Why We Built SQL Tran: Automated Oracle to PostgreSQL Migration in Seconds - Spectral Core Blog
Back<br>News Why We Built SQL Tran: Automated Oracle to PostgreSQL Migration in Seconds<br>Damir Bulic<br>05 Jun 2026<br>- 10 min read
SQL Tran translates Oracle to PostgreSQL automatically at 95 to 99 percent accuracy in seconds, with built in testing. See the live launch demo.
Automated Oracle to PostgreSQL Migration | SQL Tran
SQL Tran is live.<br>We ran a live, end to end demo: a real Oracle database translated to PostgreSQL, the target database created, and thousands of tests generated and run, all in seconds. If you missed it, the full recording is on our YouTube channel and the link is at the bottom of this post.<br>Here is what SQL Tran does, what it took to build, and why we think the industry standard for automated SQL translation needs to change.<br>The problem with Oracle to PostgreSQL migration<br>Industry standard for automated SQL translation success is 50 to 70 percent, whatever vendors say.<br>The remaining 30 to 50 percent is dropped on a senior engineer who now has to read through translated code looking for the parts the tool got wrong.<br>That is not translation. That is shifting the work.<br>There is a second problem nobody likes to talk about. On a serious codebase, say half a million lines of database code, the legacy tools need a day or two just to run the assessment, then more time for the translation, before anyone has even started thinking about testing. Slow iteration means weak verification, and weak verification means surprises in production.<br>What SQL Tran does<br>SQL Tran is a high speed, fully automated SQL code translator with a built in static analysis engine and automated test generation.<br>In the launch demo, we translated an anonymized customer schema of 85,000 lines of Oracle code, with roughly 2,000 objects including 1,200 tables, 96 procedures, and 500 triggers, to nearly 100 percent in seconds. Then we did it again with a 500,000 line database. Same result, same speed.<br>On our largest tested codebase, SQL Tran translates 2.7 million lines of Oracle code to PostgreSQL in about 20 seconds. It typically lands at 95 to 99 percent translation accuracy on real codebases, and it is up to 1,000 times faster than competing products.<br>Currently supported translations: Oracle to PostgreSQL, SQL Server to Microsoft Fabric Warehouse, and Synapse Analytics to Fabric Warehouse.<br>It includes the parts most database migration tools leave to you:<br>Full semantic analysis with data lineage and impact analysis, so you know what every identifier refers to and what depends on it<br>Automatic creation of the live target PostgreSQL database, straight from your Oracle schema dump<br>Business logic correctness testing that compares source and target behavior<br>Synthetic test data generation that conforms to your schema, respecting data types, check constraints, uniqueness, and foreign keys<br>Hundreds of built in emulations for cases where naive syntax translation would change behavior<br>What we showed in the live demo<br>A few things from the launch that are worth calling out, because they are what set SQL Tran apart from a syntax converter.
Idiomatic, lock free PostgreSQL. We work hard to generate clean, idiomatic PostgreSQL and to avoid external extensions wherever possible. Roughly 99 percent of the time we translate everything out of the box, so your result runs on open source PostgreSQL. No lock in to EDB, no lock in to a specific cloud provider's extensions. That matters because the whole point of leaving Oracle is to stop being locked in.<br>Hard things, handled. Oracle packages do not exist in PostgreSQL, so we emulate them. CONNECT BY, LISTAGG, complex aggregations, recursive logic, nested procedures: all emulated faithfully. In the demo we navigated the data lineage of a single table and saw every object that reads from it and every object that writes to it, without needing a live database to figure it out.<br>From script to tested database in seconds. We started from nothing but an Oracle schema dump, created a live PostgreSQL database, generated synthetic data conforming to the schema, then generated and ran roughly 1,700 tests across tables, views, functions, procedures, and packages. The result: 99.9 percent of the code behaved identically in source Oracle and target PostgreSQL. That is the number that should matter to anyone scoping a migration.<br>How it works<br>Short version: we did not build a syntax substitution tool. We built a full SQL compiler.<br>SQL Tran parses your codebase into a complete semantic model. Every identifier is bound to its definition. Every reference is tracked. As far as we know, we are the only vendor doing full static analysis across an entire database. We know what your code means before we translate it, and we translate to preserve meaning, not surface syntax. Static analysis also catches broken references in your existing code, so SQL Tran finds bugs you have been carrying for years.<br>This is also why we are fast....