id_after)
+            order by last_merged_on desc
+        ) as row_num
+    from grouped
+)
+
+select
+    'mitxonline' as platform
+    , platform_user_id_before
+    , platform_user_id_after
+    , last_merged_on as merged_on
+from ranked
+where row_num = 1

src/ol_dbt/models/reporting/_reporting__models.yml

@@ -1381,3 +1381,37 @@ models:
   - name: user_latest_row_desc
     description: integer, row number indicating the latest row for each user based
       on application and course run dates
+
+- name: edxorg_mitlearn_user_mapping
+  description: >
+    Cross-platform identity crosswalk between edX.org and MIT Learn accounts,
+    connected via each account's link to MITx Online. Only includes users
+    linked on at least one side. Grain: one row per MITx Online account
+    (merged-away accounts are redirected to their canonical replacement via
+    bridge_user_account_link).
+  tests:
+  - dbt_utils.expression_is_true:
+      expression: "edxorg_user_id is not null or mitlearn_application_user_id is not\
+        \ null"
+  columns:
+  - name: mitxonline_openedx_user_id
+    description: MITx Online Open edX user ID, nullable. The connecting hop between
+      edX.org and MIT Learn -- not itself part of the crosswalk.
+    tests:
+    - unique:
+        config:
+          severity: warn
+  - name: edxorg_user_id
+    description: edX.org user ID, nullable. Not populated if dim_user shows a link
+      that bridge_user_account_link can't independently confirm for that MITx Online
+      account (see model for the identity-collapsing collision this guards against).
+    tests:
+    - unique:
+        config:
+          severity: warn
+  - name: mitlearn_application_user_id
+    description: MIT Learn user ID, nullable.
+    tests:
+    - unique:
+        config:
+          severity: warn

src/ol_dbt/models/reporting/edxorg_mitlearn_user_mapping.sql

@@ -0,0 +1,67 @@
+-- Referencing bridge_user_account_link (not just dim_user) is intentional:
+-- dim_user can fold unrelated accounts together when their computed emails
+-- collide, and its own output is the thing being validated, so it can't be
+-- used to catch its own collisions -- the bridge has the pre-collapse link.
+with mitxonline_edxorg_link as (
+    select from_user_id as mitxonline_application_user_id, to_user_id as user_edxorg_id
+    from {{ ref('bridge_user_account_link') }}
+    where from_platform = 'mitxonline' and to_platform = 'edxorg'
+)
+
+, users as (
+    select
+        dim_user.mitxonline_application_user_id
+        , dim_user.mitxonline_openedx_user_id
+        , case
+            when mitxonline_edxorg_link.user_edxorg_id = dim_user.edxorg_openedx_user_id
+                then dim_user.edxorg_openedx_user_id
+        end as edxorg_user_id
+        , dim_user.mitlearn_user_id as mitlearn_application_user_id
+    from {{ ref('dim_user') }} as dim_user
+    left join mitxonline_edxorg_link
+        on dim_user.mitxonline_application_user_id = mitxonline_edxorg_link.mitxonline_application_user_id
+    where dim_user.mitxonline_application_user_id is not null
+)
+
+, merges as (
+    select from_user_id as platform_user_id_before, to_user_id as platform_user_id_after
+    from {{ ref('bridge_user_account_link') }}
+    where from_platform = 'mitxonline' and to_platform = 'mitxonline'
+)
+
+-- A canonical account can have more than one account merged into it, so
+-- aggregate per canonical account. max() only stands in for coalesce here --
+-- the count(distinct ...) guard nulls the result out instead of guessing if
+-- two merged-away accounts carry genuinely conflicting non-null values.
+, merged_away_links as (
+    select
+        merges.platform_user_id_after as mitxonline_user_id
+        , case
+            when count(distinct users.edxorg_user_id) <= 1
+                then max(users.edxorg_user_id)
+        end as edxorg_user_id
+        , case
+            when count(distinct users.mitlearn_application_user_id) <= 1
+                then max(users.mitlearn_application_user_id)
+        end as mitlearn_application_user_id
+    from merges
+    inner join users
+        on merges.platform_user_id_before = users.mitxonline_application_user_id
+    group by merges.platform_user_id_after
+)
+
+, mapped as (
+    select
+        users.mitxonline_openedx_user_id
+        , coalesce(users.edxorg_user_id, merged_away_links.edxorg_user_id) as edxorg_user_id
+        , coalesce(users.mitlearn_application_user_id, merged_away_links.mitlearn_application_user_id)
+            as mitlearn_application_user_id
+    from users
+    left join merged_away_links
+        on users.mitxonline_application_user_id = merged_away_links.mitxonline_user_id
+    where users.mitxonline_application_user_id not in (select platform_user_id_before from merges)
+)
+
+select mitxonline_openedx_user_id, edxorg_user_id, mitlearn_application_user_id
+from mapped
+where edxorg_user_id is not null or mitlearn_application_user_id is not null

src/ol_dbt/models/staging/mitxonline/_mitxonline__sources.yml

@@ -411,6 +411,23 @@ sources:
         warn_after: {count: 1, period: day}
         error_after: {count: 2, period: day}
       loaded_at_field: "from_iso8601_timestamp(created_on || 'Z')"
+  - name: raw__mitxonline__app__postgres__courses_courserunenrollmentaudit
+    columns:
+    - name: id
+      description: int, sequential ID for this audit record
+    - name: enrollment_id
+      description: int, foreign key referencing courses_courserunenrollment
+    - name: acting_user_id
+      description: int, ID of the user (often an admin/support account) who made the
+        change
+    - name: data_before
+      description: JSON string, full enrollment record state before the change
+    - name: data_after
+      description: JSON string, full enrollment record state after the change
+    - name: created_on
+      description: timestamp, specifying when this audit record was created
+    - name: updated_on
+      description: timestamp, specifying when this audit record was last updated
   - name: raw__mitxonline__app__postgres__courses_programenrollment
     columns:
     - name: id
@@ -431,6 +448,24 @@ sources:
     - name: enrollment_mode
       description: str, enrollment mode for user
 
+  - name: raw__mitxonline__app__postgres__courses_programenrollmentaudit
+    columns:
+    - name: id
+      description: int, sequential ID for this audit record
+    - name: enrollment_id
+      description: int, foreign key referencing courses_programenrollment
+    - name: acting_user_id
+      description: int, ID of the user (often an admin/support account) who made the
+        change
+    - name: data_before
+      description: JSON string, full enrollment record state before the change
+    - name: data_after
+      description: JSON string, full enrollment record state after the change
+    - name: created_on
+      description: timestamp, specifying when this audit record was created
+    - name: updated_on
+      description: timestamp, specifying when this audit record was last updated
+
   - name: raw__mitxonline__app__postgres__courses_courserun
     columns:
     - name: id