Calling #OracleACE: we want your assertion use cases! We’re opening a bounty for creative, real-world solutions using assertions Submit your blog post, video, script or slide deck the contribution app by the end of 15 March Full details at buff.ly/fNgDHQv

Think you know about #SQL performance? Take this 3-minute indexing quiz by @winand.at A simple test that 60% fail! buff.ly/5C4XZ7N Did you pass? Let us know how you did!

Struggling to remember which #SQL statements you just ran? Set sql_history_enabled = true in Oracle AI Database 26ai and it records the last 50 statements in each session Query v$sql_history to find them @brendan-tierney.bsky.social demos

Understanding a complex data model is challenging @connormcd.bsky.social explains how you can make this easier in Oracle AI Database with Constraints to improve data quality Usecase domains to describe values Schema annotations to add metadata

MERGE in Oracle AI Database 26ai supports the RETURNING clause: MERGE INTO ... USING ... ON ( ... ) WHEN ... RETURNING OLD ..., NEW ... INTO ..., ... This enables you to get pre (OLD) and post (NEW) merge values from the target @andrejsql.bsky.social demos

Limit the rows processed by aggregate and window functions with the FILTER clause e.g. COUNT (*) FILTER ( WHERE ... ) The COUNT only includes rows WHERE the condition is true Each function in a query can have a different FILTER Available in Oracle AI Database from 23.26.1

ICYMI Oracle AI Database 26ai Enterprise Edition, 23.26.1.0.0, is available for on-premises download on Linux x86-64 platforms @mikedietrichde.com shows you where to get it

Oracle AI Database 26ai added operators for sounds like/similar to logic: Sounds like => PHONIC_ENCODE ( <algorithm>, <str> ) Similar to => FUZZY_MATCH ( <algorithm>, <str1>, <str2> ) These help you find related strings @oraclebase.bsky.social demos

The QUALIFY clause enables you to filter rows using window functions e.g. get the top ranked rows with SELECT ... FROM ... QUALIFY RANK () OVER ( ORDER BY ... ) = ... @danischnider.bsky.social looks at how it works in Oracle AI Database 26ai

AI vector search enables you to find rows based on meaning e.g. “Toys for big dogs” can match “Heavy Duty Rope for Large Dog Breeds” @alina-yurenko.bsky.social shows how to build this with GraalVM, Spring Boot, and Oracle AI Database

Create cross-row, cross-table constraints in Oracle AI Database (23.26.1) with CREATE ASSERTION ... CHECK ( <rule> ) Define the rule as either an Existential expr [ NOT ] EXISTS ( <query> ) Universal expr ALL ( <query> ) alias SATISFY ( <condition> )

It's amazing what you can do with #SQL Patrick has built a SQL-powered Doom clone using DuckDB

Parallelizing a task can make it faster; this is different to simultaneous: Simultaneous – Occurring, operating, or done at the same time. Parallel – Dividing a task into smaller sub-tasks that can be executed concurrently Ric Van Dyke discusses

The view GV$SYSMETRIC_HISTORY in Oracle AI Database shows all the database metrics available This can be useful to help troubleshoot performance issues @tanelpoder.com has built scripts to display these with 1-minute granularity in a tabular way

The ACID properties of database transactions help keep data consistent But these aren't enough to stop all data errors! Failures may cause users to keep retrying and duplicating work Arkadiusz Chmura shows how to avoid this with effectively once processing

Caching can improve performance by bringing data closer to the client @anders-swanson.bsky.social gives an overview of the caches in Oracle AI Database, including Oracle True Cache Oracle Active Data Guard Database In-Memory Buffer Cache Server-Side Result Cache

Preparing to upgrade to Oracle AI Database 26ai? @dohdatabase.com covers important to add to your post-upgrade checklist Backup Gather Fixed Objects Statistics Underscore Parameters Enable Optimizer Fixes Data Guard Config Full details at

