gitoriaLog in with ident

tracker

All repositories: gitoria

ReadmeCodePull requestsReleasesTicketsSettings
Commitf83571c3f83571c3tracker#18 (mission 067): daily sync by change lists — TMDB /tv|movie/changes (since the stored day, paged) + TVmaze /updates/shows → only our changed titles (followed: full step, unfollowed: light step — changed seasons, no TVmaze), full walk on first run / gap > 14 days / failed list; show record refreshed (title, tmdbSummary, tagline, status, genres …; renamed titles re-indexed); summary = the creator's own text (page: summary > tmdbSummary > tvmazeSummary), one-time clear of copied summaries (9,647 on the live copy); gate 261, tests/realdata-018*.mjs, README + STATUSmref83571c3/plugins/http1/README.md

13.4 KB

  1. # hl:http1 Plugin
  2. HTTP/1.1 server plugin for Hybriel. Epoll-based acceptor with I/O thread pool, optional TLS, keep-alive support.
  3. ## Two Usage Patterns
  4. ### 1. Event-based (class with `on request`)
  5. Extend `NativeHttpServer`, override `on request`, call `listen()`.
  6. ```hybriel
  7. import { NativeHttpServer, Response } from 'hl:http1';
  8. MyServer < NativeHttpServer {
  9. Number port = 8091;
  10. on request(req) {
  11. return new Response('<!DOCTYPE>
  12. <html>
  13. <head>
  14. <title>Hello World!</title>
  15. </head>
  16. <body>
  17. <h1>Hello World!</h1>
  18. <p>This could be your front page!</p>
  19. </body>
  20. </html>
  21. ', {
  22. status = 200;
  23. headers = {
  24. "Set-Cookie" = 'session=test';
  25. };
  26. });
  27. }
  28. }
  29. myServer = { port = 8091; } > new MyServer();
  30. ```
  31. The server runs in the event loop. Each incoming request is dispatched to `on request` as a fiber, so the process stays alive and can handle concurrent requests.
  32. ### 2. Iterator-based (`for (req of server)`)
  33. Call `listen(port)` directly, iterate requests in a loop.
  34. ```hybriel
  35. import { listen, Response } from 'hl:http1';
  36. server = listen(9123);
  37. for (req of server) {
  38. if (req.path == "/hello") {
  39. emit req.response(new Response('<!DOCTYPE>
  40. <html>
  41. <head>
  42. <title>Hello World!</title>
  43. </head>
  44. <body>
  45. <h1>Hello World!</h1>
  46. <p>This could be your front page!</p>
  47. </body>
  48. </html>
  49. ', {
  50. status = 200;
  51. headers = {
  52. "Set-Cookie" = 'session=test';
  53. };
  54. }));
  55. } else {
  56. emit req.response(new Response('Not found', { status = 404; }));
  57. }
  58. }
  59. ```
  60. This blocks on each iteration waiting for the next request. Simpler but sequential.
  61. ## Request Object
  62. Each request has these fields:
  63. | Field | Type | Description |
  64. |-------|------|-------------|
  65. | `method` | String | `"GET"`, `"POST"`, etc. |
  66. | `path` | String | `"/hello"`, `"/api/users"` |
  67. | `query` | Object | Parsed query params (keys/values) |
  68. | `headers` | Object | HTTP headers (keys/values) |
  69. | `body` | String or null | Request body — a `Content-Length` one, or a `Transfer-Encoding: chunked` one de-chunked |
  70. | `bytes` | Bytes | The same body as a Bytes, byte for byte (empty when there is none) |
  71. | `respond` | Handle | Response handle |
  72. ## Response Object
  73. Create a `Response` with body and optional options:
  74. ```hybriel
  75. // Basic
  76. new Response("Hello!")
