Complete function reference for SAP data integration in DuckDB
DuckDB Version: >= v1.2.0 Extensions: erpl_rfc, erpl_bics, erpl_odp
The ERPL extension suite brings SAP data integration directly into DuckDB. It enables analysts, data engineers, and scientists to query SAP ERP tables, SAP BW cubes, and replicate data using familiar SQL — no middleware, no data movement.
Key Benefits:
- SQL-native: All operations expressed as SQL functions, pragmas, and secrets
- Zero middleware: Connect directly to SAP via RFC and BICS protocols
- Parallel extraction: Multi-threaded reads for large tables
- Virtual catalog: ATTACH SAP systems as DuckDB databases
| Extension | Purpose |
|---|---|
| erpl_rfc | Core SAP RFC connectivity — table reads, function calls, metadata |
| erpl_bics | SAP BW queries via BICS — cubes, hierarchies, lineage |
| erpl_odp | SAP ODP data extraction and replication |
| (SSH tunnelling) | moved to the erpl_tunnel extension |
-- 1. Install and load
INSTALL 'erpl' FROM 'http://get.erpl.io';
LOAD 'erpl';
-- 2. Create SAP connection secret
CREATE SECRET my_sap (
TYPE sap_rfc,
ASHOST 'sap-server.example.com',
SYSNR '00',
CLIENT '100',
USER 'DEVELOPER',
PASSWD 'secret',
LANG 'EN'
);
-- 3. Read a table
SELECT * FROM sap_read_table('SFLIGHT');
-- 4. Attach SAP as a virtual database
ATTACH '' AS sap (TYPE sap_rfc);
SELECT * FROM sap."SFLIGHT" WHERE CARRID = 'LH';
-- 5. Query SAP BW
SELECT * FROM sap_bics_show_cubes();| Function | Purpose | Example |
|---|---|---|
sap_read_table |
Read SAP table data | SELECT * FROM sap_read_table('SFLIGHT') |
sap_rfc_invoke |
Call any RFC function | SELECT * FROM sap_rfc_invoke('STFC_CONNECTION', {'REQUTEXT': 'Hi'}) |
sap_show_tables |
Search SAP tables | SELECT * FROM sap_show_tables(TABLENAME='*FLIGHT*') |
sap_describe_fields |
Get table field metadata | SELECT * FROM sap_describe_fields('SFLIGHT') |
sap_rfc_authorizations |
List RFC modules each function uses (for S_RFC) | SELECT * FROM sap_rfc_authorizations() |
sap_bics_show_cubes |
List BW cubes | SELECT * FROM sap_bics_show_cubes() |
sap_bics_hierarchy |
Extract BW hierarchy | SELECT * FROM sap_bics_hierarchy('MY_HIER') |
sap_bics_set_char_prop |
AO-style char property (Display/Sort/Totals) | SELECT * FROM sap_bics_set_char_prop('q1', '0CNTRY', 'DISPLAY', 'TEXT') |
sap_odp_read_full |
Extract ODP data (full snapshot) | SELECT * FROM sap_odp_read_full('BW', 'MY_ODP') |
sap_odp_read_delta |
Extract ODP data (incremental delta) | SELECT * FROM sap_odp_read_delta('BW', 'MY_ODP', 'MY_PIPELINE') |
sap_odp_get_last_modified |
Last-modified timestamp of an ODP object (cheap delta probe) | SELECT * FROM sap_odp_get_last_modified('ABAP_CDS', 'MY_CDS$E') |
sap_odp_get_subscriptions |
List subscriptions for one ODP object | SELECT * FROM sap_odp_get_subscriptions('ABAP_CDS', 'MY_CDS$E') |
ATTACH |
Mount SAP as database | ATTACH '' AS sap (TYPE sap_rfc) |
Read data from an SAP table or CDS view. Supports projection pushdown, filter pushdown, and parallel reads.
| Parameter | Type | Default | Description |
|---|---|---|---|
table_name |
VARCHAR | required | SAP table or CDS view name |
THREADS |
UINTEGER | 0 | Number of parallel read threads |
COLUMNS |
LIST(VARCHAR) | all | Columns to retrieve |
FILTER |
VARCHAR | — | SAP WHERE clause filter |
fetch_size |
UINTEGER | erpl_rfc_fetch_size |
Concurrent result rows per round-trip; transport only, never changes the rows returned |
partitions |
UINTEGER | erpl_rfc_partitions |
Read this many row ranges in parallel. Same rows, unspecified order |
MAX_ROWS |
UINTEGER | 0 (all) | Maximum rows to return |
READ_TABLE_FUNCTION |
VARCHAR | 'RFC_READ_TABLE' |
RFC function to use (see note) |
READ_TABLE_DELIMITER |
VARCHAR | — | Delimiter for TABLE2 variants |
SECRET |
VARCHAR | — | Named secret to use |
Supported READ_TABLE_FUNCTION values: RFC_READ_TABLE, /BODS/RFC_READ_TABLE, /SAPDS/RFC_READ_TABLE, /BODS/RFC_READ_TABLE2, /SAPDS/RFC_READ_TABLE2
-- Basic read
SELECT * FROM sap_read_table('SFLIGHT');
-- With DuckDB filter pushdown (automatically pushed to SAP)
SELECT * FROM sap_read_table('SFLIGHT') WHERE CARRID = 'LH';
-- SAP-side filter, column selection, parallel threads, row limit
SELECT * FROM sap_read_table('SFLIGHT',
COLUMNS=['CARRID', 'CONNID', 'FLDATE'],
FILTER='CARRID = ''LH''',
THREADS=4,
MAX_ROWS=1000
);
-- Using a named secret
SELECT * FROM sap_read_table('SFLIGHT', SECRET='my_sap');Invoke any SAP RFC function module. Accepts variable arguments as STRUCT or scalar values.
| Parameter | Type | Default | Description |
|---|---|---|---|
function_name |
VARCHAR | required | RFC function module name |
...args |
ANY (varargs) | — | Function parameters as structs/values |
path |
VARCHAR | — | Path to select specific output (e.g. '/RFCTABLE') |
secret |
VARCHAR | — | Named secret to use |
-- Simple function call
SELECT * FROM sap_rfc_invoke('STFC_CONNECTION', {'REQUTEXT': 'Hello SAP'});
-- Select specific output table via path
SELECT * FROM sap_rfc_invoke('BAPI_FLIGHT_GETLIST',
{'AIRLINE': 'LH'},
path='/FLIGHT_LIST'
);Search for SAP tables and views. No positional arguments.
| Parameter | Type | Default | Description |
|---|---|---|---|
TABLENAME |
VARCHAR | '%' |
Table name pattern (supports * and % wildcards) |
TEXT |
VARCHAR | '%' |
Description text pattern |
THREADS |
UINTEGER | 0 | Parallel read threads |
Returns: table_name, text, class (VIEW, TRANSP, POOL, CLUSTER)
SELECT * FROM sap_show_tables(TABLENAME='*FLIGHT*');
SELECT * FROM sap_show_tables(TEXT='%booking%');Get field metadata for an SAP table or view.
| Parameter | Type | Default | Description |
|---|---|---|---|
table_name |
VARCHAR | required | Table/view name |
LANGUAGE |
VARCHAR | 'E' |
Language for descriptions |
Returns: pos, is_key, field, text, sap_type, length, decimals, check_table, ref_table, ref_field, language
SELECT field, is_key, sap_type, length FROM sap_describe_fields('SFLIGHT');Search for RFC function modules by name or group.
