Oracle SQL

Let a single query build a temporary model from its own rows and score them, with the analytic form of PREDICTION.

Train a one-class support vector machine on normal data and rank rows by how likely they are to be anomalies.

Compress several attributes into a few features with a feature extraction model, and see which feature describes each row best.

Predict a number with a generalized linear model and get a 95 percent range around each prediction, not just one value.

Tell a classification model that some mistakes cost more than others, and let predictions choose the cheapest outcome.

See how likely every class is for each row, not only the top prediction, and use the probabilities in your decisions.

Build a classification model in the database, score rows in SQL, and explain each prediction with its most important attributes.

Speed up compute-heavy PL/SQL with native compilation, and hide the source of procedures and package bodies by wrapping them.

Find code that compiles but is probably wrong, such as unreachable lines, with PLSQL_WARNINGS, and make the worst ones errors.

Keep debug code and version-specific code in one PL/SQL source, and choose what is compiled with PLSQL_CCFLAGS and DBMS_DB_VERSION.

Build parameterized views and reusable query shapes with functions marked SQL_MACRO(TABLE), which Oracle expands into the calling query.

Record who connects with AFTER LOGON triggers and keep a history of schema changes with AFTER DDL triggers, and manage your triggers.

Let many transactions update seats, stock, or balances on the same row at once with reservable columns, while constraints still hold.

See how a deadlock happens and how Oracle breaks it with ORA-00060, where it is recorded, and how to keep deadlocks from recurring.

Lock the rows you read until you update them with SELECT FOR UPDATE, and use NOWAIT, WAIT, and SKIP LOCKED to control waiting.

Choose between READ COMMITTED, SERIALIZABLE, and READ ONLY transactions in Oracle, and know what each query sees and when ORA-08177 appears.

Reuse an expression across queries with a function marked SQL_MACRO(SCALAR), which Oracle expands into plain SQL when the query is parsed.

Create a subtype that inherits from a supertype, override its methods, and test and narrow values with IS OF and TREAT.