  77. // With status and headers
  78. new Response('{"ok":true}', {
  79. status = 200;
  80. headers = {
  81. "Content-Type" = 'application/json';
  82. "Set-Cookie" = 'session=abc';
  83. };
  84. })
  85. // Error response
  86. new Response('Not found', { status = 404; })
  87. ```
  88. | Field | Type | Default | Description |
  89. |-------|------|---------|-------------|
  90. | `body` | String or Bytes | `''` | Response body; a Bytes is sent raw, as `application/octet-stream` unless `headers` name a type |
  91. | `status` | Number | `200` | HTTP status code |
  92. | `headers` | Object | `{}` | HTTP headers (key-value pairs) |
  93. ## NativeHttpServer Properties
  94. | Property | Type | Default | Description |
  95. |----------|------|---------|-------------|
  96. | `port` | Number | 8080 | Listen port |
  97. | `host` | String | "0.0.0.0" | Bind address |
  98. | `tls_cert` | String | null | Path to TLS certificate |
  99. | `tls_key` | String | null | Path to TLS private key |
  100. | `threads` | Number | 4 | I/O worker threads |
  101. ## Architecture
  102. ```
  103. Client → epoll acceptor thread → PARK (epoll) → I/O thread pool (TLS + parse)
  104. → request queue → event loop → on request fiber
  105. ```
  106. - Acceptor thread uses epoll to accept connections without blocking
  107. - I/O workers handle TLS handshake and HTTP parsing in parallel
  108. - Parsed requests go into an MPSC queue
  109. - The runtime event loop polls the queue (non-blocking) and spawns a fiber per request
  110. ### Parking: an idle connection costs no thread (mission 084)
  111. A worker's `read()` is blocking, so a connection handed to the pool occupies a whole
  112. worker until bytes arrive. Keep-alive connections used to be re-enqueued straight into the
  113. pool after each response, so an idle one still held a thread — with the default
  114. `threads = 4`, **four quiet sockets starved the server** and every later request hung
  115. (mission 075 saw this as "the second browser session's page never fires `load`").
  116. Connections are therefore **parked** in a dedicated epoll instance (`ParkedConns`) and only
  117. enter the worker queue once they are actually readable — or have hung up, which a worker
  118. reads as EOF and closes. Both the acceptor and the post-response keep-alive path park.
  119. An idle connection now costs one fd and zero threads.
  120. - A parked connection that stays silent for `PARK_IDLE_TIMEOUT_MS` (60s) is closed.
  121. - Workers set `SO_RCVTIMEO` (30s) on every connection as a backstop against a client that
  122. trickles headers; it is no longer the idle-timeout mechanism.
  123. - **An established TLS connection is never parked**: OpenSSL may hold already-decrypted
  124. plaintext that epoll on the raw fd cannot see. Those go straight to a worker, as before,
  125. so a TLS server is still starvable by idle keep-alive connections — the plain-HTTP path
  126. (and the pre-handshake TLS socket, whose ClientHello does arrive on the raw fd) is fixed.
  127. - Regression test: `tests/browser/tests/13-keepalive-starvation.mjs`.
  128. ### WebSocket liveness sweep (mission 091)
  129. The WS reader loop (one thread, epoll, level-triggered EPOLLIN) already wakes on a 500ms
  130. timeout, so liveness costs no timer and no extra thread: on every tick it walks the
  131. upgraded sockets, pings the ones that have gone quiet, and drops the ones whose pong is
  132. overdue.
  133. A drop enqueues the **same `close` event a real FIN would have produced**, so every
  134. consumer above this plugin — hl:web's push subscription registry included — prunes through
  135. the path it already had. No new concept travels upward, and nothing at the `.hl` level
  136. needs a timer facility.
  137. Why it is needed: a peer that vanishes without closing (dead wifi, a suspended laptop, a
  138. host that never sent the FIN) leaves a socket that stays open and **writable**. A server
  139. with nothing to push at it never discovers the corpse, so before this the subscription was
  140. immortal. A closed tab is not this case — the browser's stack sends the FIN, and even under
  141. CDP offline emulation it still answers protocol pings (measured, mission 091), which is why
  142. the regression test's dead peer is a raw socket that goes silent rather than a tab.
