Languages

pgzest analyzes SQL once, whatever language calls it. Each query becomes a descriptor — its parameters, its columns with their types and nullability, the number of rows it returns, the objects it depends on — and a generator per language turns descriptors into typed code.

What happens on the SQL side is shared: the analysis, the migrations, the editor on .sql files, the instrumented routines. What is per language is the generated code and its runtime: the functions and types, SQL written inside the language’s own strings, the collector that turns routine runs into OpenTelemetry spans.

Language	Status	Driver	What you get
Rust	Available	tokio-postgres	Named queries, sql! and query! compiled at build time; parameter and row structs, enums, COPY; routine traces as OpenTelemetry spans
TypeScript	Next	pg or postgres (planned)	Named queries to .d.ts with a thin runtime first, then queries written in TypeScript files
.NET	Considered	Npgsql	Not started

Any other language: the descriptor as JSON

Until a language has its generator, pgzest describe prints the descriptors of a project’s named queries as JSON: everything a generator needs, checked against the project’s schema. See the CLI reference.

pgzest describe queries/restaurants.sql