| Parameter | Type | Default | Description |
|---|---|---|---|
FUNCNAME |
VARCHAR | '*' |
Function name pattern (wildcard *) |
GROUPNAME |
VARCHAR | — | Function group name pattern |
LANGUAGE |
VARCHAR | — | Language for descriptions |
SELECT * FROM sap_rfc_show_function(FUNCNAME='BAPI_FLIGHT*');
SELECT * FROM sap_rfc_show_function(GROUPNAME='RSBOLAP_BICS');Search for RFC function groups.
| Parameter | Type | Default | Description |
|---|---|---|---|
GROUPNAME |
VARCHAR | '*' |
Group name pattern (wildcard *) |
LANGUAGE |
VARCHAR | — | Language for descriptions |
Returns: name, text
SELECT * FROM sap_rfc_show_groups(GROUPNAME='RSBOLAP*');Get detailed metadata about an RFC function module (parameters, types, structures).
| Parameter | Type | Default | Description |
|---|---|---|---|
function_name |
VARCHAR | required | Function module name |
SELECT * FROM sap_rfc_describe_function('STFC_CONNECTION');List which SAP RFC function modules each ERPL function invokes, so an SAP admin can scope the
S_RFC authorization object for the ERPL service user to least privilege. This is a static
reference — it needs no SAP connection (no secret required) and makes no RFC calls.
Returns: extension, duckdb_function, rfc_function_module, invocation, purpose
invocation is one of: always (called every time), fallback (one of a runtime-selected,
capability-dependent chain — e.g. the RFC_READ_TABLE variants), optional (attempted, skipped
gracefully on failure), metadata (a secondary DDIC/describe call), or user-specified (the FM is
the one you pass to sap_rfc_invoke). Note: opening a connection and sap_rfc_ping use the SDK
directly and invoke no function module.
-- All RFC modules to authorize for the whole ERPL suite
SELECT DISTINCT rfc_function_module FROM sap_rfc_authorizations()
WHERE rfc_function_module NOT IN ('<user-specified>', '<none>') ORDER BY 1;
-- Just what sap_read_table needs
SELECT rfc_function_module, invocation, purpose
FROM sap_rfc_authorizations() WHERE duckdb_function = 'sap_read_table';Test SAP RFC connection. Returns 'PONG' on success.
| Parameter | Type | Default | Description |
|---|---|---|---|
secret |
VARCHAR | — | Named secret to use |
PRAGMA sap_rfc_ping;
PRAGMA sap_rfc_ping(secret='my_sap');Which implementation is serving SAP RFC calls: 'nwrfc' or 'proto'. Resolves the backend
if that has not happened yet. See Selecting the RFC backend.
SELECT sap_rfc_backend();Set SAP NetWeaver RFC SDK trace level.
| Parameter | Type | Description |
|---|---|---|
level |
INTEGER | 0=Off, 1=Brief, 2=Verbose, 3=Detailed, 4=Full |
PRAGMA sap_rfc_set_trace_level(2);Set directory for SAP RFC SDK trace files.
PRAGMA sap_rfc_set_trace_dir('/tmp/sap_traces');Set maximum size for SAP RFC trace files.
| Parameter | Type | Description |
|---|---|---|
size |
INTEGER | File size value |
unit |
ENUM | 'M' (megabytes) or 'G' (gigabytes) |
PRAGMA sap_rfc_set_maximum_trace_file_size(100, 'M');Set maximum number of trace files to keep.
PRAGMA sap_rfc_set_maximum_stored_trace_files(10);Set path to SAP RFC INI configuration file (sapnwrfc.ini).
PRAGMA sap_rfc_set_ini_path('/etc/sap/sapnwrfc.ini');Reload the SAP RFC INI configuration file.
PRAGMA sap_rfc_reload_ini_file;Attach an SAP system as a virtual DuckDB database. Tables appear as views backed by sap_read_table().
-- Attach with default secret
ATTACH '' AS sap (TYPE sap_rfc);
SELECT * FROM sap."SFLIGHT";
-- Attach with named secret
ATTACH '' AS sap (TYPE sap_rfc, SECRET 'my_sap');
-- Restrict to specific tables
ATTACH '' AS sap (TYPE sap_rfc, TABLES 'SFLIGHT,SPFLI,SCARR');
-- Scope to tables matching a glob pattern (resolved against the dictionary at ATTACH)
ATTACH '' AS sap (TYPE sap_rfc, TABLES '/DMO/*,Z*');
SHOW TABLES FROM sap; -- lists the resolved set
-- Detach
DETACH sap;| Option | Type | Description |
|---|---|---|
TYPE |
— | Must be sap_rfc |
SECRET |
VARCHAR | Named secret for SAP connection |
TABLES |
VARCHAR | Comma-separated list of exact table names and/or glob patterns (*, ?) to expose. Empty = on-demand lookup. |
SHOW TABLES and table enumeration. A SAP system exposes tens of thousands of
tables, so an attached catalog does not list them all. SHOW TABLES FROM <catalog>
(and information_schema.tables) reflect only the tables named or matched by TABLES:
- With
TABLES(exact names and/or patterns) → those tables are listed and queryable; tables outside the set are not accessible through the catalog. - Without
TABLES→ tables are resolved on demand when referenced by name (SELECT * FROM sap."SFLIGHT"), andSHOW TABLESis empty by design.
Patterns use * (any run of characters) and ? (single character) and are resolved once,
at ATTACH time, against the data dictionary (DD02V, the same source as
sap_show_tables()). To browse the full catalog without scoping, use
sap_show_tables().
List InfoProviders, cubes, queries, or info areas.
| Parameter | Type | Default | Description |
|---|---|---|---|
obj_type |
ENUM | — | 'INFOPROVIDER', 'QUERY', 'CUBE', 'INFOAREA' |
search |
VARCHAR | — | Search pattern (supports wildcards) |
search_in_key |
BOOLEAN | — | Search in technical names |
search_in_text |
BOOLEAN | — | Search in descriptions |
fetch_levels |
UINTEGER | — | Number of hierarchy levels to fetch |
secret |
VARCHAR | — | Named secret to use |
Returns: technical_name, text, type, cube_name, is_folder, last_changed, last_changed_by, level
SELECT * FROM sap_bics_show();
SELECT * FROM sap_bics_show(obj_type='QUERY', search='*SALES*');Convenience function to list only BW cubes. Same parameters as sap_bics_show minus obj_type.
SELECT * FROM sap_bics_show_cubes();
SELECT * FROM sap_bics_show_cubes(search='*SALES*');Convenience function to list only BW queries.
SELECT * FROM sap_bics_show_queries();List available BW hierarchies.
| Parameter | Type | Default | Description |
|---|---|---|---|
search |
VARCHAR | — | Search pattern for hierarchy names |
info_object |
VARCHAR | — | Filter by InfoObject |
secret |
VARCHAR | — | Named secret |
Returns: technical_name, text, version, valid_to_date
SELECT * FROM sap_bics_show_hierarchies();
SELECT * FROM sap_bics_show_hierarchies(info_object='0COSTCENTER');Retrieve metadata structure for cubes or queries.
| Parameter | Type | Default | Description |
|---|---|---|---|
cube_name |
VARCHAR | required | Cube technical name |
query_name |
VARCHAR | — | Query technical name |
id |
VARCHAR | — | State ID for persisted session |
version |
UINTEGER | — | State version |
secret |
VARCHAR | — | Named secret |
-- Describe a cube
SELECT * FROM sap_bics_describe('MY_CUBE');
-- Describe a query
SELECT * FROM sap_bics_describe('MY_CUBE', 'MY_QUERY');Get technical details about a BW InfoObject.
