fix: support no-expression template literals in computed member access by aryanraj45 · Pull Request #20607 · webpack/webpack

aryanraj45 · 2026-03-07T14:43:21Z

Currently, import.meta[\url`] triggers a "Critical dependency" warning and generates incorrect output (({})[`url`]), while import.meta["url"] works correctly. Both are statically analyzable — the difference is only the node type (TemplateLiteralvsLiteral`).

Root cause: extractMemberExpressionChain in JavascriptParser.js only handled Literal string properties in computed member expressions. No-expression template literals (expressions.length === 0) were falling through to the break, losing the member chain.

Fix: Add a TemplateLiteral branch parallel to the existing Literal branch. When a template literal has no expressions (i.e. it's a static string), extract quasis[0].value.cooked the same way Literal extracts property.value.

This allows import.meta[\url`]to be resolved asimport.meta.url`, which is already handled by existing plugins.

What kind of change does this PR introduce?

fix

Did you add tests for your changes?

Yes — test/configCases/parsing/import-meta-computed/ with three cases: ["url"] (existing behavior), [`url`] (the fix), and ["webpackHot"].

Does this PR introduce a breaking change?

No. Existing behavior for Literal, Identifier, and truly dynamic computed access is unchanged.

If relevant, what needs to be documented once your changes are merged?

N/A

Use of AI

No AI

changeset-bot · 2026-03-07T14:43:24Z

🦋 Changeset detected

Latest commit: 5beb6bc

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
webpack	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

alexander-akait · 2026-03-07T14:54:49Z

+					);
+				} else {
+					break;
+				}


Oh, I think we can use const member = evaluateExpression(...) here, it should handle all cases, let's try

And using if (typeof member !== "string") continue;

codspeed-hq · 2026-03-07T15:28:14Z

Merging this PR will degrade performance by 51.46%

⚡ 3 improved benchmarks
❌ 1 regressed benchmark
✅ 140 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	`BASE`	`HEAD`	Efficiency
❌	Memory	`benchmark "many-modules-esm", scenario '{"name":"mode-development-rebuild","mode":"development","watch":true}'`	391 KB	805.5 KB	-51.46%
⚡	Memory	`benchmark "many-modules-commonjs", scenario '{"name":"mode-development-rebuild","mode":"development","watch":true}'`	807.3 KB	390.9 KB	×2.1
⚡	Memory	`benchmark "future-defaults", scenario '{"name":"mode-production","mode":"production"}'`	9.3 MB	7.5 MB	+22.86%
⚡	Memory	`benchmark "many-modules-esm", scenario '{"name":"mode-development","mode":"development"}'`	1.4 MB	1.1 MB	+20.99%

_{Comparing aryanraj45:fix/import-meta-computed-template-literal (6c5c3dc) with main (d6927b4)}

aryanraj45 · 2026-03-07T16:45:46Z

@alexander-akait I have added the test case plss take a look thanks!

alexander-akait · 2026-03-07T18:35:20Z

+				if (member.isString()) {
+					members.push(/** @type {string} */ (member.string));
+				} else if (member.isNumber()) {
+					members.push(`${/** @type {number} */ (member.number)}`);


Do we need to handle number? If yes, please add a test case, but I don't know any numbers in import.meta

aryanraj45 · 2026-03-07T19:07:04Z

@alexander-akait I actually tried using .isString() first as you suggested, but it broke one of the existing tests (import-by-name). The issue was with numeric computed keys like c[2] evaluateExpression(2).isString() returns false, so it completely skips the number and the test fails so i added the number one check . but now i have remove the numerical check as well and to fix this, I used .asString(). Since it's a built-in method of BasicEvaluatedExpression, it neatly handles strings, numbers, template literals, and concatenations all in one single call (basically covering "all cases" like you mentioned!) i hope this approach make sense.Thanks!

alexander-akait · 2026-03-07T19:16:41Z

tests are failed, something was broken

aryanraj45 · 2026-03-07T23:09:26Z

@alexander-akait I have fixed it and implemented the same logic altogether as discuss and suggested please re-review it thanks!

alexander-akait · 2026-03-08T11:11:18Z

-			const property =
-				/** @type {Identifier} */
-				(expression.property).name ||
-				`${/** @type {Literal} */ (expression.property).value}`;


Why this was changed?

