OraMatt: YABAOracle

One of the persistent myths in data engineering is that SQL is portable.

While SQL is based on ANSI standards, every database engine implements its own dialect. These differences appear in:

• pagination syntax
• quoting rules
• data types
• functions
• identifier handling
• DDL semantics

These incompatibilities become immediately obvious during:

• database migrations
• cross-database development
• platform modernization projects

To help streamline these workflows, I built MADSQL, a deterministic command-line tool that converts SQL scripts between dialects using SQLGlot.

GitHub Repository:
https://github.com/oramatt/madsql

Why SQL Dialect Translation Matters

Consider something as simple as limiting query results.

SQL Server

SELECT TOP 3 [name] FROM dbo.users;

PostgreSQL

SELECT name FROM users LIMIT 3;

Oracle

SELECT name FROM users FETCH FIRST 3 ROWS ONLY;

These statements all accomplish the same goal, but they are not interchangeable across database platforms.

Now imagine converting an application containing thousands of SQL statements between databases. This is where dialect incompatibilities quickly become a major engineering challenge.

MADSQL was created to help automate part of that work.

What MADSQL Is

MADSQL is a deterministic command-line tool for translating SQL scripts between database dialects using the SQLGlot parser and transpiler.

It supports workflows such as:

• converting SQL between dialects
• splitting large SQL files into individual statements
• processing entire directory trees of SQL scripts
• generating deterministic output suitable for CI pipelines
• producing structured error reports and run logs

MADSQL is designed specifically for batch SQL transformation workflows.

Core Design Goals

Deterministic Output

One of the key guarantees of MADSQL is deterministic output.

Given the same:

• input SQL
• SQLGlot version
• CLI arguments

MADSQL will generate identical output every time.

This makes the tool ideal for:

• automated migration pipelines
• CI/CD workflows
• reproducible SQL transformations

CLI-First Design

MADSQL is intentionally designed as a terminal-based tool.

There is no GUI and there are currently no plans to add one. The CLI design makes it easy to integrate into:

• automation scripts
• migration frameworks
• developer workflows
• batch SQL processing pipelines

Batch SQL Processing

MADSQL can process SQL from multiple input sources:

• single SQL files
• directory trees
• glob patterns
• stdin streams

Example:

madsql convert --source tsql --target postgres "./sql/**/*.sql" --out ./converted

This allows entire repositories of SQL code to be processed in a single run.

Example Usage

Convert a Single Statement

echo "SELECT TOP 3 [name] FROM dbo.users;" | madsql convert --source tsql --target postgres

Convert a SQL File

madsql convert --source postgres --target mysql ./input.sql

Convert a Directory of SQL Scripts

madsql convert --source postgres --target mysql --in ./sql --out ./converted

When converting directories, MADSQL preserves the relative directory structure in the output folder.

Splitting SQL Statements

Another useful feature is the ability to split SQL scripts into individual statements.

madsql split-statements --in ./input.sql --out ./split

This produces deterministic files such as:

0001_stmt.sql
0002_stmt.sql
0003_stmt.sql

MADSQL attempts to detect statement boundaries using SQLGlot first. If parsing fails, it automatically falls back to sqlparse for split-only boundary detection.

Deterministic File Naming

When splitting or converting SQL, MADSQL uses deterministic naming.

Example output:

0001_stmt.oracle.sql
0002_stmt.oracle.sql

If a statement fails conversion, later statements retain their original index numbers.

Example:

0001_stmt.oracle.sql
0003_stmt.oracle.sql

This makes it easy to correlate outputs with the original SQL source.

Error Handling and Reporting

MADSQL provides structured error handling designed for automation workflows.

Write JSON error reports:

--errors errors.json

Stop immediately on failure:

--fail-fast

Exit codes are predictable:

This makes MADSQL safe to integrate into automated pipelines.

Markdown Reports

MADSQL can also generate logs and structured reports.

Markdown reports:

--report

Reports include:

• command invocation
• source and target dialects
• success and failure counts
• overall success rate

Reports are timestamped and written to the output directory.

Example Migration Workflow

MADSQL fits naturally into database migration pipelines.

Application SQL
      │
      ▼
MADSQL Convert
      │
      ▼
Target Dialect SQL
      │
      ▼
Database Validation

Because the tool is deterministic and script-friendly, it works well as part of larger migration automation systems.

Why I Built MADSQL

There are already SQL translators available.

MADSQL exists because I wanted a tool that supported the specific workflows I use when working on migration projects.

In particular:

• deterministic outputs
• reproducible CLI workflows
• batch processing
• structured error reporting
• reliable statement splitting

In many migration projects, engineers spend significant time rewriting SQL manually. MADSQL helps automate parts of that process.

Project Repository

https://github.com/oramatt/madsql

The project is licensed under the Universal Permissive License (UPL).

Final Thoughts

SQL dialect differences are one of the most underestimated challenges in database migrations.

Even small differences between database engines can require large-scale SQL refactoring across applications.

MADSQL provides a practical CLI tool to help automate SQL dialect translation and reduce manual effort in cross-database development and modernization projects.