Returns: info_object, data_type, conv_exit, output_length, length, decimals
SELECT * FROM sap_bics_describe_infoobject('0COSTCENTER');Extract hierarchy structure with node relationships and paths.
| Parameter | Type | Default | Description |
|---|---|---|---|
hierarchy_name |
VARCHAR | required | Hierarchy technical name |
version |
VARCHAR | 'A' |
Hierarchy version |
date_to |
VARCHAR | — | Valid-to date (YYYYMMDD) |
secret |
VARCHAR | — | Named secret |
Returns: node_id, parent_id, child_id, next_id, info_object, node_name, node_value, date_from, date_to, level, path
SELECT * FROM sap_bics_hierarchy('ZCOSTCENTER_H01');
SELECT * FROM sap_bics_hierarchy('ZCOSTCENTER_H01', version='A', date_to='20251231');BICS queries use a stateful workflow: initialize a session, configure axes and filters, then fetch results.
Initialize a BICS query session. cube_name may also be a BEx query technical
name, which is resolved to its InfoProvider via RSRREPDIR.
| Parameter | Type | Default | Description |
|---|---|---|---|
cube_name |
VARCHAR | required | Cube or BEx query technical name |
id |
VARCHAR | — | User-defined state ID |
return |
BICS_RETURN | 'DESCRIBE' |
'DESCRIBE' for metadata, 'RESULT' for data |
variables |
LIST<STRUCT> | — | BEx variable values (see below) |
hierarchy_variables |
LIST<STRUCT> | — | Hierarchy-node variable values (see below) |
variant |
VARCHAR | — | Name of a saved BEx variant to fill the variables |
secret |
VARCHAR | — | Named secret |
SELECT * FROM sap_bics_begin('MY_CUBE', id='my_session');variables shape — LIST<STRUCT(NAME, SIGN, OP, LOW, HIGH)>, mapped onto
SAP's BICS_PROV_STATE_INIT_VARIABLES row type and submitted with
BICS_PROV_OPEN. It mirrors the ODP filters shape, with NAME naming a BEx
variable instead of a field:
| Field | Type | Semantics |
|---|---|---|
NAME |
VARCHAR | BEx variable technical name |
SIGN |
VARCHAR | 'I' (include) or 'E' (exclude); defaults to 'I' |
OP |
VARCHAR | 'EQ', 'BT', 'GE', 'LE', 'CP', …; defaults to 'BT' when HIGH is given, else 'EQ' |
LOW |
VARCHAR | Value, or lower bound of an interval |
HIGH |
VARCHAR | Upper bound of an interval; empty otherwise |
Repeat NAME to fill a multi-value variable — that is how multiple values
travel on the wire.
SELECT * FROM sap_bics_begin('MY_QUERY', id='q1',
variables => [
{'NAME':'ZVAR_YEAR', 'SIGN':'I', 'OP':'EQ', 'LOW':'2026', 'HIGH':''},
{'NAME':'ZVAR_DATE', 'SIGN':'I', 'OP':'BT', 'LOW':'20260101', 'HIGH':'20260131'}
]);Values are stored with the session state and replayed whenever the session is
restored, so a chained begin → rows → result workflow stays restricted.
hierarchy_variables shape — LIST<STRUCT(NAME, LOW, HIERARCHY_NAME, HIERARCHY_VERSION, HIERARCHY_DUE_DATE)>. Separate from variables so the
common case does not have to spell out the hierarchy fields;
HIERARCHY_DUE_DATE defaults to '99991231'.
A variable name the query does not expose as input-ready is rejected — BW would otherwise ignore it silently and return an unrestricted result.
List the BEx variables of a query, to find out what variables has to fill.
| Column | Type | Description |
|---|---|---|
name |
VARCHAR | Technical name |
text |
VARCHAR | Description |
mandatory |
BOOLEAN | BW refuses to produce a result until this one has a value |
input_enabled |
BOOLEAN | Can be filled by the caller (exit variables cannot) |
is_exit_variable |
BOOLEAN | Filled by a BW customer-exit, not by the caller |
var_type |
VARCHAR | CHARACTERISTIC_VALUE, HIERARCHY, TEXT, FORMULA, HIERARCHY_NODE |
selection_type |
VARCHAR | SINGLE_VALUE, INTERVAL, SELECTION_OPTION, MULTIPLE_VALUES, PRECALCULATED_VALUE_SET |
entry_type |
VARCHAR | OPTIONAL, MANDATORY, MANDATORY_NOT_INITIAL |
reference_char |
VARCHAR | Characteristic a hierarchy-node variable refers to |
SELECT * FROM sap_bics_variables('MY_QUERY');Configure row axis characteristics.
| Parameter | Type | Default | Description |
|---|---|---|---|
state_id |
VARCHAR | required | State ID from sap_bics_begin |
char1, ... |
VARCHAR | required | Characteristics to add/set |
op |
BICS_OPERATION | 'ADD' |
'SET', 'ADD', or 'REMOVE' |
return |
BICS_RETURN | 'DESCRIBE' |
Return format |
SELECT * FROM sap_bics_rows('my_session', '0CALDAY', '0MATERIAL');Configure column axis characteristics. Same signature as sap_bics_rows.
SELECT * FROM sap_bics_columns('my_session', '0AMOUNT', op='SET');Restrict a characteristic to specific member values (AO "Keep Only" /
"Exclude" gesture). Members are positional varargs after the characteristic
name; pass zero members with op='SET' to clear the filter.
| Parameter | Type | Default | Description |
|---|---|---|---|
state_id |
VARCHAR | required | State ID |
char_name |
VARCHAR | required | Characteristic technical name |
member1, … |
VARCHAR (varargs) | — | Member values to include (or remove with op='REMOVE') |
op |
BICS_OPERATION | 'ADD' if members supplied, else 'SET' |
'SET', 'ADD', 'REMOVE' |
return |
BICS_RETURN | 'DESCRIBE' |
Return format |
-- "Keep Only Germany and France" on country
SELECT * FROM sap_bics_filter('q1', '0D_NW_CNTRY', 'DE', 'FR', op='SET');
-- Append USA to the existing selection
SELECT * FROM sap_bics_filter('q1', '0D_NW_CNTRY', 'US', op='ADD');
-- Clear the filter
SELECT * FROM sap_bics_filter('q1', '0D_NW_CNTRY', op='SET');Set an AO-style per-characteristic property (Display, Sort, or Totals
visibility). Mutates /E_TH_STATE_CHARACTERISTICS/<id>/<field> on the
server and the change is honoured by the next sap_bics_result call.
| Parameter | Type | Default | Description |
|---|---|---|---|
state_id |
VARCHAR | required | State ID |
char_name |
VARCHAR | required | Characteristic technical name |
prop |
VARCHAR | required | One of 'DISPLAY', 'TOTALS', 'SORT' |
value |
VARCHAR | required | See table below |
return |
BICS_RETURN | 'DESCRIBE' |
Return format |
prop |
Allowed value |
Server field mutated |
|---|---|---|
DISPLAY |
'KEY' | 'TEXT' | 'BOTH' |
RESULT_SET_PRESENTATION + MEMBER_ACCESS_PRESENTATION (bitflag: KEY=4, TEXT=32) |
TOTALS |
'SHOW' | 'HIDE' |
RESULT_VISIBILITY ('A' / 'N') |
SORT |
'ASC' | 'DESC' | 'NONE' |
RESULT_SET_SORTING.DIRECTION ('A' / 'D' / '') |
-- Show member texts instead of keys
SELECT * FROM sap_bics_set_char_prop('q1', '0D_NW_CNTRY', 'DISPLAY', 'TEXT');
-- Sort products descending
SELECT * FROM sap_bics_set_char_prop('q1', '0D_NW_PROD', 'SORT', 'DESC');The result-set parser is display-mode aware: when BICS returns both key and
text presentation entries for a member, the renderer picks key / text / both
according to the per-char setting. The DESCRIBE payload of state_rows,
state_columns, and state_free now carries {display, totals, sort} so
clients can read the current values without an extra round-trip.
