username sanitisation for registration

d2m/actions/register-user.js

@@ -4,39 +4,16 @@ const assert = require("assert")
 
 const passthrough = require("../../passthrough")
 const { discord, sync, db } = passthrough
-/** @type {import("../../matrix/mreq")} */
-const mreq = sync.require("../../matrix/mreq")
+/** @type {import("../../matrix/api")} */
+const api = sync.require("../../matrix/api")
 /** @type {import("../../matrix/file")} */
 const file = sync.require("../../matrix/file")
 
-async function registerUser(username) {
-	assert.ok(username.startsWith("_ooye_"))
-	/** @type {import("../../types").R.Registered} */
-	const res = await mreq.mreq("POST", "/client/v3/register", {
-		type: "m.login.application_service",
-		username
-	})
-	return res
-}
-
 /**
  * A sim is an account that is being simulated by the bridge to copy events from the other side.
  * @param {import("discord-api-types/v10").APIUser} user
  */
 async function createSim(user) {
 	assert.notEqual(user.discriminator, "0000", "user is not a webhook")
-	fetch("https://matrix.cadence.moe/_matrix/client/v3/register", {
-		method: "POST",
-		body: JSON.stringify({
-			type: "m.login.application_service",
-			username: "_ooye_example"
-		}),
-		headers: {
-			Authorization: `Bearer ${reg.as_token}`
-		}
-	}).then(res => res.text()).then(text => {
-
-		console.log(text)
-	}).catch(err => {
-		console.log(err)
-	})
+	api.register("_ooye_example")
+}

d2m/converters/user-to-mxid.js

@@ -0,0 +1,74 @@
+// @ts-check
+
+const assert = require("assert")
+
+const passthrough = require("../../passthrough")
+const { sync, db } = passthrough
+
+/**
+ * Downcased and stripped username. Can only include a basic set of characters.
+ * https://spec.matrix.org/v1.6/appendices/#user-identifiers
+ * @param {import("discord-api-types/v10").APIUser} user
+ * @returns {string} localpart
+ */
+function downcaseUsername(user) {
+	// First, try to convert the username to the set of allowed characters
+	let downcased = user.username.toLowerCase()
+		// spaces to underscores...
+		.replace(/ /g, "_")
+		// remove disallowed characters...
+		.replace(/[^a-z0-9._=/-]*/g, "")
+		// remove leading and trailing dashes and underscores...
+		.replace(/(?:^[_-]*|[_-]*$)/g, "")
+	// The new length must be at least 2 characters (in other words, it should have some content)
+	if (downcased.length < 2) {
+		downcased = user.id
+	}
+	return downcased
+}
+
+/** @param {string[]} preferences */
+function* generateLocalpartAlternatives(preferences) {
+	const best = preferences[0]
+	assert.ok(best)
+	// First, suggest the preferences...
+	for (const localpart of preferences) {
+		yield localpart
+	}
+	// ...then fall back to generating number suffixes...
+	let i = 2
+	while (true) {
+		yield best + (i++)
+	}
+}
+
+/**
+ * @param {import("discord-api-types/v10").APIUser} user
+ * @returns {string}
+ */
+function userToSimName(user) {
+	assert.notEqual(user.discriminator, "0000", "cannot create user for a webhook")
+
+	// 1. Is sim user already registered?
+	const existing = db.prepare("SELECT sim_name FROM sim WHERE discord_id = ?").pluck().get(user.id)
+	if (existing) return existing
+
+	// 2. Register based on username (could be new or old format)
+	const downcased = downcaseUsername(user)
+	const preferences = [downcased]
+	if (user.discriminator.length === 4) { // Old style tag? If user.username is unavailable, try the full tag next
+		preferences.push(downcased + user.discriminator)
+	}
+
+	// Check for conflicts with already registered sims
+	/** @type {string[]} */
+	const matches = db.prepare("SELECT sim_name FROM sim WHERE sim_name LIKE ? ESCAPE '@'").pluck().all(downcased + "%")
+	// Keep generating until we get a suggestion that doesn't conflict
+	for (const suggestion of generateLocalpartAlternatives(preferences)) {
+		if (!matches.includes(suggestion)) return suggestion
+	}
+
+	throw new Error(`Ran out of suggestions when generating sim name. downcased: "${downcased}"`)
+}
+
+module.exports.userToSimName = userToSimName

d2m/converters/user-to-mxid.test.js

@@ -0,0 +1,33 @@
+const {test} = require("supertape")
+const tryToCatch = require("try-to-catch")
+const assert = require("assert")
+const {userToSimName} = require("./user-to-mxid")
+
+test("user2name: cannot create user for a webhook", async t => {
+   const [error] = await tryToCatch(() => userToSimName({discriminator: "0000"}))
+   t.ok(error instanceof assert.AssertionError, error.message)
+})
+
+test("user2name: works on normal name", t => {
+   t.equal(userToSimName({username: "Harry Styles!", discriminator: "0001"}), "harry_styles")
+})
+
+test("user2name: works on emojis", t => {
+   t.equal(userToSimName({username: "Cookie 🍪", discriminator: "0001"}), "cookie")
+})
+
+test("user2name: works on crazy name", t => {
+   t.equal(userToSimName({username: "*** D3 &W (89) _7//-", discriminator: "0001"}), "d3_w_89__7//")
+})
+
+test("user2name: adds discriminator if name is unavailable (old tag format)", t => {
+   t.equal(userToSimName({username: "BOT$", discriminator: "1234"}), "bot1234")
+})
+
+test("user2name: adds number suffix if name is unavailable (new username format)", t => {
+   t.equal(userToSimName({username: "bot", discriminator: "0"}), "bot2")
+})
+
+test("user2name: uses ID if name becomes too short", t => {
+   t.equal(userToSimName({username: "f***", discriminator: "0001", id: "9"}), "9")
+})