VARCHAR2 in Oracle AI Database can store up to 32k of data, provided MAX_STRING_SIZE = EXTENDED @gvenzl.bsky.social shows that extended VARCHAR2s are stored as BLOBs, storing bytes exactly as-is like regular VARCHAR2s CLOBs use 2-byte fixed-width characters

The default fetch size in the Oracle JDBC driver is 10 rows This is too small when reading large data sets => looooong query processing times To avoid this, the 26ai driver can automatically increase the fetch size when getting many rows @ppiipuu.bsky.social demos

Want to improve the quality of your Oracle #SQL and PL/SQL? @salvis.com has built the automatic rule checking suite, dbLinter This defines 187 rules you can apply to check your code Philipp has picked out 14 to always run Do you agree with his list? buff.ly/Nc5t1Qd

Got #SQL which works in an anonymous block ...but when you place it in a PL/SQL stored procedure you're hit with ORA-01031: insufficient privileges ?!? It's because you need direct privileges on the objects, not via a role @cafedatabase.com explains

Limit the rows returned in ISO standard #SQL with FETCH FIRST n ROWS ... ONLY => At most n WITH TIES => Up to n + all rows with the same sort value as the Nth Get a fraction of the total rows with ... n PERCENT ROWS ... Without ORDER BY, the rows returned are undefined

Sort the output of #SQL queries with ORDER BY exp1, exp2, ... This sorts in ascending order: low to high, A-Z For high to low, use ORDER BY exp DESC Rows with ties for exp1 are then sorted by exp2, etc. The order is undefined for rows with the same value for all exps

Filter the rows returned by a #SQL query with WHERE <condition> <condition> can be any valid SQL condition, e.g. number = 1 letter like 'A%' You can combine conditions with AND/OR; the database returns rows where AND - all conditions are TRUE OR - any condition is TRUE

De-duplicate rows in #SQL query results with SELECT DISTINCT col1, col2, ... This returns one row for each unique combination of values stored in the DISTINCT columns This should be a rare operation! Repeated use of DISTINCT is often a sign of flaws in the query or schema

The SELECT clause in a #SQL query defines which values it returns SELECT { * | v1, v2, v3, ... } The values can be Any column from the source tables Literal values Expressions Asterisk is the all-column wildcard; it returns every (visible) column in the tables listed in FROM

You can give AI agents access to Oracle AI Database via the SQLcl MCP Server @thatjeffsmith.com uses this to generate a report of viewing habits in his Netflix data using Opus 4.1 buff.ly/89106db Remember: keep an eye on your agents & check their SQL

The method_opt parameter of dbms_stats controls histogram generation @martindba.bsky.social shows how to use it to Delete all histograms except for 1 Create this histogram using the AUTO clause With for all columns size 1, for columns data_skew size AUTO

Want to find the value at the nth percentile with #SQL? Use PERCENTILE_CONT ( n ) WITHIN GROUP( ORDER BY close_price ) n is between 0 and 1, so 0.5 => median 0.75 => 75th percentile 0.99 => 99th percentile @vladmihalcea.bsky.social demos

The Expand/Contract Method is a tool for safe database changes Expand: add new schema elements Migrate: move code to these Contract: remove old elements @jasminfluri.bsky.social discusses using it to Rename a column Split a table Change the data type of a column

Learning #SQL for your 2026 New Year's Resolution? Head to Oracle Dev Gym This has FREE classes to take you from Your first SELECT & CREATE TABLE statements On to recursive WITH and window function queries And learning how to make these fast Join at buff.ly/7gWfOdq

We're taking a break from the #SQL tips over the festive period UPDATE sql_tips SET next_tip_on = DATE'2026-01-05' We hope you have a wonderful Christmastime and a fantastic New Year Here's a SQL snowflake generator to celebrate buff.ly/yjbOwiK See you in 2026!