Note: the grand-total "SUMME" / "Overall Result" row in
sap_bics_resultis not affected by per-charTOTALS='HIDE'— BICS does not expose a state field for grand-total visibility. The flag is applied to the server state (visible in DESCRIBE) but the SUMME row still appears in the result set; downstream clients can filter it out client-side. AO does the same.
Fetch result set from configured query state.
| Parameter | Type | Default | Description |
|---|---|---|---|
state_id |
VARCHAR | required | State ID |
Result shape — one VARCHAR column per row-axis element carrying the formatted
member key or text, one INTEGER <name>_HIER_LEVEL column per row-axis element, then
one DOUBLE column per column-axis leaf.
The number of row-axis columns follows the result set the BW server returns; when a result carries no rows at all, the session state's axis is used instead so a query's schema does not change shape just because it returned nothing.
Column names are resolved in order: the session state, then the query's design-time
metadata (one extra RFC call, made only when the state cannot name the axis — the case
for BEx queries whose drilldown lives in the query definition), then ROW_1, ROW_2,
…. Both metadata sources are matched on the member's characteristic id, never by
position, so a column is never labelled with an unrelated characteristic.
-- Complete workflow
SELECT * FROM sap_bics_begin('MY_CUBE', id='q1');
SELECT * FROM sap_bics_rows('q1', '0CALDAY', '0MATERIAL');
SELECT * FROM sap_bics_columns('q1', '0AMOUNT');
SELECT * FROM sap_bics_set_char_prop('q1', '0CALDAY', 'SORT', 'DESC'); -- optional
SELECT * FROM sap_bics_filter('q1', '0MATERIAL', 'M01', 'M02', op='SET'); -- optional
SELECT * FROM sap_bics_result('q1');Functions for extracting BW system metadata. All accept an optional secret parameter.
| Function | Description |
|---|---|
sap_bics_meta_providers([provider_type, secret]) |
List BW InfoProviders by type (CUBE, ADSO, HCPR, ODSO, ODSVIEW) |
sap_bics_meta_provider_fields(provider_name [, secret]) |
Fields/characteristics for a provider |
sap_bics_meta_datasources([secret]) |
List all DataSources |
sap_bics_meta_datasource_fields(datasource_name [, secret]) |
Fields for a DataSource |
sap_bics_meta_transformations([secret]) |
List all transformations |
sap_bics_meta_transform_fields(transformation_name [, secret]) |
Field mappings and rules |
sap_bics_meta_hcpr([secret]) |
List all CompositeProviders |
sap_bics_meta_hcpr_mapping(hcpr_name [, secret]) |
Field mappings for CompositeProviders |
sap_bics_meta_queries([secret]) |
List all BW queries |
sap_bics_meta_query_usage(query_name [, secret]) |
Query usage and dependencies |
sap_bics_meta_query_elements(query_name [, secret]) |
Detailed query structure elements |
sap_bics_meta_query_stats([secret]) |
Query performance and usage statistics |
sap_bics_meta_objxref([secret]) |
Dependencies between BW objects |
sap_bics_meta_infoobjects([secret]) |
List all InfoObjects |
SELECT * FROM sap_bics_meta_providers(provider_type='CUBE');
SELECT * FROM sap_bics_meta_provider_fields('MY_CUBE');
SELECT * FROM sap_bics_meta_queries();Functions for BW data lineage extraction. All accept an optional secret parameter.
Extract lineage edges between BW objects.
Returns: edge_type, src_kind, src_name, src_field, tgt_kind, tgt_name, tgt_field
SELECT * FROM sap_bics_lineage_edges();Full lineage graph in JSON format.
SELECT * FROM sap_bics_lineage_graph_json();Trace lineage from a specific object.
SELECT * FROM sap_bics_lineage_trace('MY_CUBE');Complete field-level lineage for a specific BW query.
SELECT * FROM sap_bics_query_lineage('MY_QUERY');List available ODP contexts.
Returns: technical_name, text, release
SELECT * FROM sap_odp_show_contexts();
-- Typical contexts: BW, ABAP_CDS, SAPI, SLT, HANAList ODP providers/sources within a context.
| Parameter | Type | Default | Description |
|---|---|---|---|
odp_context |
VARCHAR | required | ODP context (e.g. 'BW', 'ABAP_CDS', 'SAPI') |
search |
VARCHAR | — | Search pattern (wildcards) |
secret |
VARCHAR | — | Named secret |
Returns: technical_name, text, semantics, semantics_text
SELECT * FROM sap_odp_show('BW');
SELECT * FROM sap_odp_show('ABAP_CDS', search='*FLIGHT*');Get detailed metadata about an ODP provider.
SELECT * FROM sap_odp_describe('BW', 'MY_ODP_SOURCE');Preview a small sample of data from an ODP provider.
SELECT * FROM sap_odp_preview('BW', 'MY_ODP_SOURCE');Full extraction of ODP provider data with parallel processing. The cursor on
the SAP side is opened in FULL mode and auto-closed when the scan ends —
this is one-shot. For incremental extraction use
sap_odp_read_delta.
| Parameter | Type | Default | Description |
|---|---|---|---|
odp_context |
VARCHAR | required | ODP context |
odp_name |
VARCHAR | required | ODP provider name |
threads |
UINTEGER | 5 | Parallel read threads |
columns |
LIST(VARCHAR) | all | Columns to extract |
filters |
LIST(STRUCT) | — | Server-side selection filters (see below) |
secret |
VARCHAR | — | Named secret |
filters shape — LIST<STRUCT(FIELDNAME, SIGN, OP, LOW, HIGH)> mapped
onto SAP's RODPS_REPL_S_SELECTION row type. Predicates are OR-combined
within the list:
| Field | Type | Semantics |
|---|---|---|
FIELDNAME |
VARCHAR | ODP source field name |
SIGN |
VARCHAR | 'I' (include) or 'E' (exclude) |
OP |
VARCHAR | 'EQ', 'NE', 'GT', 'LT', 'GE', 'LE', 'BT', 'CP', ... |
LOW |
VARCHAR | Lower / single-value comparand |
HIGH |
VARCHAR | Upper comparand (only for BT) |
-- Simple equality
SELECT * FROM sap_odp_read_full('BW', 'MY_ODP_SOURCE');
SELECT * FROM sap_odp_read_full('ABAP_CDS', 'MY_CDS_VIEW', threads=8);
-- Server-side range filter (between two business partners, inclusive)
SELECT * FROM sap_odp_read_full('ABAP_CDS', 'SEPM_IBUPA$P',
COLUMNS=['BUSINESSPARTNER','COMPANYNAME'],
FILTERS=[{
'FIELDNAME': 'BUSINESSPARTNER',
'SIGN': 'I',
'OP': 'BT',
'LOW': '0100000000',
'HIGH': '0100000099'
}]);Incremental delta extraction. The first call with a given subscriber_process
performs SAP's auto-DELTAINIT — returns the full current snapshot AND
registers a server-side delta pointer keyed by the subscriber tuple. Subsequent
calls with the same subscriber_process resume from the previous pointer and
return only the changes since then.
