Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add string literal completions for package.json imports field #57718

Draft
wants to merge 2 commits into
base: main
Choose a base branch
from

Conversation

Andarist
Copy link
Contributor

@Andarist Andarist commented Mar 11, 2024

closes #52460
closes #57680
closes #57777

Currently, this only has tests based on #55015 but I still have to add more

@typescript-bot typescript-bot added the For Backlog Bug PRs that fix a backlog bug label Mar 11, 2024
@@ -6288,6 +6290,21 @@ export function getPossibleOriginalInputExtensionForExtension(path: string) {
[Extension.Tsx, Extension.Ts, Extension.Jsx, Extension.Js];
}

/** @internal */
export function getPossibleOriginalInputPathWithoutChangingExt(
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this is kinda a reverse of getOutputPathWithoutChangingExt:

function getOutputPathWithoutChangingExt(
inputFileName: string,
ignoreCase: boolean,
outputDir: string | undefined,
getCommonSourceDirectory: () => string,
): string {
return outputDir ?
resolvePath(
outputDir,
getRelativePathFromDirectory(getCommonSourceDirectory(), inputFileName, ignoreCase),
) :
inputFileName;
}

@@ -6288,6 +6290,21 @@ export function getPossibleOriginalInputExtensionForExtension(path: string) {
[Extension.Tsx, Extension.Ts, Extension.Jsx, Extension.Js];
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

tangent: would it make any sense to limit this to avoid returning js/jsx for path with ts/tsx?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It seems like it. If it doesn’t affect anything negatively, I'd say go for it.

host: LanguageServiceHost,
moduleSpecifierResolutionHost: ModuleSpecifierResolutionHost | undefined
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TODO: it was annoying to pass moduleSpecifierResolutionHost here from all of the callers but I probably should do that

matches = concatenate(matches, inputMatches);
}

matches = concatenate(matches, directories);
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TODO: this lacks inputDirectories


function trimPrefixAndSuffix(path: string): string | undefined {
if (inputBaseDirectory && inputBaseDirectory !== baseDirectory) {
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TODO: a helper has to be introduced to deduplicate this branch with the ones that creates original matches

Comment on lines 849 to +850
isExports: boolean,
isImports: boolean,
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

it's somewhat annoying to have those separate parameters but I need to distinguish between those 2 cases. Alternatively, I could introduce an enum

@@ -875,7 +887,7 @@ function addCompletionEntriesFromPathsOrExports(
if (typeof pathPattern === "string" || matchedPath === undefined || comparePaths(key, matchedPath) !== Comparison.GreaterThan) {
pathResults.push({
matchedPattern: isMatch,
results: getCompletionsForPathMapping(keyWithoutLeadingDotSlash, patterns, fragment, baseDirectory, extensionOptions, isExports && isMatch, host)
results: getCompletionsForPathMapping(keyWithoutLeadingDotSlash, patterns, fragment, baseDirectory, extensionOptions, (isExports || isImports) && endsWith(keyWithoutLeadingDotSlash, '*'), isImports, compilerOptions, host, moduleSpecifierResolutionHost)
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this isExportsWildcard parameter here is a little bit surprising. It feels like it could be checked further down the callee chain - but I'm not sure yet

}
if (!seenPackageScope) {
const packageFile = combinePaths(ancestor, "package.json");
if (seenPackageScope = tryFileExists(host, packageFile)) {
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TODO: this has to be heavily deduplicated with the existing branch that handles exports

@@ -938,10 +951,41 @@ function getCompletionEntriesForNonRelativeModules(
}
}
if (!foundGlobal) {
let seenPackageScope = false;
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TODO: add a test case that will show that it's needed

): readonly NameAndKind[] {
if (!endsWith(path, "*")) {
Copy link
Contributor Author

@Andarist Andarist Mar 11, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This should also fix completions when * is somewhere in the middle of paths. The algorithm is largely the same between paths/exports/imports after all (I need to add a test case for this).

Note that nothing here fixes completions with multiple wildcards in the value - I don't want to work on that.

@@ -1072,9 +1153,11 @@ function getModulesForPathsPattern(

const normalizedSuffix = normalizePath(parsed.suffix);
const declarationExtension = normalizedSuffix && getDeclarationEmitExtensionForPath("_" + normalizedSuffix);
const matchingSuffixes = declarationExtension ? [changeExtension(normalizedSuffix, declarationExtension), normalizedSuffix] : [normalizedSuffix];
const inputExtension = isImports && normalizedSuffix ? getPossibleOriginalInputExtensionForExtension("_" + normalizedSuffix) : undefined;
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All of the logic in this function assumes the same extensions between potential inputs and outputs and stuff. This can impact the used globs but I think it's largely fine right now - unless somebody would come up with a compelling test case that should work differently.

});
verify.completions({
marker: ["2"],
exact: ["foo.js"],
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This currently fails. I took this test case from https://github.com/microsoft/TypeScript/pull/55015/files#diff-c8e8521e91d6946ce7869359a143f8cf7b65ee15703a4e04c7d73dcbaa5682a4 but I think that perhaps it should just be removed/tweaked in both places. A trailing slash is already not supported by node@>=17 and it was deprecated before that. There is little value in supporting this - in my opinion, it would be better to discourage this completely.

In addition to that, I noticed that the previous segment gets suggested in a situation like this today:

{
    "name": "pkg_test",
    "exports": {
        "./test/": "./"
    }
}
import { bazinga } from 'pkg_test/test//*1*/'

At marker 1 I get test again, it's imho a clear bug and I can fix it here - after we agree on not supporting this old deprecated pattern at all.

@Andarist
Copy link
Contributor Author

@andrewbranch I opened this as a draft because there are some minor cleanups to be done here. I'd appreciate an early review here though - in case there is something fundamentally wrong with this.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I need more test cases like this but with a pattern. Atm, I didn't add any special logic for declarationDir - this "exact" match works solely based on the fact that it's exact. It doesn't even need to check if the corresponding file exists on the disk or not so it doesn't perform any output->input mapping etc. This has to be fixed.

@DanielRosenwasser
Copy link
Member

@Andarist did you still want to pursue this? We're looking to get this done in the TS 5.7 timeframe.

@Andarist
Copy link
Contributor Author

@DanielRosenwasser yes, it would be great if @andrewbranch could give this a quick look to check if im not doing anything overly wrong so I dont spend too much time on cleaning up things that will turn out to be wrong in the end ;p

Copy link
Member

@andrewbranch andrewbranch left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, this looks on the right track to me, thanks!

@@ -6288,6 +6290,21 @@ export function getPossibleOriginalInputExtensionForExtension(path: string) {
[Extension.Tsx, Extension.Ts, Extension.Jsx, Extension.Js];
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It seems like it. If it doesn’t affect anything negatively, I'd say go for it.


verify.completions({
marker: ["1"],
exact: ["#something.js"],
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would be good to have a version of this test with allowImportingTsExtensions where this appears as "#something.ts". (Pretty sure it will work without additional changes)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
For Backlog Bug PRs that fix a backlog bug
Projects
None yet
4 participants