Snippets and Thoughts of One Oracle Developer

H Here is the situation I kept running into. I want an AI agent to help me with a real schema, look up a table, count some rows, sanity-check a column, find which view a column comes from. None of that is risky on paper, it is just SELECT. But the moment you hand a database connection to an agent, you are one bad prompt away from a disaster. TRUNCATE CASCADE, DROP PURGE, even simple UPDATE can do a lot of damage. So I built a small command into ADT.ai for exactly this case. It is called "discovery", and the whole point is that an AI (or any muggle) can poke around a schema without being able to break anything. What it does You give it a SELECT and an environment, it runs the query, prints the result, and writes a Markdown report. That is it. No DML, no DDL, no PL/SQL, no multi-statement smuggling. If you try anything other than a single SELECT, the validator rejects the statement before it ever reaches the database. adtai discovery -env DEV -sql "SELECT t...

I I have been shipping APEX Deployment Tool for a few years now. It is a Python CLI that extracts Oracle database objects, APEX applications, and data exports into files you can version in Git, then via clever patch files deploys them anywhere. Other people also use it, which is great. The problem is I am not a professional Python developer, and after a couple of years of adding features on top of features, the code shows it. So a two weeks ago I started a full rewrite. Not just a refactor, but a full rewrite, done almost entirely by AI, module by module. This post is about why, how, and what is still ahead. Why the rewrite The code quality problem was the most obvious one. The largest files had grown into multi-responsibility monsters that were hard to read and harder to change safely. Normalizers (the functions that clean up DDL before writing files) were woven into the main export script instead of living in their own testable place. Adding a new object type or tweakin...

I I read Tim Urban's "Your Life in Weeks" piece years ago, the one where he draws every week of a long life as a single grid of small squares. The image stuck with me for a reason I could not explain at the time. I rebuilt it as a tiny Claude Code skill that lives in my Obsidian vault and refreshes itself. Let me start with why the idea matters, then the skill, then how to build your own. The power of seeing your weeks A life of 90 years is about 4,680 weeks. Written as a number it means nothing; you read it, you nod, you move on. Drawn as a grid where each week is one square, it stops you cold. The whole thing fits on a single screen. That is the trick, and it is not a gimmick. The reason it works is the unit. A year is too big to feel and a day is too small to matter, but a week is the exact size of a human plan. You already think in weeks. You know what one feels like. So when you see all of them laid out and notice how few are left, the abstraction of ...

A APEX 26.1 dropped and everyone wants to spin it up locally. Good news, you don't have to build anything from scratch. The community have already done the heavy lifting. Here is whats out there. uc-local-apex-dev (United Codes) GitHub: https://github.com/United-Codes/uc-local-apex-dev This is my first recommendation for anyone who just wants to get going. Philipp Hartenfeller from United Codes maintains it actively, and it is kept up with every release since 23ai. Runs Oracle DB 26ai + APEX 26.1 + ORDS 26.1 out of the box. A single "./install.sh" handles everything, including waiting for the database to come up and configuring ORDS. Works with both Docker and Podman (it auto-detects your engine), and it is Mac/Linux friendly. ORDS Docker Compose (Anders Swanson, Oracle) GitHub: https://github.com/anders-swanson/oracle-database-code-samples/blob/main/ords-docker-compose/ Anders Swanson works on Oracle AI Database at Oracle, and he published ...

A APEX 26.1 shipped a small package called APEX_DB_DICTIONARY , and at first glance you might ignore it. We already have DBMS_METADATA for pulling object definitions, right? Well, these two solve different problems, and once I saw the difference I started reaching for the new one a lot. Here is the short version. DBMS_METADATA gives you DDL meant to recreate an object. APEX_DB_DICTIONARY gives you a description meant to be read, by a human or by an LLM. Noisy old way Say you want to hand the structure of EMP to your AI, so it can write you some queries. The reflex is GET_DDL. SELECT DBMS_METADATA.GET_DDL('TABLE', 'EMP') FROM dual; And what you get back is a wall of storage clauses, tablespace names, segment attributes and other junk you don't care about. To make it readable you have to do a lof of transformations first. BEGIN DBMS_METADATA.SET_TRANSFORM_PARAM(DBMS_METADATA.SESSION_TRANSFORM, 'STORAGE', FALSE); DBMS_METADATA...

