Skip to content

Reference ​

The reference is the Weft specification itself. Where the guide teaches, these pages define: what the compiler accepts, what it rejects, and what a program does when it runs.

  • The language: the syntax, types, values, functions, components, modules, the server and the limits.
  • The prelude: the types every file has without an import and their methods, the error types, and the modules every free function comes from (std/math, std/json, std/log, std/time…).
  • The app framework: the file tree of an app: pages, layouts, errors, components, modules and routes in code.
  • The server runtime: how a program serves: one Weft thread, I/O beside it, limits, the wire and server functions over HTTP.
  • std/http: the HTTP server (routes, middleware, sessions, CORS, static files) and client.
  • std/auth: sign-in: passwords, passkeys, email links and codes, OAuth and OpenID Connect, roles, guards and CSRF.
  • std/app: serving an app: pages rendered on the server, the client, endpoints, user trees.
  • std/db: the database: the schema in Weft, queries at three levels, policies, migrations, on SQLite, Postgres and MySQL.
  • std/live: live connections, channels and live queries: a server function call kept current in the page.
  • std/s3: buckets on any S3-compatible store: objects, listing, presigned URLs, streamed uploads and downloads.
  • std/time: instants, durations, dates, times and time zones, the clock and timers.
  • std/theme: the theme as Weft: colours, fonts, radii and shadows, evaluated into CSS variables when the project is built.
  • std/sandbox: user code: trees checked when saved, run with only what the app passes in.
  • Users' files: the API an app exports to user code, its compatibility, the export index, history, migrations and deprecations.
  • std/process: the arguments, environment, input, output and exit of the running program, and the programs it runs.
  • std/fs: files and directories for scripts and servers, and paths.
  • std/net: TCP and TLS sockets: clients, listeners and STARTTLS.
  • std/regex: regular expressions in a linear-time dialect, the same on every backend, regex"…" checked at compile time.
  • std/crypto: hashes, HMAC, constant-time comparison, random bytes, UUID v7, signatures and JWTs.
  • std/url: URLs as WHATWG parses them, query strings, percent-encoding and base64url.
  • The MCP server: weft mcp: the tools an AI agent works on a project with, from reading and editing to screenshots and tests.

How these pages are made ​

Each page is a file in the repository's spec/ directory (spec/language.md, spec/std.md, spec/app.md, spec/std/*.md), converted when the site is built. Nobody edits the pages here, so they always say what the spec says.

The specification uses a few words precisely:

  • error: the compiler rejects the program, with a diagnostic. An error value is what a function fails with: data of its error type (R ! E), which the caller handles or passes on with try. See Errors.
  • panics: the program hits a bug when it runs (an index out of range, ! on null). Ordinary code can't handle a panic: it ends the turn, and only @try, the server's 500 and assertPanics see it. See Panics.
  • stop: the program hits a limit (fuel, memory, stack, the deadline). Nothing handles a stop. See Limits.

Some parts of the specification refer to files that are not on this site, such as expressions.md or docs/plan.md. They live in the repository beside the spec.

Weft is in development. The reference is the spec as it stands today.