  143. | Env var | Default | Meaning |
  144. |---------|---------|---------|
  145. | `HL_WS_PING_MS` | `15000` | a socket quiet this long is pinged |
  146. | `HL_WS_PONG_TIMEOUT_MS` | `10000` | a ping unanswered this long ⇒ the peer is gone |
  147. Read once when the engine starts; timeouts are measured on `CLOCK_MONOTONIC`, so an NTP
  148. step cannot make one fire early or never. Any inbound frame — of any opcode — counts as
  149. proof of life and clears an outstanding ping.
  150. - Regression test: `tests/browser/tests/23-ws-liveness-sweep.mjs` (includes the
  151. falsification: the same silent peer against a sweep-disabled server stays registered).
  152. ### The handshake's cookies + a cookie-grade random (mission 093)
  153. A WebSocket upgrade is an ordinary HTTP request, so the browser attaches the site's cookies
  154. to it. That is the one moment an upgraded socket can be attributed to whoever loaded the
  155. page — after the 101 the request and its headers are freed and the connection carries no
  156. identity of its own. The `connect` event therefore gained a sixth field:
  157. ```
  158. { kind, id, data, binary, code, cookie }
  159. ```
  160. `cookie` is the handshake's `Cookie:` header verbatim, non-empty on `connect` only. Nothing
  161. is parsed here: hl:web's session layer (`plugins/web/web_session.hl`) owns the cookie's name
  162. and meaning, and this plugin owns only its delivery.
  163. ```
  164. randomToken(Number n) // n lowercase hex chars, kernel CSPRNG (getrandom(2)), max 128
  165. ```
  166. A session id is the only thing between a stranger and someone else's session, so it may not
  167. come from a seeded PRNG — `hl:math`'s `random()` is a clock-seeded xoshiro whose state a
  168. handful of outputs reveals, which would make every other id derivable from one's own. This
  169. lives here because a session cookie is an HTTP artifact and this is the plugin that parses
  170. and writes one; it is currently the only CSPRNG reachable from `.hl`.
  171. - Regression test: `tests/browser/tests/25-sessions.mjs` (two real browser instances = two
  172. cookie jars = two sessions).
  173. Ticket #74 added `host` (the handshake's `Host:` header) and ticket #105 `headers`: every
  174. header of the handshake as `name: value` lines, names lowercase, non-empty on `connect` only.
  175. `NativeWebSocketServer` hands them to the `WsClient` as `client.headers`, a hash by name —
  176. which is where hl:web reads the headers a navigation over that socket is constructed with.
  177. ## The outgoing WebSocket client (ticket #109)
