Conversation
…i-session-agent) The previous implementation treated every uppercase letter as a word boundary, breaking agent names with acronyms like AI, API, UI, DB. Now consecutive uppercase letters are kept together as a single unit before splitting. Fixes #924
Co-authored-by: whoiskatrin <whoiskatrin@users.noreply.github.com>
|
commit: |
…age-lock conflict)
|
Got to block this, the fix needs to be in partyserver, and I'll do that myself, I need to check for backward compat as well. PR #941 Review: Fix kebab-case acronym splitting bugIssue SummaryIssue #924 reports that What the PR Does
Blocking Issues1. Critical: Routing mismatch with partyserverThis is a show-stopper. Partyserver bundles its own internal copy of // node_modules/partyserver/dist/index.js:250-255
function camelCaseToKebabCase(str) {
if (str === str.toUpperCase() && str !== str.toLowerCase())
return str.toLowerCase().replace(/_/g, "-");
let kebabified = str.replace(/[A-Z]/g, (letter) => `-${letter.toLowerCase()}`);
kebabified = kebabified.startsWith("-") ? kebabified.slice(1) : kebabified;
return kebabified.replace(/_/g, "-").replace(/-$/, "");
}This function is used by // node_modules/partyserver/dist/index.js:276-287
async function routePartykitRequest(req, env$1 = env, options) {
if (!serverMapCache.has(env$1)) {
const namespaceMap = {};
const bindingNames$1 = {};
for (const [k, v] of Object.entries(env$1))
if (v && typeof v === "object" && "idFromName" in v && typeof v.idFromName === "function") {
const kebab = camelCaseToKebabCase(k); // <-- uses OLD algorithm
namespaceMap[kebab] = v;
bindingNames$1[kebab] = k;
}
serverMapCache.set(env$1, namespaceMap);
bindingNameCache.set(env$1, bindingNames$1);
}And // packages/agents/src/index.ts:4282-4287
export async function routeAgentRequest<Env>(request: Request, env: Env, options?: AgentOptions<Env>) {
return routePartykitRequest(request, env as Record<string, unknown>, {
prefix: "agents",
...(options as PartyServerOptions<Record<string, unknown>>)
});
}The failure scenario after this PR:
Before this PR, both sides agree on Resolution options:
2. Missing changesetPer repo policy, changes to Non-Blocking Issues3. Digit handling added but untestedLines 30–31 of kebabified = kebabified.replace(/([a-zA-Z])(\d)/g, "$1-$2");
kebabified = kebabified.replace(/(\d)([a-zA-Z])/g, "$1-$2");No tests cover this. Should either add tests (e.g., 4.
|
2 participants
Here's a summary of the changes:
Fix:
camelCaseToKebabCasenow handles acronyms correctlyProblem: The function at
packages/agents/src/utils.ts:6treated every uppercase letter as a word boundary, soAISessionAgentbecamea-i-session-agentinstead ofai-session-agent.Root cause: The old implementation used a simple
/[A-Z]/gregex that replaced each uppercase letter individually with-<lowercase>.Fix: Replaced the single regex with a two-pass approach:
/([A-Z]+)([A-Z][a-z])/g— splits before the last capital of consecutive uppercase letters (handles acronyms likeAI,API,HTTP)/([a-z])([A-Z])/g— splits between lowercase-to-uppercase transitions (standard camelCase boundaries)Results:
AISessionAgenta-i-session-agentai-session-agentAPIEndpointa-p-i-endpointapi-endpointMyUIComponentmy-u-i-componentmy-ui-componentTestStateAgenttest-state-agenttest-state-agent(unchanged)Files changed:
packages/agents/src/utils.ts— fixed the function with improved JSDocpackages/agents/src/tests/utils.test.ts— added 22 tests covering acronyms, basic camelCase, kebab-case passthrough, underscores, and edge casesVerification: All 552 existing workers tests pass, plus the 22 new unit tests. Typecheck and lint pass.
Closes #924
github run