The cursor persists across calls (FULL cursors auto-close on scan completion;
DELTA cursors do not). Close them explicitly with
PRAGMA sap_odp_close_delta_cursor when
your pipeline is done.
| Parameter | Type | Default | Description |
|---|---|---|---|
odp_context |
VARCHAR | required | ODP context (e.g. 'BW', 'ABAP_CDS') |
odp_name |
VARCHAR | required | ODP provider name |
subscriber_process |
VARCHAR | required | Stable identifier that keys the server-side delta pointer across calls. Choose a deterministic name per pipeline (e.g. 'MY_ETL_BUPA_DAILY'). |
threads |
UINTEGER | 1 | Accepted for API symmetry but delta/RECOVER fetch is serialized (capped to 1). Parallel package fetch can race a multi-package delta into an under-count, and delta packets are small; use sap_odp_read_full threads for large parallel snapshots. Must be ≥ 1. |
columns |
LIST(VARCHAR) | all | Columns to extract (not allowed together with recover=true) |
filters |
LIST(STRUCT) | — | Server-side selection predicates (same shape as sap_odp_read_full; not allowed together with recover=true) |
recover |
BOOLEAN | false |
When true, re-stream the last unconfirmed packet (I_EXTRACTION_MODE='R') without advancing the pointer, using the original open's projection/filters. Useful after a fetch was interrupted. Cannot be combined with columns/filters. |
secret |
VARCHAR | — | Named secret |
Concurrency note: do not run two sap_odp_read_delta calls with the same
subscriber_process in parallel — they will race the server-side pointer.
Change semantics: for ABAP_CDS byElement-tracked sources, inserts and
updates both surface as after-images (ODQ_CHANGEMODE='U'); physical deletes are
not reported by the byElement annotation (the row simply leaves the snapshot).
-- First call: SAP auto-DELTAINIT — returns full current snapshot, registers
-- a server-side delta pointer under subscriber_process 'NIGHTLY_ETL'.
SELECT * FROM sap_odp_read_delta('BW', '0D_FC_C01$F', 'NIGHTLY_ETL');
-- Run the same call later (e.g. next day): returns only the rows that
-- changed in the cube since the previous call.
SELECT * FROM sap_odp_read_delta('BW', '0D_FC_C01$F', 'NIGHTLY_ETL');
-- If a previous DELTA call was interrupted, re-stream its last packet:
SELECT * FROM sap_odp_read_delta('BW', '0D_FC_C01$F', 'NIGHTLY_ETL',
recover=true);
-- Release the cursor when done:
PRAGMA sap_odp_close_delta_cursor('BW', 'NIGHTLY_ETL', '0D_FC_C01$F');Delta-capable sources are identified by supports_delta=true in
sap_odp_describe. Not every CDS view is delta-capable —
the underlying DDL must carry the @Analytics.dataExtraction.delta.byElement
annotation; BW fact tables (*$F) are typically delta-capable.
List active ODP subscriptions.
SELECT * FROM sap_odp_show_subscriptions();List ODP extraction cursors for delta tracking. Cursors created by
sap_odp_read_delta appear here with is_delta_extension=true; their
subscriber_proc column matches the subscriber_process argument you passed.
SELECT * FROM sap_odp_show_cursors();Return the last-modified UTC timestamp of an ODP object (invokes
RODPS_REPL_ODP_GET_LAST_MODIF) without fetching any rows. Use it as a cheap
probe before sap_odp_read_delta: if the timestamp has not advanced since your
last run, there is nothing to extract.
Output columns: odp_name (VARCHAR), last_modified (DECIMAL(21,7) — a SAP UTC
timestamp; comparable against the pointer column of sap_odp_show_cursors).
SELECT * FROM sap_odp_get_last_modified('ABAP_CDS', 'MY_CDS_VIEW$E');List the subscriptions registered for a specific ODP object (invokes
RODPS_REPL_ODP_GET_SUBSCR). Unlike sap_odp_show_subscriptions (which lists all
subscriptions), this is scoped to one odp_name and can be further filtered.
Output columns: subscriber_type, subscriber_name, subscriber_process,
queue_name, subscription_id.
| Parameter | Type | Description |
|---|---|---|
odp_context |
VARCHAR | ODP context (e.g. 'BW', 'ABAP_CDS') |
odp_name |
VARCHAR | ODP object name |
subscriber_name |
VARCHAR | (named) optional subscriber-name filter |
subscriber_process |
VARCHAR | (named) optional subscriber-process filter |
secret |
VARCHAR | (named) optional secret name |
SELECT * FROM sap_odp_get_subscriptions('ABAP_CDS', 'MY_CDS_VIEW$E');Scalar functions returning how many SAP RFC connections erpl has opened and closed in
this process, and how many it currently holds open (opened - closed).
Every open connection is a session and a work-process reservation on the SAP system, so
sap_rfc_live_connections() is the number that matters to a Basis team. Between
queries it should be 0. A non-zero value means erpl is still holding SAP sessions.
SELECT sap_rfc_live_connections(); -- 0 between queriesThese are process-wide counters, not per-connection state, and they are VOLATILE so
DuckDB never constant-folds them at bind time. Their intended use is asserting in tests
and in the field that a scan released what it acquired — client-side timing shows nothing
when a connection is never released, because the entire cost falls on the SAP system.
Graceful counterpart to sap_odp_drop. Looks up the cursor for the given
subscriber tuple and calls RODPS_REPL_ODP_CLOSE on its pointer. Idempotent:
returns 'CLOSED' if a cursor existed (open or already closed) and
'NOT_FOUND' if no cursor of that name was found.
Prefer this over sap_odp_drop at the end of a delta pipeline — close leaves
the subscription registered and resumable from its last pointer; drop wipes the
subscription so the next call performs DELTAINIT again.
| Parameter | Type | Description |
|---|---|---|
odp_context |
VARCHAR | ODP context (e.g. 'BW', 'ABAP_CDS') |
subscriber_process |
VARCHAR | Subscriber process identifier (the one passed to sap_odp_read_delta) |
odp_name |
VARCHAR | ODP object name |
Named parameters: secret — optional secret name.
PRAGMA sap_odp_close_delta_cursor('BW', 'NIGHTLY_ETL', '0D_FC_C01$F');Cleanup hygiene matters: leaked open cursors accumulate server-side. If a
pipeline crashes mid-fetch the cursor may not be closeable via this pragma —
fall back to sap_odp_drop to fully reset, accepting that the next call
will re-snapshot via DELTAINIT.
Drop an ODP subscription/cursor on the SAP system (invokes RODPS_REPL_ODP_RESET).
A subsequent sap_odp_read_full will re-create the subscription from scratch.
| Parameter | Type | Description |
|---|---|---|
odp_context |
VARCHAR | ODP context (e.g. 'ABAP_CDS', 'BW', 'SAPI') |
subscriber_name |
VARCHAR | Subscriber name registered against the queue |
subscriber_process |
VARCHAR | Subscriber process identifier |
odp_name |
VARCHAR | ODP object name |
Named parameters: secret — optional secret name.
