Datoria SQL Compiler
One grammar definition generates parsers, renderers, and semantic analyzers for 15 SQL dialects. Lossless, fully typed, fast.
At a glance
| Dialects | 15: ANSI, PostgreSQL, BigQuery, Snowflake, DuckDB, Spark, Databricks, T-SQL, Oracle, DB2, MariaDB, Presto, Trino, Redshift, SQLite |
| Identity tests | 177,197+ roundtrip tests from 34+ real-world SQL sources, 99.7%+ pass rate |
| AST types | 5,988 distinct typed interfaces across all dialects |
| Parse speed | 44 microseconds/file |
| Semantic analysis | Multi-pass optimizer, column lineage, type inference, scope resolution |
| dbt support | Full Jinja compilation, tested against 59 public dbt projects (9,925 models) |
| Formatter | AST-aware, adaptive, configurable, dialect-specific defaults |
| Target | Generated Java (JVM); architecture supports Rust, TypeScript, Go targets |
What's different
One grammar, many dialects. A single declarative grammar generates a specialized parser for each dialect. No hand-written parsers, no runtime grammar interpretation. Add a dialect feature once and it works everywhere the grammar says it should.
Lossless by design. Every token is preserved: whitespace, comments, keyword spelling, parenthesization. Parse a file, render it back, get byte-identical output. That's what makes it safe for refactoring tools, formatters, and migration scripts that need to touch SQL without breaking it.
Full semantic stack. Parsing is the foundation. On top of it: scope resolution, column qualification, type inference, column lineage, query optimization, and SQL formatting — all working on the same typed, immutable AST.
Generated, not interpreted. Each dialect's parser compiles to specialized Java. The JIT can inline, devirtualize, and optimize it the way it would hand-written code. That's why it outperforms parsers written in Rust and C.
Try it
Each dialect has an interactive playground with live parsing, formatting, optimization, column lineage, and type inference. Pick a dialect and paste your SQL:
ANSI SQL | PostgreSQL | BigQuery | Snowflake | DuckDB | Spark | Databricks | T-SQL | Oracle | DB2 | MariaDB | Presto | Trino | Redshift | SQLite
Capabilities
| Capability | Description |
|---|---|
| Dialect Support | 15 SQL dialects from a single shared grammar definition |
| Typed AST & Roundtripping | 5,988 typed interfaces, immutable records, lossless roundtripping, semantic comparison |
| dbt & Jinja | Full Jinja template compilation with dbt project support |
| SQL Formatting | Adaptive, AST-aware formatter with configurable defaults |
| Column Lineage | Trace every output column back to base table sources |
| Type Inference | Nullable-aware typing for every expression |
| Query Optimization | Multi-pass rule-based optimizer with predicate pushdown and subquery elimination |
| Error Recovery | Parse broken SQL into partial ASTs with zero overhead on valid input |
| Performance | Fastest full-featured SQL parser benchmarked |
| Built for LLMs | Tight feedback loops for code agents and SQL generation |
| Why Datoria | How we compare to SQLGlot, Calcite, jOOQ, and SDF |
Strategic context
The SQL tooling landscape consolidated in 2025. Fivetran acquired SQLGlot (via Tobiko Data), dbt Labs acquired SDF Labs, and after the Fivetran-dbt merger one entity controls the dominant SQL infrastructure in the data ecosystem: SQLGlot, SDF/Fusion, and SQLMesh.
For anyone building SQL-powered products (data governance, migration tools, observability, IDE plugins), that's a strategic dependency on a competitor. Datoria is the only independent, production-grade SQL compiler covering this breadth of dialects with this depth of semantic analysis.
See Why Datoria for a feature-by-feature comparison with SQLGlot, Calcite, jOOQ, and SDF.
Evaluate or license
Datoria is a commercial SQL compiler available for technical evaluation and licensing. The interactive playgrounds on this site let you test parsing, formatting, lineage, and optimization against your own SQL across all 15 dialects.
Request a technical evaluation. Test the compiler against your production SQL corpus — your dialects, your queries, your edge cases. Free 30-day evaluations available.
OEM licensing. Embed the full stack in your product with redistribution rights. Per-dialect or all-dialect bundles. Category-exclusive terms available.
Contact us at hello@datoria.ai.