PLSQL

Avoid opening an explicit cursor twice by checking %ISOPEN, closing cursors after errors, or using cursor FOR loops.

Read an Oracle error stack from the top, follow ORA-06512 to the code line that raised the error, and keep the location when handling errors.

Make SELECT INTO return exactly one row with the full key or a stated rule, or read all rows with BULK COLLECT.

Understand why SELECT INTO raises NO_DATA_FOUND, handle it where a missing row is normal, and watch for it in functions called from SQL.

Read the query string, headers, and other request data in PL/SQL web procedures, and test them without a web server.

Write HTML pages from PL/SQL procedures with HTP and HTF, and test them in a SQL session by printing the page buffer.

Write one table function that works on any table, with output columns decided when the query is parsed, using DBMS_TF.

Find out at run time what type a value has, its name, and the name and type of each of its attributes.

Keep numbers, strings, dates, and objects in one column with ANYDATA, and read each value back according to its type.

Find out how full a table's blocks are, and estimate how much space a planned table of millions of rows will need.

Record which blocks of PL/SQL code your tests actually run, see the percentage covered, and find the untested branches.

Record every PL/SQL call and SQL statement while code runs, and find the functions and statements where the time goes.

See cached query results being reused, and check, flush, or invalidate the server result cache with DBMS_RESULT_CACHE.

Measure the work of one end user, module, or session with targeted statistics and SQL tracing, instead of the whole database.

Change how one table's statistics are gathered with preferences, and freeze them with a lock to keep plans stable.

Keep the optimizer's picture of your data current by gathering table, column, and histogram statistics after large changes.

See the plan a statement actually ran with, and compare the optimizer's estimated rows with the actual rows of each step.

Recompute the hash chain of a blockchain table to prove no row was changed, and delete rows once their retention ends.