-- Find the subscription to drop
SELECT queue_name, subscriber_name, subscriber_proc
FROM sap_odp_show_subscriptions();
-- Drop it (values come from the row above)
PRAGMA sap_odp_drop(
'ABAP_CDS',
'ERPL_BW',
'ERPL_ABAP_CDS_SEPM_IBUPA_4271',
'SEPM_IBUPA$P'
);The SSH tunnel that erpl used to bundle now lives in a dedicated extension, erpl_tunnel, which does more: reverse tunnels, Tailscale and NetBird backends, and peer discovery.
INSTALL erpl_tunnel FROM 'http://get.erpl.io';
LOAD erpl_tunnel;tunnel_create, tunnel_close, tunnel_close_all and tunnels() remain registered in
erpl as stubs that raise and point here, so an old script says where the function went
rather than failing with "does not exist". Loading erpl_tunnel replaces them.
Two differences from the version erpl bundled: tunnels() gained backend and
direction columns, so column order changed (queries naming columns are unaffected),
and tunnel_create is a deprecated alias for tunnel_import. The ssh_tunnel secret type
is unchanged, so existing secrets keep working.
All parameters are VARCHAR. Choose either direct connection or load-balanced parameters.
| Parameter | Description | Direct | Load-Balanced |
|---|---|---|---|
ashost |
Application server host | required | — |
sysnr |
System number | required | — |
mshost |
Message server host | — | required |
msserv |
Message server service | — | optional |
sysid |
System ID | — | required |
group |
Logon group | — | required |
client |
SAP client number | required | required |
user |
Username | required | required |
passwd |
Password (redacted) | required¹ | required¹ |
lang |
Language (e.g. 'EN') |
optional | optional |
snc_mode |
Activate SNC — '1' on, '0' off |
optional | optional |
snc_sso |
Use the SNC identity for logon ('1'/'0') |
optional | optional |
snc_qop |
SNC quality of protection | optional | optional |
snc_myname |
SNC client name | optional | optional |
snc_partnername |
SNC server name | optional | optional |
snc_lib |
SNC library path | optional | optional |
mysapsso2 |
SSO2 ticket (redacted) | optional | optional |
x509cert |
X.509 certificate for logon (redacted) | optional | optional |
saprouter |
SAProuter string, e.g. '/H/routerhost/S/3299' |
optional | optional |
gwhost |
Gateway host | optional | optional |
gwserv |
Gateway service | optional | optional |
codepage |
Logon codepage, e.g. '4103' |
optional | optional |
trace |
SAP RFC trace level '0'–'3' |
optional | optional |
dest |
Destination in sapnwrfc.ini |
optional | optional |
¹ Not required when logging on via SNC or an SSO2 ticket.
The parameter names are passed through to RfcOpenConnection unchanged, so they
follow the SAP NetWeaver RFC SDK documentation.
-- Direct connection
CREATE SECRET my_sap (
TYPE sap_rfc,
ASHOST 'sap.example.com', SYSNR '00', CLIENT '100',
USER 'sapuser', PASSWD 'password', LANG 'EN'
);
-- Load-balanced connection
CREATE SECRET my_sap_lb (
TYPE sap_rfc,
MSHOST 'sapms.example.com', SYSID 'PRD', GROUP 'PUBLIC',
CLIENT '100', USER 'sapuser', PASSWD 'password'
);
-- SNC / Kerberos logon, without a password
CREATE SECRET my_sap_snc (
TYPE sap_rfc,
ASHOST 'sap.example.com', SYSNR '00', CLIENT '100', USER 'sapuser',
SNC_MODE '1',
SNC_PARTNERNAME 'p:CN=SAP/sap.example.com@EXAMPLE.LOCAL',
SNC_LIB '/usr/lib/libgsskrb5.so'
);A secret is selected by name via the secret argument, e.g.
PRAGMA sap_rfc_ping(secret='my_sap_snc') or
SELECT * FROM sap_read_table('SFLIGHT', secret='my_sap_snc'). Without it, the
best-matching sap_rfc secret in scope is used.
| Parameter | Type | Default | Description |
|---|---|---|---|
ssh_host |
VARCHAR | required | SSH jump host |
ssh_port |
INTEGER | 22 | SSH port |
ssh_user |
VARCHAR | required | SSH username |
password |
VARCHAR | — | SSH password (redacted) |
private_key_path |
VARCHAR | — | Path to SSH private key (redacted) |
passphrase |
VARCHAR | — | Private key passphrase (redacted) |
auth_method |
VARCHAR | auto | 'password', 'key', or 'agent' (auto-detected if omitted) |
-- Password authentication
CREATE SECRET my_ssh (
TYPE ssh_tunnel,
ssh_host 'jumphost.example.com', ssh_user 'myuser',
password 'secret'
);
-- Key-based authentication
CREATE SECRET my_ssh_key (
TYPE ssh_tunnel,
ssh_host 'jumphost.example.com', ssh_user 'myuser',
private_key_path '/home/user/.ssh/id_rsa'
);| Type | Values | Used By |
|---|---|---|
BICS_RETURN |
'DESCRIBE', 'RESULT' |
BICS query workflow functions |
BICS_OPERATION |
'SET', 'ADD', 'REMOVE' |
BICS axis/filter configuration |
ODP_REPLICATION_MODE |
'FULL', 'DELTA', 'RECOVER' |
ODP extraction mode |
The DDIC name on the left is what sap_describe_fields() reports; the RFCTYPE
shown in parentheses is the internal RFC-SDK type the ERPL scanners use.