Please provide description why this was changed

@aryanraj45 I don't see your answer on this

alexander-akait · 2026-03-08T11:13:12Z

-				memberRanges.push(/** @type {Range} */ (expr.object.range)); // the range of the expression fragment before the literal
+				const evaluated = this.evaluateExpression(expr.property);
+				const member = evaluated.asString();
+				if (typeof member !== "string") break;


You are using const evaluated = this.evaluateExpression(expr.property); result after this check, why do you make unnecessary work? Move it after this check

I don't see any fixes here

…terals

aryanraj45 · 2026-03-08T12:45:38Z

@alexander-akait The Cli.basictest.js failure in the basic CI job isn't related to this PR because I checked and it reproduces on upstream/main as well. It seems to be a terminal color detection issue ("reset" not being found in supported colors) specific to the CI/macOS environment. Happy to investigate further if needed, but I don't think it's a blocker for this change thanks!

alexander-akait · 2026-03-08T12:55:44Z

We have green tests in the main, so it is related

aryanraj45 · 2026-03-08T19:49:07Z

@alexander-akait, I investigated this further and I'm confident this failure is not caused by my changes. Here's the proof:

I ran Cli.basictest.js twice — once with my PR code, and once after reverting JavascriptParser.js back to upstream/main:

PR branch: Tests: 1 failed, 28 passed, 29 total

upstream/main code: Tests: 1 failed, 28 passed, 29 total

The failing test is Cli › createColors › simple (colors by default) — it's a supports-color detection issue on macOS (Node v24 + Darwin). The CI passes because it runs on Linux where color detection works correctly. My PR doesn't touch any color-related code.Please take a view thanks!

alexander-akait · 2026-03-10T10:50:39Z

@aryanraj45 Please answer questions above

aryanraj45 · 2026-03-10T11:00:44Z

@aryanraj45 Please answer questions above

Alexander I have already addressed the question above below your comment of why I am removing (expression.property).name || `${(expression.property).value} in this PR and using members[members.length - 1] for this PR.

alexander-akait

Sorry I don't see your answer above on my questions

aryanraj45 · 2026-03-10T11:35:43Z

@alexander-akait I have already added although please see the ss its mentioned in that as well :

aryanraj45 · 2026-03-10T11:39:56Z

if u say I can describe the changes in a new comment as well for better veiw and clarity sorry for the inconvience

alexander-akait · 2026-03-10T11:45:26Z

Now we are getting this

moduleIdentifier = /home/runner/work/webpack/webpack/test/configCases/css/css-modules/use-style.js; moduleName = ./use-style.js; loc = 8:31-58; message = export 'class' (imported as 'notACssModule') was not found in './style.module.css.inval...; moduleId = ./use-style.js; moduleTrace = [{"originIdentifier":"/home/runner/work/webpack/webpack/test/configCases/css/css-mo...; details = undefined; stack = at HarmonyImportSpecifierDependency.getLinkingErrors (/home/runner/work/webpack/webpack/lib/dependencies/HarmonyImportDependency.js:252:8)

It means something was broken

…t.meta only

aryanraj45 · 2026-03-10T13:19:00Z

@alexander-akait yeah actually the earlier version used [evaluateExpression()] on all computed properties without restricting which object types it applies to. The problem is that [evaluateExpression]( is very aggressive it resolves things like "c" + "lass" into "class" at build time.

This broke the CSS modules test (css-modules/use-style.js), where notACssModule["c" + "lass"] is intentionally written as a concatenation specifically so webpack does not statically resolve it. My previous code was resolving it anyway, which changed the module graph and broke the test.

So What I did to fix it: i have Added a type guard in extractMemberExpressionChain BinaryExpression computed properties are now only evaluated when [expr.object.type === "MetaProperty"] (i.e., the object is [import.meta]. For any other object, BinaryExpression hits the existing break and stays opaque, just like before.

