Field noteNov 16, 20265 min read1 source

Reylo 0.7.0 adds DuckDB and replaces separate engine packages with a native engine

In one minute

Reylo 0.7.0 adds DuckDB as a fourth SQL dialect and replaces separate database-engine packages with one native engine installed through reylo. For teams using the TypeScript SQL tool, the release brings broader database support and more precise inferred types, but also requires attention to platform and production-install behavior.

What changed: Reylo 0.7.0 adds DuckDB support and replaces separate engine packages with a native engine installed alongside Reylo.

What happened

Reylo's GitHub release notes say version 0.7.0 makes the engine one native program, built for supported platforms, instead of a WebAssembly component packaged separately for each database. The reylo package declares eight platform packages as optional dependencies so a package manager can install the one matching the machine: macOS on Apple silicon or Intel; Linux on x64 or arm64 with glibc or musl; and Windows on x64 or arm64. The release instructs projects to remove @reylosql/postgres, @reylosql/mysql, and @reylosql/sqlite. It also warns that production installs need the matching engine, so teams must not install with --omit=optional where the preload runs. REYLO_ENGINE_PATH can point to a different Reylo binary. If the platform package is missing or the platform is unsupported, Reylo reports the supported platforms and that variable.

The release adds DuckDB as a fourth dialect. A duckdb: URL identifies the database file, and query .run() and .try() methods accept a DuckDBConnection. Type checking prepares each statement using the project's own @duckdb/node-api, checking parameter and column types. Reylo infers nullability and result cardinality using the statement and catalog information such as NOT NULL columns and keys. It opens the file read-only for analysis and closes it afterward; when the application holds a copy for writing, the analyzer reads a copy. DuckDB values have explicit TypeScript mappings in the release notes: BIGINT and HUGEINT become bigint, decimals become strings, and types such as STRUCT, MAP, and UNION receive corresponding object forms. Untyped DuckDB parameters, such as in select :x, produce an error asking for a cast.

Reylo also changes inference across existing dialects. The release describes keys carried through set operations and column lists, more accurate handling of nullable values in joins and grouping sets, and corrections to cardinality in cases such as set-returning functions. Postgres dates and timestamps outside years 0000–9999 now use ISO 8601 extended years; for example, 0002-06-01 BC becomes -000001-06-01. A potentially breaking parsing detail is that tags[:i] is interpreted as a slice, while tags[(:i)] expresses a parameterized subscript.

The engine change also affects performance and package size. On macOS, for first analysis of select 1 as x, three runs reported 64–69 ms for Postgres, 47–56 ms for MySQL, and 43–49 ms for SQLite with the native engine, versus 230–274 ms, 210–240 ms, and 184–234 ms respectively with the WebAssembly engine. A second Postgres analysis is reported at 1.2–1.7 ms, down from 6–8 ms. The release adds a JSON analysis API and schema endpoint to the Reylo playground. It was released November 3, according to GitHub.

What the archive adds

New Runtime covered earlier Reylo releases: version 0.3.0 on October 24, and version 0.6.0 on October 28. Version 0.7.0 is a new release in that ongoing product thread, with a new dialect and a substantial engine-packaging change rather than a restatement of the earlier capabilities.

Trend signal

This release supports agent-ready software exposes capabilities only weakly: Reylo makes SQL analysis available through a playground API, including query parameters, return type, cardinality, and diagnostics. The core release is a developer-tool upgrade, not evidence by itself of a broader shift toward agent-facing product interfaces.

Why it matters

For a team evaluating Reylo, DuckDB support may bring the same query-type inference workflow to local analytical databases, while the more precise nullability and cardinality analysis can prevent TypeScript code from assuming a row or non-null value when the database can return something else. The release's migration notes are operationally important: the new engine depends on an optional platform package, which can be dropped by an installation command that omits optional dependencies. A deployment that works locally can therefore fail if its production install or target architecture differs. Teams should treat 0.7.0 as a compatibility migration, not just a dependency bump: review SQL typing changes, test supported environments, and verify the install artifact in CI and production.

Limits and uncertainty

The timing figures are Reylo's own measurements: the notes specify macOS, three runs for first analysis, and a simple query, but do not establish performance across other platforms, larger queries, or independent test environments. The release does not quantify the effect of the type-inference changes on real projects or report adoption of DuckDB support. Although the notes describe eight supported platform combinations, they do not establish compatibility with every package manager, bundler, deployment image, or unusual runtime configuration. Teams should validate those conditions against their own build and deployment paths.

Reader takeaway

Reylo 0.7.0 broadens the database set and tightens query typing, but the upgrade's most immediate risk is installation behavior: the new native engine arrives as an optional platform package. Before adopting it, check your deployment target and ensure production installs retain optional dependencies; then run type checks against queries whose inferred nullability or cardinality may change.

Evidence / sources

  1. 1github.com/colinhacks/reylo/releases/tag/v0.7.0
All Field Notes

Next reads

Read next

Open the graph
  1. 01topicDeveloper ToolsExplore the developer-tools topic hub.
  2. 02archiveField NotesOpen the latest editorial analysis.
  3. 03sourcesSourcesSee the original sources this archive cites.
Who read this page?Machine requests, hidden until opened

Loading the privacy-safe route aggregate...

Open the JSON contract