| DDIC type | RFCTYPE | DuckDB type | Notes |
|---|---|---|---|
| CHAR / CLNT / LANG / CUKY / UNIT | RFCTYPE_CHAR | VARCHAR | Fixed-length, right-trimmed |
| NUMC / ACCP | RFCTYPE_NUM | VARCHAR | Numeric characters; preserves leading zeros |
| INT4 | RFCTYPE_INT | BIGINT | 4-byte integer (signed) |
| INT1 | RFCTYPE_INT1 | TINYINT | 1-byte integer |
| INT2 / PREC | RFCTYPE_INT2 | SMALLINT | 2-byte integer (PREC is a DDIC alias) |
| INT8 | RFCTYPE_INT8 | BIGINT | 8-byte integer (signed) |
| FLTP | RFCTYPE_FLOAT | DOUBLE | IEEE 754 floating point |
| DEC / CURR / QUAN | RFCTYPE_BCD | DECIMAL(min(2N-1,38), S) | Packed-decimal; precision capped at DuckDB's max |
| DECF16 / D16D / D16N / D16R / D16S | RFCTYPE_DECF16 | DECIMAL(16, S) | IEEE 754-2008 decimal floating point; precision is the spec-defined 16, scale clamped to it |
| DECF34 / D34D / D34N / D34R / D34S | RFCTYPE_DECF34 | DECIMAL(34, S) | IEEE 754-2008 decimal floating point; precision is the spec-defined 34, scale clamped to it |
| STRING / STRG / SSTR / LCHR | RFCTYPE_STRING | VARCHAR | Variable-length character data |
| XMLDATA | RFCTYPE_XMLDATA | VARCHAR | XML payload (text) |
| RAW / LRAW | RFCTYPE_BYTE | BLOB | Fixed-length raw bytes (incl. RAW(16) UUIDs). See note below |
| RAWSTRING / RSTR | RFCTYPE_XSTRING | BLOB | Variable-length raw bytes. See note below |
| DATS | RFCTYPE_DATE | DATE | Format YYYYMMDD |
| TIMS | RFCTYPE_TIME | TIME | Format HHMMSS |
| UTCLONG / UTCL | RFCTYPE_UTCLONG | TIMESTAMP | UTC timestamp, microsecond precision |
| UTCS | RFCTYPE_UTCSECOND | TIMESTAMP | UTC timestamp, second precision |
| UTCM | RFCTYPE_UTCMINUTE | TIMESTAMP | UTC timestamp, minute precision |
| (n/a — RFC parameter only) | RFCTYPE_DTDAY / DTWEEK / DTMONTH | INTEGER | Date durations (days / weeks / months) |
| (n/a — RFC parameter only) | RFCTYPE_TSECOND / TMINUTE | INTEGER | Time durations (seconds / minutes) |
| (n/a — RFC parameter only) | RFCTYPE_CDAY | INTEGER | Calendar day |
| (n/a — RFC parameter only) | RFCTYPE_STRUCTURE | STRUCT | Nested structure |
| (n/a — RFC parameter only) | RFCTYPE_TABLE | LIST | Table parameter (each row is a STRUCT) |
(any unrecognized RFCTYPE, with erpl_rfc_strict_type_check=false) |
— | VARCHAR | Lenient fallback (issue #53) |
Notes:
- The same mapping applies to
sap_read_table,sap_rfc_invoke, andsap_odp_*scanners because they shareRfcType::CreateDuckDbType(). BICS characteristic axes use the equivalentbicstype2rfctype → rfctype2logicaltypepath. - DDIC
CLNT,LANG,CUKY, andUNITare CHAR aliases — they surface as VARCHAR with a fixed width, not as a dedicated type. - BCD precision is capped at DuckDB's
DECIMAL(38)maximum; SAP fields wider thanDEC(20)will be reported withprecision=38and scale clamped to it. - Raw byte columns in
sap_read_table:RFC_READ_TABLEcannot carry binary data, so it spellsRAW/LRAW/RAWSTRING/RSTRcolumns out as hex text in its characterDATAline. ERPL decodes that back into the bytes it stands for, so the BLOB holds the payload andoctet_length()is the true byte count. An empty raw column arrives as the field delimiter and is reported asNULL. Before v2026.08.18 the hex text was stored verbatim, which doubledoctet_length()and required a manualunhex()(#109).
| Option | Type | Default | Description |
|---|---|---|---|
erpl_telemetry_enabled |
BOOLEAN | true |
Enable telemetry (erpl.io/telemetry) |
erpl_telemetry_key |
VARCHAR | (built-in) | Telemetry API key |
erpl_trace_enabled |
BOOLEAN | false |
Enable ERPL tracing |
erpl_trace_level |
VARCHAR | 'INFO' |
Trace level: TRACE, DEBUG, INFO, WARN, ERROR, NONE |
erpl_trace_output |
VARCHAR | 'console' |
Output: console, file, both |
erpl_trace_file_path |
VARCHAR | 'trace' |
Trace file directory |
erpl_trace_max_file_size |
BIGINT | 0 (unlimited) | Max trace file size in bytes |
erpl_trace_rotation |
BOOLEAN | false |
Enable trace file rotation |
erpl_rfc_strict_type_check |
BOOLEAN | false |
When true, throw an error on unsupported SAP RFC types instead of falling back to VARCHAR |
erpl_rfc_persistent_connections |
BOOLEAN | true |
Cache one RFC connection + function descriptor per column for a sap_read_table scan instead of reopening per batch |
erpl_rfc_max_persistent_connections |
UINTEGER | 16 | Upper bound on RFC connections a scan caches concurrently (issue #67); columns past the cap use per-batch open/close |
erpl_rfc_read_table_batch_budget |
UINTEGER | 1310720 | Target max concurrent result rows (projected columns × per-column batch) for sap_read_table; bounds peak memory on wide tables (issue #69). Lower = less memory but more RFC round-trips; 0 disables the cap |
erpl_rfc_fetch_size |
UINTEGER | 1310720 | How much sap_read_table asks SAP for per round-trip, as concurrent result rows. Alias of erpl_rfc_read_table_batch_budget. See Tuning large reads |
erpl_rfc_max_threads |
UINTEGER | 0 | Default for the threads named parameter; 0 lets erpl choose |
erpl_rfc_partitions |
UBIGINT | 0 | Split a sap_read_table scan into this many row ranges read in parallel. 0 reads in one pass, parallelising across columns instead. See Narrow tables |
erpl_rfc_partition_window_rows |
UBIGINT | 0 | Rows a partition worker claims at a time; 0 uses one RFC batch per window |
erpl_rfc_pushdown_filters |
BOOLEAN | true |
Translate SQL WHERE predicates into RFC_READ_TABLE's OPTIONS table so SAP filters the rows instead of sending them all. Turning it off never changes which rows come back, only how many cross the wire. See Filter Pushdown |
erpl_rfc_backend |
VARCHAR | 'nwrfc' |
Which implementation serves RFC calls: 'nwrfc' (SAP's NetWeaver RFC SDK) or 'proto' (the pure-Rust erpl-proto implementation). Must be set before the first SAP call; frozen for the life of the process once resolved. Environment override: ERPL_RFC_BACKEND |
erpl_rfc_backend_path |
VARCHAR | '' |
Explicit path to the RFC backend shared library, overriding the search. Empty means: next to the extension, then the loader's library path. Environment override: ERPL_RFC_BACKEND_PATH |
SET erpl_trace_enabled = TRUE;
SET erpl_trace_level = 'DEBUG';
SET erpl_trace_output = 'both';erpl does not link the SAP SDK. It resolves the RFC entry points at runtime, so the implementation behind them is a runtime choice:
-- Must come before any SAP call in this process.
SET erpl_rfc_backend = 'proto';
SELECT sap_rfc_backend(); -- 'proto'The backend freezes at the first SAP call. Connections and function descriptors belong to
one implementation, so a later SET is refused rather than allowed to hand a handle from
one library to the other — start a new process to switch.
If 'proto' is selected and its library cannot be found, the call fails. erpl never falls
back to the SDK, because a silent downgrade would leave you believing you had tested a
backend you had not.
| Option | Type | Default | Description |
|---|---|---|---|
erpl_bics_trace |
BOOLEAN | false |
Enable BICS trace logging |
erpl_bics_trace_dir |
VARCHAR | './trace' |
BICS trace directory |
erpl_bics_stream_result_tables |
BOOLEAN | true |
Stream the large BICS response tables off the SAP SDK handle instead of materialising them |
erpl_bics_max_result_memory |
VARCHAR | '' (DuckDB's memory_limit) |
Memory a single BICS result set may use, e.g. '8GB' |
erpl_bics_max_data_cells |
BIGINT | 0 (derive from the memory budget) |
Explicit cell budget (I_MAX_DATA_CELLS), for pinning exactly what BW receives |
erpl_bics_stream_result_tables controls how the response is read. With it on (the
default) the four tables that grow with the result — E_T_DATA_CELLS, E_T_ROWS,
E_T_MEMBER, E_T_MEMBER_PRESENTATION — are read row by row from the SDK handle and only
the fields actually needed are converted. Turning it off restores the pre-2026.08
behaviour of building the whole response as a value tree, which costs roughly 2.2 KB per
data cell and 0.8 KB per row-axis element. It exists as an escape hatch; there is no
reason to turn it off except to compare behaviour.
