Impact Analysis — user guide

Impact Analysis is a command-line tool that extracts SAP BusinessObjects universe and Web Intelligence document metadata into a relational database. Once the data is there, you query it with whatever you already have — SQL Server Management Studio, a Webi report, your favourite IDE, or an AI assistant pointed at the same database via dbquerymcp.

The first run loads everything; subsequent runs are incremental — only changed universes and documents are re-extracted. A water-mark timestamp and a per-cuid status table track what’s been seen and what failed.

Setup

1. Configure the database

Edit config/jdbc.json. SQLite is the default — nothing to install, the file sits next to the launcher:

{
  "driver": "org.sqlite.JDBC",
  "url": "jdbc:sqlite:impactanalysis.db",
  "username": "",
  "password": ""
}

For SQL Server:

{
  "driver": "com.microsoft.sqlserver.jdbc.SQLServerDriver",
  "url": "jdbc:sqlserver://HOST:1433;databaseName=DBNAME;encrypt=true;trustServerCertificate=true",
  "username": "USERNAME",
  "password": "PASSWORD"
}

For Oracle:

{
  "driver": "oracle.jdbc.OracleDriver",
  "url": "jdbc:oracle:thin:@HOST:1521:SID",
  "username": "USERNAME",
  "password": "PASSWORD"
}

All three drivers are bundled — no separate downloads required.

2. Configure BIP and SAP BO connection

Edit config/systems.json with the BIP REST endpoint plus BO Client Tools credentials (the tool drives webimaster for documents over REST and unxdoc for universes through the local Semantic Layer SDK):

{
    "lastSystem": "prod",
    "systems": [
        { "system": "prod", "username": "Administrator", "password": "..." }
    ]
}

3. Choose what to extract

Edit config/folders.txt. One folder path per line; comments start with #:

# CMS folders to scan for Webi documents and universes.
/Reporting/Sales
/Reporting/Operations
/Universes

Sub-folders are walked recursively. Anything outside the listed folders is ignored.

4. Apply the schema

impactanalysis-cli initdb

This applies impactanalysis’s own schema, plus the bundled webimaster and unxdoc schemas — one self-contained database, no sibling tools required.

Commands

A typical operational loop is refresh on a schedule (nightly, hourly), status to check health, retry-failures after fixing anything broken.

What’s in the database

Three logical layers, one schema:

The two join keys that bridge documents to universes:

These are stable cuids assigned by BO, not surrogate keys generated at extraction time — they survive renames and folder moves.

Example queries

Which documents are affected if I change a universe object?

Given a universe cuid and an object name, find the result objects that pull from it plus the variables and layout sites that reference those result objects.

WITH target_obj AS (
  SELECT universe_cuid, object_id
  FROM   universe_object
  WHERE  universe_cuid = :univ
    AND  name = :object_name              -- or filter by select_expr LIKE '%column%'
),
hit_ro AS (
  SELECT dp.document_cuid, ro.result_object_id, ro.name AS result_object_name
  FROM   document_data_provider dp
  JOIN   document_result_object ro ON ro.document_cuid = dp.document_cuid AND ro.dp_id = dp.dp_id
  JOIN   target_obj t              ON t.universe_cuid  = dp.data_source_cuid
                                  AND t.object_id      = ro.data_source_object_id
)
SELECT d.path, d.name AS document, hit_ro.result_object_name,
       ref.source_kind, ref.source_id, ref.target_name AS referenced_as
FROM   hit_ro
JOIN   document d              ON d.document_cuid = hit_ro.document_cuid
LEFT JOIN document_reference ref ON ref.document_cuid    = hit_ro.document_cuid
                                AND ref.target_kind      = 'result_object'
                                AND ref.target_id        = hit_ro.result_object_id
ORDER BY d.path, d.name;

source_kind = 'variable' rows tell you a variable formula references the object; source_kind = 'site' rows mean a report element (block, axis, sort, filter) references it. A NULL source_kind row means the object is selected in a query but never referenced downstream — usually safe to remove.

Which DB column does this universe object resolve to?

SELECT object_id, name, select_expr, where_expr
FROM   universe_object
WHERE  universe_cuid = :univ
  AND  (select_expr LIKE :pattern OR where_expr LIKE :pattern);

pattern is something like '%SCHEMA.TABLE.COLUMN%'. For derived tables you may need to walk universe_table.expression as well.

Which variables transitively depend on a result object?

document_reference is the explicit edge table. Variables → result objects is one hop; variables → variables → result objects is N hops. Use a recursive CTE:

WITH RECURSIVE chain(document_cuid, var_id, hop) AS (
  SELECT document_cuid, source_id, 1
  FROM   document_reference
  WHERE  source_kind = 'variable' AND target_kind = 'result_object' AND target_id = :ro_id
  UNION ALL
  SELECT r.document_cuid, r.source_id, c.hop + 1
  FROM   document_reference r
  JOIN   chain c
         ON r.document_cuid = c.document_cuid
        AND r.target_kind   = 'variable'
        AND r.target_id     = c.var_id
  WHERE  r.source_kind = 'variable' AND c.hop < 10
)
SELECT DISTINCT v.name, c.hop
FROM   chain c
JOIN   document_variable v ON v.document_cuid = c.document_cuid AND v.variable_id = c.var_id
ORDER BY c.hop, v.name;

Freshness

impactanalysis_run.water_mark is the timestamp the most recent refresh extracted up to. To know what to trust:

• Cuids with impactanalysis_cuid.status = 'ok' were extracted successfully on the last run that touched them.
• Cuids with status = 'failed' were attempted but did not land in the DB — treat them as stale until retry-failures succeeds.
• The water mark itself answers “as of when was this DB current?”.

Always check impactanalysis-cli status (or query impactanalysis_run) before relying on the data for a critical decision.

Limitations

• Universe expressions like @Select(), @DerivedTable(), @Variable() are stored verbatim — not expanded. A query against select_expr may need to traverse them.
• BEX queries are not processed.
• Documents that failed to open (corrupt, broken data provider, permissions) are skipped and logged as failures; their previous DB rows (if any) remain stale until the doc is fixed and retry-failures is run.