W When you decide alone, your strongest critic is missing. Friends are busy or too polite, and the AI you ask just agrees with whatever you already lean toward, or the push back is very soft. So the brain settles into "this feels right" the moment you have half-committed to a direction. That is how I end up shipping things I should have killed and killing things I should have shipped. So I built a Claude Code skill that provides this honest feedback. It is called Devil's Advocates . You bring it a real decision and before it argues with you, it ask you few claryfing questions. One of them always forces you to say out loud what could go wrong, what you would lose if this call is the wrong one. Half the time that round alone is the value, because most bad decisions are just badly framed ones. A vague question gets you vague answer. Once the question is sharp, it dispatches five adversarial sub-agents in parallel. Parallel matters: each one answers without seeing th...

A APEX 26.1 ships tomorrow. After 2 years of massive team effort, we have Oracle APEX Special Update office hours scheduled for Thursday May 14th, 16:00 CEST, and the marquee feature is finally coming out of the oven. Six new icons! The feature is APEXlang , an open application specification language, and it's the biggest change to APEX since its introduction in 2004. APEXlang in one paragraph APEXlang is a new export/import format for APEX apps. Instead of a single ".sql" script with environment-specific numeric IDs, you get a zipped package of ".apx" files (one per page, one per shared component, one at app level) plus native ".sql", ".css", ".png", ".js" files living at proper paths. It is something like ".yaml" files, but on steroids and editable. Human readable, diffable, mergeable, Git-friendly. Same IDs across environments (static and non-numeric). The classic SQL script export isn't goin...

C Claude Code ships a "/insights" command that generates a usage report. It covers which sessions succeeded, where things broke, tool usage patterns, cost distribution. It is a HTML file and extracting anything useful from it is just enough friction to lose the habit. The big issue There's a bigger problem though. By default, "/insights" generates a report from the past 30 days. That's a huge window. I don't want feedback once per month, I want it once per week. At that point most of the recommendations there are for things you may have solved two or three weeks ago. The report feels stale before you even open it. I found a solution by manipulating the source data. Claude Code reads "~/.claude/projects/*/*.jsonl" to compute the report. The skill temporarily moves JSONL files older than 7 days out of that directory before running "/insights", then restores them immediately after regardless of success or failure. This constra...

W When Claude finishes a long task, you get silence. That's fine, but there's a better option. Claude Code supports hooks, which are commands that fire on specific events. So I wired up Mario sounds you might know from my APEX Deployment Tool , and now I get a little coin chime when Claude is done, a warning tone when it's waiting for me, and a dying-mushroom sound right before it compacts the conversation. Here's how to set it up. Install chime The chime Python library does all the work. One install, no system dependencies. pip install chime The script Save this anywhere stable. I keep mine at "~/.claude/scripts/play_sound.py". #!/usr/bin/env python3 import argparse import chime SOUNDS = { 'success' : chime.success, 'error' : chime.error, 'warning' : chime.warning, 'info' : chime.info, } def main(): parser = argparse.ArgumentParser() parser.add_argument('sound', ...

S So I just published two Claude Code plsql-formatter and sql-formatter skills. The first one handles packages, procedures, functions, and triggers. The second handles the SQL side: standalone queries, any SQL embedded inside PL/SQL and views. Together they cover everything that lands in a real project. I split them into two skills because that is how Claude routes, by intent. Two skills, two trigger sets, no token waste on rules that don't apply. The Problem SQL Developer's built-in formatter is configurable, but every developer has a slightly different preferences and not everything can be customized there as you want. The result on a shared repo is a diff full of whitespace noise, columns realigned three different ways, WHERE clauses re-indented on every commit. So you spend time in code review fixing formatting instead of logic. Eventually you settle with a style which you don't really like. A skill fixes this differently. Instead of running a tool that rew...

I I live in Obsidian . My whole second brain sits in a git-versioned vault, and I open it every single day – for notes, work tasks, personal tasks, blog drafts, projects, everything. So if I'm staring at the same editor for hours, it might as well look the way I want. That's how I ended up building my own theme: One Oracle Developer . Catchy name, huh? It's a fork of the excellent Baseline theme, with a few tweaks on top. Why fork Create a theme from scratch is extremely demanding and time consuming. I checked all themes in Obsidian, and 95% of them were total garbage. Baseline was almost perfect, but I had to make a few small tweaks here and there so it works better with lists, checklists, and tags. I also adjusted the headers and fonts. The biggest issue was the folders/files tree, which felt a bit dull. Install Baseline and my theme and compare them yourself. What's different The theme is opinionated but small. Here's what it changes: Serif h...

