Pattern Matching Champions on 2024-04-09

02:57	<Jack Works>	`{ length: 0 }` needs to match an empty string, for example We don't do type coerce, but do we need to do primitive boxing?
02:57	<ljharb>	i mean, we could decide not to. but that would be a radical departure from the rest of the language.
02:57	<ljharb>	if you can do `x.length` on it, and you can do `const { length } = x` on it, you should be able to `{ length }` pattern match on it
02:58	<ljharb>	so i think "consistency with destructuring" forces us to do boxing, full stop
03:00	<Jack Works>	let's decide it next group meeting
03:00	<ljharb>	sure, that's fine, but i'm not clear on what possible argument could justify "different from destructuring"
03:01	<Jack Works>	Probably still a little too early for a thorough review of the spec, but I noticed that https://tc39.es/proposal-pattern-matching/#sec-match-expression-clauses-runtime-semantics-evaluation only cares about ECMAScript language values rather than completion values, which doesn't work as it prevents ThrowCompletion from bubbling out of match (or `return`/`break`/`continue` in the event `do` expressions are supported). returns either a normal completion containing either an ECMAScript language value or not-matched, or an abrupt completion
03:01	<ljharb>	also MM just posted an issue moments ago, and i'm not sure i understand it, but it seems like it's in direct conflict with the "bindings by default" change we made in response to yulia's feedback
03:01	<Jack Works>	Why do you say so? It contains Abrupt Completion
03:54	<rbuckton>	returns either a normal completion containing either an ECMAScript language value or not-matched, or an abrupt completion Not the return. Step 2 of the 2nd and 3rd productions: Let result be ? MatchExpressionClausesEvaluation of MatchExpressionClauses with arguments subject and cacheGroup. If result is an ECMAScript language value, return result. Return ? MatchExpressionClauseEvaluation of MatchExpressionClause with arguments subject and cacheGroup. result will never be an ECMAScript language value. We need to check for a normal completion whose value is `not-matched`.
03:55	<rbuckton>	Wait, I'm an idiot. That's handled by 1.
03:55	<rbuckton>	That's what I get for reading spec text on a 5hr car trip.
03:57	<rbuckton>	The use of ReturnIfAbrupt (?) sort of obfuscates the intention, IMO. It was easy to overlook the fact that what we care about is `not-matched`.
03:58	<rbuckton>	Either by explicitly looking for `not-matched`, or by adding an "Assert: result is `not-matched`"
04:08	<Jack Works>	I'll do that next time