rakus — a static HTTP file server

Source on GitHub ↗

Point it at a directory and it serves the files inside over HTTP/1.1 — the Raku answer to python3 -m http.server, on nothing but IO::Socket::INET.

This is the showcase copy: one file, runnable straight from a checkout. The installable distribution lives in raku-modules/App-Rakus as App::Rakus, where the routing is split into a library so it can be tested without a socket. It is published to the zef ecosystem (raku.land), so zef install App::Rakus puts a rakus command on PATH — one that serves the current directory by default, where this copy serves its bundled public/. Edits here do not reach there, or the other way about. Where the

pastebin is a single-purpose app with a hand-wired route table, rakus is a reusable server: give it a folder and it figures out the rest.

Run it

build/rakupp showcase/rakus/rakus.raku                  # serves ./public on :8080
build/rakupp showcase/rakus/rakus.raku 9000             # choose the port
build/rakupp showcase/rakus/rakus.raku 9000 ~/site      # choose port and root
# or compile a standalone binary:
build/rakupp --exe -o rakus showcase/rakus/rakus.raku && ./rakus 8080 ~/site

Then open http://127.0.0.1:8080/. With no root given it serves the bundled public/ folder — a landing page, a stylesheet, an SVG logo, and a files/ directory with no index (so you can see the auto listing).

What it does

Correct Content-Type by extension — text and binary. HTML, CSS, JS, JSON, SVG, PNG, JPEG, … (files are read as raw bytes and streamed back, so images arrive intact).
index.html when a directory has one; otherwise an auto directory listing with sizes and links.
GET and HEAD; other methods get 405.
301 to add a missing trailing slash on a directory (so relative links resolve), 403 on .. path traversal, 404 for anything missing.
Concurrent — one start thread per connection.
Logs each request to stderr: 200 GET /style.css.

From the command line

$ curl -sI http://127.0.0.1:8080/logo.svg
HTTP/1.1 200 OK
Content-Type: image/svg+xml
Content-Length: 289
Server: rakus
Connection: close

$ curl -s http://127.0.0.1:8080/files/ | grep -o 'Index of[^<]*'
Index of /files/

$ curl -so /dev/null -w '%{http_code}\n' http://127.0.0.1:8080/nope
404

How it works

No HTTP library. Each connection is read with recv until the header terminator; the request line is split into method and target by hand.
Path handling. The target is URL-decoded and the query stripped; a .. segment is rejected; the rest is joined onto the (absolute) document root.
Serving. A directory resolves to index.html or a generated listing; a file is slurp-ed as a Buf and its bytes written straight to the socket with a Content-Length. The headers go out with print, the body with write — so binary files are byte-exact.
Byte-exact bodies. Every response body is a Buf (generated HTML is .encode-d), so Content-Length is always the true byte count.