  178. Everything above is the SERVER half — a browser (or anything else) dialling in.
  179. `WebSocketClient` is the other direction: Hybriel dialling OUT to someone else's
  180. WebSocket server. Subclass it and answer its three events. A subclass does NOT inherit
  181. the parent's run body (same rule as `NativeHttpServer`/`NativeWebSocketServer`: a parent's
  182. own root never runs for it) — end your own file's root with `connect()`, exactly as those
  183. two end theirs with `listen()`:
  184. ```hybriel
  185. import { WebSocketClient } from 'hl:http1'
  186. inherit WebSocketClient
  187. String url = "wss://example.com/socket"
  188. Hybrid headers = { "Authorization" = "Bearer …" }
  189. on open() {
  190. send('hello')
  191. }
  192. on message(text) {
  193. console.log('got: ' + text)
  194. }
  195. on close(code) {
  196. console.log('closed: ' + code)
  197. }
  198. connect()
  199. ```
  200. Or attach handlers to an instance directly, without subclassing at all — the same
  201. `on instance.event(...)` form used throughout this plugin's own tests:
  202. ```hybriel
  203. import { WebSocketClient } from 'hl:http1'
  204. client = new WebSocketClient(url = 'ws://127.0.0.1:8091/')
  205. on client.open() { client.send('hello') }
  206. on client.message(text) { console.log('got: ' + text) }
  207. on client.close(code) { console.log('closed: ' + code) }
  208. ```
  209. ### Properties
  210. | Property | Type | Default | Description |
  211. |----------|------|---------|-------------|
  212. | `url` | String | `null` | `ws://host[:port][/path]` or `wss://…` |
  213. | `headers` | Object | `{}` | Extra headers sent on the Upgrade request |
  214. | `caFile` | String | `null` | `wss://` only — an EXTRA trust anchor for a self-signed fixture, exactly hl:fetch's field of the same name (never a replacement for the system trust store) |
  215. | `connected` | Boolean | `false` | True between `open` and `close` |
  216. ### Methods
  217. | Method | Answers | Description |
  218. |--------|---------|-------------|
  219. | `connect()` | `true`/`false` | Dial `url` (idempotent — a no-op while already dialling or open). The file root calls this for you. |
  220. | `send(text)` | `1`/`0` | One text frame to the peer; `0` when not open |
  221. | `close([code])` | `1`/`0` | Start the close handshake; `close` still fires once the peer's echo (or a timeout) completes it |
  222. ### Events
  223. ```
  224. emit open() // the connection is open and ready for send()
  225. emit message(String data) // one complete text (or binary) message
  226. emit close(Number code) // the connection ended — RFC 6455 §7.4 code
  227. ```
  228. At most one `open` ever precedes a `close`: a connection that never opened still gets
  229. exactly one `close` (`1006` for a dial that never reached a peer — DNS failure, refused,
  230. TLS handshake failure; `1002` for a handshake the peer answered but got wrong).
  231. ### `wss://` uses hl:fetch's TLS
  232. The client half of `plugins/http/tls_common.zig` — the SAME module hl:fetch uses for
  233. `https://`. The system trust store, hostname verification always on, and `caFile` for an
  234. extra anchor. There is no verify-off switch, on this client either.
  235. ### Ping/pong is automatic; reconnect never is
  236. A quiet connection is pinged after `HL_WS_PING_MS` and dropped if the pong doesn't land
  237. within `HL_WS_PONG_TIMEOUT_MS` — the SAME two env vars the server's liveness sweep reads
  238. (mission 091, above). Any inbound ping is answered with a pong immediately. Neither ever
  239. reaches `.hl`: the protocol requires both, and they are not application information.
  240. **Reconnect is never automatic.** A dropped connection reaches you as exactly one `close`
  241. event and nothing here dials again on its own — call `connect()` again yourself (typically
  242. from your own `on close`) if you want a reconnect policy; `connect()` after a `close` opens
  243. a fresh connection, it is not a one-shot object.
  244. ### Architecture
  245. One dedicated thread per connection — the same shape as hl:fetch's one-thread-per-ticket
  246. model, not the server's shared epoll pool: a client dials a handful of long-lived peers,
  247. not thousands of short ones, and a private thread means a private event queue with no risk
  248. of misdelivering one connection's frames onto another instance's `on open`/`on
  249. message`/`on close`. The thread does DNS + TCP connect, the TLS handshake for `wss://`, the
  250. HTTP Upgrade handshake, then a read loop over the shared RFC 6455 framing
  251. (`plugins/http/ws_common.zig`) — this side's own frames go out through
  252. `writeFrameMasked`/`writeCloseMasked` (RFC 6455 §5.1: "a client MUST mask all frames"),
  253. freshly masked per frame from the kernel CSPRNG.
  254. - Fixtures: `tests/pass/plugins/039_http1_ws_client.hl` (open, send, a server-pushed
  255. frame, close from the client) and
  256. `tests/pass/plugins/040_http1_ws_client_close_reconnect.hl` (close from the SERVER's
  257. side, and an explicit reconnect proving it is never automatic).