So [import.meta["u" + "rl"]] also works correctly, but notACssModule["c" + "lass"] is left untouched.Please take a look thanks

alexander-akait · 2026-03-10T13:48:02Z

Please stop using AI answers, otherwise you will be ignored in future, when working on projects like this, you must use your own words, thoughts, and ideas in your answers.

alexander-akait · 2026-03-10T13:50:07Z

I spent a fair amount of time on this fix and most of it was generated by the AI without understanding the code, unfortunately I can't spend any more time trying to point out fixes

aryanraj45 · 2026-03-10T13:58:23Z

I spent a fair amount of time on this fix and most of it was generated by the AI without understanding the code, unfortunately I can't spend any more time trying to point out fixes

Alexander I am really sorry about it but ai from my side was used at the end when the test was failing one after another really sorry about your time as well but i too audited it as well when the basic cli was failing I used AI for what can be better fix also crossed checked it really sorry for the inconvince will surely surley take care of it in future or will discuss before using any AI

alexander-akait approved these changes Mar 7, 2026

View reviewed changes

alexander-akait reviewed Mar 7, 2026

View reviewed changes

Comment thread test/configCases/parsing/import-meta-computed/index.js

aryanraj45 force-pushed the fix/import-meta-computed-template-literal branch from e26a6c7 to 57c77ff Compare March 7, 2026 16:41

alexander-akait reviewed Mar 7, 2026

View reviewed changes

Comment thread lib/javascript/JavascriptParser.js Outdated

alexander-akait reviewed Mar 7, 2026

View reviewed changes

aryanraj45 force-pushed the fix/import-meta-computed-template-literal branch from fa2a210 to 2eed21b Compare March 7, 2026 18:57

alexander-akait reviewed Mar 8, 2026

View reviewed changes

aryanraj45 force-pushed the fix/import-meta-computed-template-literal branch 2 times, most recently from a27715c to b7c6694 Compare March 8, 2026 11:34

fix: support computed property access on import.meta with template li…

6c5c3dc

…terals

aryanraj45 force-pushed the fix/import-meta-computed-template-literal branch from b7c6694 to 6c5c3dc Compare March 8, 2026 12:37

alexander-akait requested changes Mar 10, 2026

View reviewed changes

fix: restrict BinaryExpression evaluation in computed access to impor…

5beb6bc

…t.meta only

alexander-akait mentioned this pull request Apr 29, 2026

fix: support no-expression template literals in computed member access #20889

Merged

alexander-akait closed this in #20889 Apr 29, 2026

Uh oh!

Conversation

aryanraj45 commented Mar 7, 2026

Uh oh!

changeset-bot Bot commented Mar 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🦋 Changeset detected

Uh oh!

alexander-akait Mar 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alexander-akait Mar 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

codspeed-hq Bot commented Mar 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will degrade performance by 51.46%

Performance Changes

Uh oh!

Uh oh!

aryanraj45 commented Mar 7, 2026

Uh oh!

Uh oh!

alexander-akait Mar 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

aryanraj45 commented Mar 7, 2026

Uh oh!

alexander-akait commented Mar 7, 2026

Uh oh!

aryanraj45 commented Mar 7, 2026

Uh oh!

alexander-akait Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

alexander-akait Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

alexander-akait Mar 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alexander-akait Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

alexander-akait Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

aryanraj45 commented Mar 8, 2026

Uh oh!

alexander-akait commented Mar 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

aryanraj45 commented Mar 8, 2026

Uh oh!

alexander-akait commented Mar 10, 2026

Uh oh!

aryanraj45 commented Mar 10, 2026

Uh oh!

alexander-akait left a comment

Choose a reason for hiding this comment

Uh oh!

aryanraj45 commented Mar 10, 2026

Uh oh!

aryanraj45 commented Mar 10, 2026

Uh oh!

alexander-akait commented Mar 10, 2026

Uh oh!

aryanraj45 commented Mar 10, 2026

Uh oh!

alexander-akait commented Mar 10, 2026

Uh oh!

alexander-akait commented Mar 10, 2026

Uh oh!

aryanraj45 commented Mar 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

changeset-bot Bot commented Mar 7, 2026 •

edited

Loading

alexander-akait Mar 7, 2026 •

edited

Loading

alexander-akait Mar 7, 2026 •

edited

Loading

codspeed-hq Bot commented Mar 7, 2026 •

edited

Loading

alexander-akait Mar 7, 2026 •

edited

Loading

alexander-akait Mar 10, 2026 •

edited

Loading

alexander-akait commented Mar 8, 2026 •

edited

Loading

aryanraj45 commented Mar 10, 2026 •

edited

Loading