A BEx query is read in one piece: BW builds the whole result set or none of it, and there
is no windowing in the BICS RFC API to page it (neither BICS_PROV_GET_RESULT_SET nor
BICS_PROV_GET_RESULTSET_DETAIL accepts a row range). So the ceiling is a memory
budget.
erpl_bics_max_result_memory accepts anything DuckDB's own memory settings accept
('8GB', '512MB', '80%'). Left empty it follows DuckDB's memory_limit, which
defaults to 80% of physical RAM — so out of the box erpl sizes itself to the machine,
and SET memory_limit moves both together.
erpl converts the budget into the cell count BW needs (I_MAX_DATA_CELLS) by dividing by
~128 bytes per data cell — the width of one BICS_PROV_RS_DATA_CELL row in the SAP RFC
container. Set erpl_bics_max_data_cells to a non-zero value to bypass the conversion and
pin the cell count yourself (capped at 2,147,483,647, the width of the ABAP integer BW
receives it in).
When a query does not fit, it fails with the size it would have needed and a value you can paste straight back:
This BEx query is too large to read in one go: it would return 63000 rows x 31 columns
= 1953000 data cells, which needs about 238.4 MB of memory. The limit is 4.0 MB. Either
raise it with SET erpl_bics_max_result_memory = '287MiB' (make sure the machine actually
has that much), or make the query smaller with sap_bics_filter(). BW reported result
state 4.
Note the budget counts data cells — the key-figure cells — not output columns. A result with 60,000 rows and 31 output columns of which 6 are key figures is 360,000 data cells, not 1,860,000.
sap_read_table supports both SAP-side and DuckDB-side filter pushdown:
-- SAP-side filter (via FILTER parameter, pushed to RFC_READ_TABLE)
SELECT * FROM sap_read_table('SFLIGHT', FILTER='CARRID = ''LH''');
-- DuckDB-side filter pushdown (automatically pushed to SAP when possible)
SELECT * FROM sap_read_table('SFLIGHT') WHERE CARRID = 'LH';On a large table this is the single biggest lever available: a predicate SAP can evaluate turns a multi-million-row transfer into a few thousand rows. What reaches the server:
| Predicate | Pushed |
|---|---|
=, <>, <, >, <=, >= |
yes |
AND / OR over one column, including BETWEEN |
yes, all arms or none |
IN (...) |
yes, until the generated clause exceeds ~4000 characters |
IS NULL / IS NOT NULL |
no — ABAP has no NULL |
Predicates on the client field (MANDT, DDIC type CLNT) |
no — RFC_READ_TABLE rejects a clause naming the client |
Literals of type TIMESTAMP, BLOB, FLOAT, DOUBLE |
no — no unambiguous ABAP spelling |
Literals are rendered the way the DDIC expects them: DATE as YYYYMMDD, TIME
as HHMMSS, character and numeric values quoted with embedded apostrophes
doubled. A type whose ABAP spelling is not established is not pushed rather than
guessed at.
Anything not pushed is still applied — erpl evaluates it after reading, so the result set is identical either way. Only the volume transferred changes.
erpl_rfc_pushdown_filters = false disables the translation entirely and makes
erpl evaluate every predicate itself. It exists as an escape hatch for SAP
releases that reject the generated syntax, and as a way to check that a filter is
not the cause of a discrepancy: the same query must return the same rows with it
on and off.
Paging and parallelism use the same two names everywhere, so what you learn on one extension applies to the others:
| Named parameter | Session setting | Meaning | |
|---|---|---|---|
| Parallelism | threads |
erpl_<ext>_max_threads |
How many SAP calls run at once. 0 = let erpl decide |
| Fetch granule | fetch_size |
erpl_<ext>_fetch_size |
How much is asked for per round-trip, in the protocol's own unit |
-- per query
SELECT * FROM sap_read_table('LARGE_TABLE', threads = 8, fetch_size = 262144);
-- or as a session default
SET erpl_rfc_fetch_size = 262144;
SET erpl_rfc_max_threads = 8;The unit of fetch_size follows the protocol: for sap_read_table it is
concurrent result rows (projected columns x per-column batch), which is what
bounds the SAP SDK's own buffer on wide tables.
Both are transport settings only. Every value returns exactly the same rows;
they trade memory against round-trips, nothing else. Lower fetch_size to cap
memory harder on a wide table, raise it for fewer round-trips on a narrow one.
How much threads helps depends on the SAP system's capacity — work processes,
application servers, database sessions — not on erpl. Raise it while watching
throughput rather than setting it blindly, and check with your Basis team before
running a large parallel extract against a production system.
erpl_rfc_read_table_batch_budget is the original spelling of
erpl_rfc_fetch_size. Both remain supported and write the same value — they are two
names for one knob, not two knobs.
threads parallelises across columns — one concurrent RFC_READ_TABLE call per
projected column. That suits a wide extract and does nothing for a narrow one: a
single-column scan issues one call, and threads has nothing to spread.
partitions splits the rows instead. Each worker claims a window of the table and
reads it with ROWSKIPS/ROWCOUNT:
SELECT * FROM sap_read_table('DD02L', partitions = 8);
-- or as a session default
SET erpl_rfc_partitions = 8;What to expect. Measured on a 164,664-row single-column extract, release build:
partitions |
1 | 2 | 4 | 8 | 16 |
|---|---|---|---|---|---|
| wall | 2.23s | 1.45s | 1.11s | 0.83s | 0.92s |
About 2.7x at eight workers, with the knee clearly visible — past it, more workers cost more than they return. Where the knee falls is a property of the SAP system's capacity, not of erpl, so find it by raising the value and watching throughput.
Rows come back unordered. An unpartitioned scan calls RFC_READ_TABLE with
GET_SORTED='X' and returns rows in that order; partitioned workers finish in whatever
order they finish. That is why this is opt-in rather than the default. Add an
ORDER BY if you need one.
erpl_rfc_partition_window_rows sets how many rows a worker claims at a time. The
default is one RFC batch per window, which shares the table evenly across workers;
larger windows mean fewer, coarser hand-offs.
Two caveats worth knowing:
ROWSKIPSis a 32-bit integer, so a partitioned scan cannot start a window beyond row 2,147,483,647. erpl refuses rather than wrapping, because a wrap would silently re-read an earlier range and duplicate rows. Narrow the read with aWHEREclause.- Windows are offsets into a server-side sort, not a snapshot. If the table is written while you read it, rows can shift between windows. That is true of an unpartitioned scan too — it also pages by offset — but partitioning makes it easier to observe.
For a genuinely large extract, running several processes against SAP still scales better than any in-process approach, because SAP-side concurrency limits apply per client program. Check with your Basis team first.
Complete workflow for connecting through an SSH jump host:
-- 1. Create SSH tunnel secret
CREATE SECRET my_ssh (TYPE ssh_tunnel, ssh_host 'jump.example.com', ssh_user 'user', password 'pw');
-- 2. Create tunnel
PRAGMA tunnel_create(secret='my_ssh', remote_host='sap.internal', remote_port=3300, local_port=13300);
-- 3. Create SAP connection pointing to local tunnel port
CREATE SECRET my_sap (TYPE sap_rfc, ASHOST 'localhost', SYSNR '00', CLIENT '100', USER 'sap', PASSWD 'pw');
-- 4. Query through tunnel
SELECT * FROM sap_read_table('SFLIGHT');
-- 5. Cleanup
PRAGMA tunnel_close_all;Last Updated: 2026-02-11