What works

Everything below has golden-corpus coverage. See the Compatibility matrix for per-area status and Limitations for what is missing.

Drivers & sessions

python-oracledb (thin), the Oracle JDBC thin driver, and ODP.NET managed (Oracle.ManagedDataAccess.Core) each pass an end-to-end probe — connect, auth (12c PBKDF2 with mutual server proof), metadata, parametrised SELECT, INSERT with row counts, ROLLBACK, a 2 500-row fetch loop, and array-bind batch DML (executemany / JDBC batch / ODP.NET ArrayBindCount) — against both the 19c and 11g personas, against PostgreSQL and MariaDB. NUMBER(p,s) precision/scale is reported for python-oracledb thin and ojdbc and ODP.NET (real precision/scale; TIMESTAMP scale included).
The OCI thick client (Instant Client, via python-oracledb thick mode) runs the golden corpus too: 634 / 637, the 3 skips being cases Oracle itself rejects over that transport (ORA-01036 ×2, ORA-00925). It uses its own TTC dialect — little-endian fixed-width integers — which dbSaci emits from the negotiated TNS_CCAP_OCI1 capability. Not run in CI (needs a licensed Instant Client).
Scalar and array binds go across as PostgreSQL parameters. A cached statement re-run with new binds (REEXECUTE) is handled. Results stream batch by batch, including results past 1M rows or one 64 MiB packet.
SELECT ... FROM DUAL, ROWNUM (→ LIMIT), FETCH FIRST n ROWS ONLY, Oracle legacy (+) outer joins, CONNECT BY hierarchical queries (NOCYCLE, CONNECT_BY_ISCYCLE / _ISLEAF / _ROOT, SYS_CONNECT_BY_PATH), MERGE (incl. WHEN MATCHED ... DELETE), INSERT ALL / INSERT FIRST, PIVOT / UNPIVOT, analytic/window functions, recursive CTEs.
NVL / DECODE / NVL2 / SYSDATE / ADD_MONTHS / TO_CHAR / SUBSTR and the rest of the common Oracle scalar library, via orafce.
REGEXP_LIKE / REGEXP_REPLACE (incl. back-references) / REGEXP_SUBSTR (nth match, capture group) / REGEXP_INSTR / REGEXP_COUNT.
Sequences (seq.NEXTVAL / .CURRVAL), identity columns.
SYS_CONTEXT('USERENV', …), USER, UID.
Session time zone — ALTER SESSION SET TIME_ZONE (IANA region or fixed offset), SESSIONTIMEZONE / DBTIMEZONE; TIMESTAMPTZ values are converted to the session zone on the wire, DST-correct for named regions.

Data dictionary

Read-only catalog views the tooling and ORMs probe, returning UPPERCASE names like a real Oracle database: USER_* / ALL_* for tables, columns, objects, constraints, indexes, sequences, comments, triggers, users; V$VERSION; NLS_SESSION_PARAMETERS (reflects the session’s NLS_* settings).

An IDE schema browser — DataGrip / IntelliJ was the reference — introspects the tree end to end against these, including DBMS_METADATA.GET_DDL for the “copy DDL” action (rendered as PostgreSQL DDL). Objects appear under the connected user’s node: dbSaci gives each user a PostgreSQL schema of its own name (Oracle’s “schema == user”), and ALL_* / USER_* report honest owners against it.

Schemas

Each user owns a schema of its own name; unqualified names resolve there first, then in public (the shared fallback, also reachable as public.<name>). Cross-schema access is a qualified name plus the usual grants (SELECT * FROM hr.employees); ALTER SESSION SET CURRENT_SCHEMA redirects unqualified resolution.

Transactions

Explicit control, SAVEPOINT / ROLLBACK TO, Oracle-style implicit commit on DDL, SET TRANSACTION READ ONLY, SELECT ... FOR UPDATE (OF <cols> dropped, WAIT n → blocking, NOWAIT / SKIP LOCKED passthrough).

DDL

Oracle column types, CREATE VIEW / CREATE TABLE AS (the SELECT body is translated), many ALTER TABLE forms, COMMENT ON, CREATE SYNONYM → view, global/private temp tables → TEMP ... IF NOT EXISTS, ordinary and BITMAP (→ plain) indexes, function-based / expression indexes, basic materialized views with explicit refresh.

Triggers

CREATE [OR REPLACE] TRIGGER, BEFORE / AFTER / INSTEAD OF, row/statement level, multi-event (INSERT OR UPDATE), WHEN (...), :NEW / :OLD, REFERENCING NEW AS x, and bodies that do column assignment, an audit INSERT, RAISE_APPLICATION_ERROR, IF … END IF, or numeric FOR loops — lowered to a PostgreSQL trigger function.

PL/SQL (the smaller end)

Anonymous blocks and standalone function/procedure bodies translate on PostgreSQL — including %TYPE / %ROWTYPE, explicit named cursors, WHERE CURRENT OF, expression and statement CASE, WHILE / LOOP, EXECUTE IMMEDIATE, EXCEPTION WHEN … handlers (Oracle predefined names mapped), user exceptions via PRAGMA EXCEPTION_INIT, and SELECT … INTO (single-row INTO STRICT semantics). DBMS_OUTPUT.PUT_LINE → RAISE NOTICE. On MariaDB, anonymous blocks and simple routines run where Oracle mode accepts them; ROWID, WHERE CURRENT OF, autonomous transactions, and EXCEPTION_INIT are not emulated (no column-named-id or drop-the-pragma shims). Interval result columns stay text on MariaDB (no type 182/183).

Errors

~40 PostgreSQL SQLSTATEs map to real ORA- numbers — deadlock, serialization failure, unique / foreign-key / not-null / check violations, NOWAIT resource-busy, statement cancellation, connection loss, invalid number/date, single-row subquery returning many rows, and more. The PostgreSQL statement character position is carried into the TTC error_pos field (python-oracledb error.offset).

Operations

TCP keepalive, idle-session reaping, per-statement timeout → ORA-01013, OCIBreak / Ctrl-C cancels the in-flight statement, graceful drain on SIGINT / SIGTERM, and dependency-free /healthz + /readyz + /metrics (Prometheus) endpoints.