Using bind variables is critical for secure #SQL But they can also give large performance benefits @martindba.bsky.social compares using binds vs literals in a simple Java app The result? Binds are 4.5x faster Use nearly 1000x less database memory

Oracle AI Database uses the Object-Relational Model to store spatial data @anders-swanson.bsky.social shows how to load spatial data and use it to Calculate the area of a geometry Find the distance between two points Find geometries close to a given shape

Want a visual guide to the structure of Oracle AI Database's Processes Memory Objects Files Check out the Technical Architecture diagrams in the docs Click through the interactive diagrams to see how the components relate Or view them all at buff.ly/Mo3goIy

Find sequences of events across rows with #SQL pattern matching e.g., W patterns where prices fall then rise twice With the MATCH_RECOGNIZE clause: MEASURES - values to return PATTERN - the sequence as a regex over variables DEFINE - the variables using SQL conditions

SQL Case Files Learn #SQL while tackling detective investigations Write queries to learn about suspects, uncover evidence, and close the cases Play at buff.ly/ANG0hA4

Schema annotations in Oracle AI Database enable you to add key-value metadata on Tables Views Indexes Domains These help document your objects @martindba.bsky.social shows how to add them to view columns with ALTER VIEW ... MODIFY ( <col> ANNOTATIONS ( ... ) )

Real-time SQL Plan Management in Oracle AI Database can Detect plan changes in #SQL queries Identify which of these plans is best Ensure future runs use the best plan Prashant Dixit shows it in action

You can run code in the BEGIN/END block of a PL/SQL package CREATE PACKAGE BODY ... AS ... BEGIN do_stuff(); END; Take care: this only runs when initializing the package @petefinnigan.bsky.social shows how security checks here can fail

Access attributes of JSON data stored in Oracle AI Database Simple dot-notation - <alias>.<column>.<path>.<to><attr> JSON_VALUE ( column, '$.<path>.<to><attr>' ) Dot-notation returns NULL if it can't process the attribute, JSON_VALUE gives full control over the behaviour #SQL

Aggregate functions in #SQL report totals across many rows Common ones are COUNT, SUM, AVG, MAX, MIN Baraa Khatib Salkini explains how these work and how to use them to answer questions like How many orders are there? What's their total sales value?

AI agents can help you build schemas and create test data @chrishoina.bsky.social shares an example prompt to store data about cars, trucks, motorcycles and their manufacturers Then shows you how to use it with SQLcl MCP Server and Cline to make the tables

The wait is (nearly) over! Oracle AI Database 26ai EE for Linux x86-64 will be released on premises in the next RU (23.26.1), due January 2026 Read full details at blogs.oracle.com/database/ora...

Finally … Oracle AI Database 26ai is going to be available on non-Oracle hardware Many of you have read my blog post from October during Oracle AI World in Las Vegas about Oracle AI Database 26ai replacing Oracle Database 23ai. But the main question remained: When is it available on-prem on…

Mouhamadou Diaw examines the differences between functions & procedures in PL/SQL Functions Return a value Can be called in #SQL - DML must be in an autonomous transaction Procedures Use OUT parameters to return values Can't be called in SQL directly

Tracking changes to values over time is a common requirement For example, recording price fluctuations for goods Alexey Makhotkin steps through designing tables to store these changes, proposing template tables with Anchor Value Valid from/to dates

Change the partitioning method for an Oracle table with: ALTER TABLE ... MODIFY PARTITION BY ... This can change The partition method (RANGE/LIST/HASH) The partition columns Add/remove subpartitioning Add the ONLINE clause to allow writes to continue while changing

Like #SQL? Like puzzles? Try #SQuizL A free, daily guess-the-SQL statement quiz You have up to six attempts to complete the statement on the top left using the fragments below There's a new statement to guess at every day at midnight UTC Play now at buff.ly/Myg8265

You can run in-database AI inference on Oracle Active Data Guard @ludovico.bsky.social shows you how to do this to find images with semantic search Loading an ONNX model & pictures Creating PL/SQL to generate embeddings Running this proc and searches on the standby