F Four years ago I wrote Sharpen the saw . Same message still holds, but the saw changed. If you are still sharpening only what you know, you are sharpening the wrong thing. So let's do the uncomfortable check again. When did you last build something with AI (and I am not talking about few ChatGPT prompts)? When did you last try a new model, a new tool, or a new workflow? How much time did you spend last week consuming AI content vs. actually using AI to change something? Snowball with AI, or get buried by it Your knowledge is the small snowball. AI is the snow around. Used right, you compound years of experience into weeks of output. Used wrong, you generate plausible looking nonsense and call it a day. The point is not to let AI do your job. The point is to let AI do the boring 80% so you can go deeper on the part that actually needs you. Talking to users. Delivering value. If you are a senior and you still hand-write boilerplate code in 2026, you are burning ...

C Claude is great at many things, but it does not know your tools. It knows Oracle, it knows APEX, it knows PL/SQL. But it does not know ADT. So when you ask it to create a patch, it will hallucinate the commands and you will waste time fixing them. That is where skills come in. That is why I created ADT skills for you. What are skills? Skills are basically text files (SKILL.md) which you put in your project's ".claude/skills/" folder. When you ask Claude to do something, it reads these files first and follows the instructions. Think of it as a cheat sheet for Claude. You write the rules once and Claude will follow them every time. You can write skills for anything. Coding standards, naming conventions, deployment procedures, formatting rules. If you can describe it, Claude can follow it. I created two skills for ADT so far: adt and adt-setup . They are available on my AI_SKILLS repo . You can copy them to your project and adjust them to your needs. The AD...

F A few years back, I created a view generator with support for JOINs. Unfortunately, it did not work for many use cases. It was bothering me for a while, so I decided to fix it. You list several tables with your preferred aliases, and it will find foreign keys between the tables (both ways) and apply them to the previous tables. So the table order is critical. I tried to adjust the table order based on the constraints, but that proved to be a bit challenging, so at least for now you have to figure out the table order yourself. Also, if you have columns with the same name, the later ones will be commented out. DECLARE in_tables CONSTANT VARCHAR2(4000) := ' TABLE_NAME1 t1, TABLE_NAME2 t2, TABLE_NAME3 t3, '; in_ignore_cols CONSTANT VARCHAR2(4000) := 'CREATED_BY,CREATED_AT,UPDATED_BY,UPDATED_AT,'; -- v_tables VAR...

A At times, I need to manage a long list of options; therefore, the logical approach is to organize them into groups rather than having everything in a single list. However, this approach introduces several challenges. Previously I just added another LOV item just with groups, but with multiple items you will quickly clutter your UI. The groups are sorted alphabetically. So if you want a different order, you have to use this neat trick, which add spaces at the start of your group name, and hence it gets sorted as you wish: SELECT NVL(g.group_name, INITCAP(g.group_code)) AS group_name, -- LPAD(' ', ROW_NUMBER() OVER ( -- reverse r# for proper grouping sets sorting hack ORDER BY g.order# DESC NULLS FIRST, g.group_code DESC ), ' ') || NVL(g.group_name, INITCAP(g.group_code)) AS group_name__, -- groups sorted, use in LOV as group -- ROW_NUMBER() OVER ( ORDER BY ...

A Any developer who has created a RESTful data service in APEX has faced the following issue: you create a service, and it does not work. It is a pain to debug REST services . This is typical of most APEX examples out there. You write your query or copy and paste your code here and there, and then move on. Low code, right? Never mind that it cannot perform proper code validation. Never mind that it cannot detect broken components in the future. As you might know from my previous blueprint articles or presentations, you should encapsulate everything in packages and views. That way, you can properly control your code and make sure your app continues to work. RESTful data services In short, REST services allow you to expose your data (basically SELECT statements) or logic (DML operations or more complex code) to other apps or systems. And it does not have to be a third-party system—you can consume your own services even from the same app. On the service side, you have to de...