Branches

Latest commits

  • f83571c3tracker#18 (mission 067): daily sync by change lists — TMDB /tv|movie/changes (since the stored day, paged) + TVmaze /updates/shows → only our changed titles (followed: full step, unfollowed: light step — changed seasons, no TVmaze), full walk on first run / gap > 14 days / failed list; show record refreshed (title, tmdbSummary, tagline, status, genres …; renamed titles re-indexed); summary = the creator's own text (page: summary > tmdbSummary > tvmazeSummary), one-time clear of copied summaries (9,647 on the live copy); gate 261, tests/realdata-018*.mjs, README + STATUSmre
  • 10bb3f93tracker#28 (mission 061): full cast (all seasons, main cast by episodes, guest stars) + crew (created by, directed by, written by, screenplay, story, music) — stored by the details completion, the daily sync, the search import (one details request) and a background credits job (resumes, RSS limit); show page collapsed after 20 with client-side Show all; showBySlug via a slug map; gate 249, tests/realdata-028.mjs, README + STATUSmre
  • 25a50bc4tracker#26 (mission 059): titles from a filmography are completed — on open (skeleton, step-wise face showComplete, no reload) and by the in-app details repair (resumes, TMDB-paced, series in parts); cast from TMDB credits; gate 231, tests/realdata-026.mjs, README + STATUSmre
  • 1704ec45tracker#17 (mission 057): season caret down/up, skeleton rows while a season loads, sessionless showSeasonEpisodes face (no page re-mount), client-only close; gate 214, tests/realdata-057.mjs, README + STATUSmre
  • f2fe3e36mission 056: README + STATUS (merge, fixes, Hybriel 8590df63, real-data check), tests/realdata-056.mjs, tools/check-public-slugs.hlmre
  • f40c250emission 056: re-vendor hybriel master 8590df63 (#121, #122); an adult title's page is Not found for non-followers; gate: leave the page before stopping the servermre
  • 2b7fdd6cmission 056: signed-out header one row on phones ("Log in", nowrap), backfill skips adult titles' posters, gate checksmre
  • 2c53d5efMerge branch 't16-person' (tracker#16 person pages) into main; filmography shows only public titles (054 adult flag), gate race fix (backfill start line)mre
  • c171227emission 054: hide adult/unknown titles from the public lists and the search; in-app adult-flag backfill (TMDB details + poster per title, resumes), gate + real-data proofmre
  • 139fafd8tracker#16: short bio (4 lines, click = all), real-data check script, README + STATUSmre
  • 93be9476tracker#16: person pages /person/<slug> with the filmography fetched from TMDB on the first visit (step by step), gatemre
  • 47a3cae6STATUS: mission 053 merge commit idsmre
  • dcc5eecaMerge branch 't14-search'mre
  • 03edc783Merge branch 't15-tvmaze'mre
  • 71b46345tracker#15: numbering check by date or title, placeholder titles in other languages, docs + real-data proofmre
  • 6bb2daf1tracker#13: homepage (tiles, intro, latest movies/shows), /shows, /movies/page/N, /my/movies; lists cached in memorymre
  • b8bd1157tracker#14: README + STATUS (search, real-data numbers, gate, merge notes)mre
  • 65c694a8tracker#14: search — header magnifier, /search/<text> (in-memory word-prefix index over titles + people), Fetch from web (TMDB search/multi, ours left out), Add = import via syncShow; gate +25 checks, real-data scriptmre
  • 34f2c15btracker#15: TVmaze merge in the sync (gaps only: new episodes/seasons, empty titles/air dates; numbering check), fake TVmaze episodes + gatemre
  • cbdc4ea7tracker#12: link icons TMDB/IMDb/TVDB/TVmaze; sync fills missing ids (TVmaze lookup); movies fetched via /movie/mre