Strict mode: hash is required to be imported, but glint reports it as not used.

[NullVoxPopuli]

Sample gjs:

import { hash } from '@ember/helper';

<template>{{yield (hash x=1)}}</template>

Sample Error:

app/components/limber/output/compiler/index.gts:3:1 - error TS6133: 'hash' is declared but its value is never read.

3 import { hash } from '@ember/helper';

Since removing the import creates a runtime error, "hash is not defined", I think Glint should not report the hash import as never read?

I'm using

• ember-source 4.5
• ember-template-imports 3.0.1
• @glint* 0.9.1

[dfreeman]

This is why I want dedicated template syntax for object and array literals 😂

This is definitely a bug, but will need to think a bit on how to best address it. (Note: {{array}} will have this same problem)

[NullVoxPopuli]

you and me both <3

I don't think I captured that in emberjs/rfcs#816 🤔
do you know of any in-progress things I should like to from that meta issue?

[dfreeman]

I remember during discussions around the default helper manager, there was some discussion that it would be easier to avoid supporting named args if there were dedicated object syntax (basically just 1:1 parity with JS function invocation), but I don't know that that was ever actually captured anywhere.

[ef4]

Instead of compiling (hash a=b) directly to ({ a: b }) it could compile to (χ.noop(hash), { a: b }).

I confirmed that this keeps typescript's strict object literal handling, in the sense that it makes unused keys into type errors.

[dfreeman]

Ah, that's clever. Using an empty array literal as a canary (both because I know that's particularly fragile inference behavior in TS and because it could affect us with {{array}}), it looks like that approach can impact the propagation of type info: https://www.typescriptlang.org/play?#code/DYUwLgBAHhC8EG0C6BuAUFFED02IEMA7AT2TQwDoAHAVwGcALACgEYAmAZgEp1Mc9CNALYAjEACcy5NKEjE4EJgAYANIiQ80xLLgiEQANwlTi1es3bcdeAILiA5sJCFIAewBmEMMSogIAckFRCX8IAEs6PVdIfDo6MPtCfBFQL1cIKnxxfCFwCQgPLx8-QMMQiiA

I think we'd want to be careful to make sure that there aren't situations where this can impact inference users may rely on in templates, but if not I like the comma expression approach. Thanks for the suggestion!

[NullVoxPopuli]

with hash, don't the individual values track separately as well? like, if we switched to an object literal, we'd forcibly entangle all values with the whole object?

[ef4]

It already compiles to an object literal. But this isn't code that runs, it's just code that tsc looks at.

[wagenet]

Is this just waiting on someone to do the work?