add refactor of extract type by Kingwl · Pull Request #30562 · microsoft/TypeScript

Kingwl · 2019-03-24T10:35:05Z

Fixes #23869

This pr implement a refactor to extract selected types into another type alias in ts or typedef jsdoc in js

The strategy is to look up the type node which is the biggest range and has been included in the selection and create a type alias node before the parent statement.
update the selected type node to reference the generated type.
the case of generic type argument will be collected and convert to type alias / type refernece
the infer type which part of a condition type and type pred will not be allowed to extract

follow things is WIP:

src/services/refactors/extractType.ts

gabritto · 2019-03-28T23:10:08Z

src/services/refactors/extractType.ts

+
+        function visitor(node: Node) {
+            if (isTypeReferenceNode(node)) {
+                if (isIdentifier(node.typeName)) {


why is it not necessary to handle the case where node.typeName is not an identifier but is a qualified name?

Could TypeParameter be a QualifiedName?
In checkAndCollectionTypeArguments we only iter the TypeParameter whitch defined in the statement

i think in the common cases where you have to collect the type, the type name is an identifier and not a qualified name. but i'm really not sure, so that's why i asked.
update: came up with an example of a qualified name for the type query check. i think there could be unusual cases where this check can fail.

gabritto · 2019-03-28T23:10:38Z

src/services/refactors/extractType.ts

+                hasError = true;
+                return;
+            }
+            else if (isTypeQueryNode(node) && isIdentifier(node.exprName)) {


same as above, why is it not necessary to handle the case where node.typeName is not an identifier but is a qualified name?

usually, I think that QualifiedName is not including in current statement

And is not a TypeParameter and do not have TypeArguments

like before, i think common cases are covered by the identifier check, but just came up with the following code:

interface I { a: string, f: (this: O, b: number) => typeof this.a };

something bad happens if we extract typeof this.a

Banned ThisTypeNode, QualityName and most left is ThisIdentifier

gabritto · 2019-03-28T23:34:30Z

src/services/refactors/extractType.ts

+                }
+            }
+            else if (isInferTypeNode(node)) {
+                const conditionalTypeNode = findAncestor(node, isConditionalTypeNode);


i think this might be wrong. we are assuming that the infer type node is only allowed in the extends clause of a conditional type. so we want to check if selection includes the conditional type node to whose extends clause the infer type node belongs. this conditional type node might not be the first conditional type ancestor we find because there could be nested conditional type nodes. for example:

type Crazy<T> = T extends [infer P, (infer R extends string ? string : never)] ? P & R : string;

we can apply the refactoring and extract (infer R extends string ? string : never) to obtain this:

type NewType = (infer R extends string ? string : never); // error: 'infer' declarations are only permitted in the 'extends' clause of a conditional type type Crazy<T> = T extends [infer P, NewType] ? P & R : string;

which gives an error.

sorry, mis operation😅

Find type node which is the first condition that contains selected infer type node in extends node

src/services/refactors/extractType.ts

Co-Authored-By: Kingwl <kingwenlu@gmail.com>

Kingwl · 2019-04-01T09:58:34Z

also this one: #29539

andrewbranch

Hey @Kingwl, thanks so much for this! It's looking great! 🌟

I thought of two small additional cases I'd like to see, something along these lines:

// typeof other parameters within function signature?
function f(a: string, b: /*a*/typeof a/*b*/): /*a*/typeof b/*b*/ {
    return '';
}

// Where do lines get inserted?
// The exact structure here doesn't matter,
// just want to see something within a block body
// to have the behavior defined in tests.
function id<T>(x: T): T {
    return (() => {
        const s: /*a*/typeof x/*b*/ = x;
        return s;
    })();
}

At a quick glance I think these will already produce the correct behavior, just want to make sure. Thanks!

Kingwl · 2019-04-23T03:57:18Z

🆙

Kingwl · 2019-05-05T05:04:39Z

⬆️

sandersn

Looks really good. The only thing is that this is incorrectly refactored.

I also had a couple of questions and two small suggestions, although they are optional.

sandersn · 2019-05-06T22:58:42Z

src/services/refactors/extractType.ts

+        return rangeContainsStartEnd(r1, skipTrivia(file.text, node.pos), node.end);
+    }
+
+    function checkAndCollectionTypeArguments(checker: TypeChecker, selection: TypeNode, statement: Statement, file: SourceFile): TypeParameterDeclaration[] | undefined {


probably should be named collectTypeParameters ? I don't think you're really checking them, and it seems like they are intended to be type parameters in addition to arguments, so matching the type name TypeParameterDeclaration feels a little better.

sandersn · 2019-05-06T23:02:11Z

src/services/refactors/extractType.ts

+    }
+
+    function isStatementButNotBlock(n: Node): n is Statement {
+        return n && isStatement(n) && !isBlock(n);


in what cases will you encounter a block before a statement in findAncestor? That is, a block inside a statement?

Yes,

() => { { // here } }

And i'm cannot remember sure why i do that🤷🏻‍♂️
Is this unnecessary?

Weird, I would have that that //here would have its own statement, but I guess not. In that case, it seems necessary.

sandersn · 2019-05-06T23:09:38Z

src/services/refactors/extractType.ts

+    }
+
+    function checkAndCollectionTypeArguments(checker: TypeChecker, selection: TypeNode, statement: Statement, file: SourceFile): TypeParameterDeclaration[] | undefined {
+        let hasError = false;


seems more elegant to turn this into a return value. It looks like the expected behaviour is to return undefined whenever hasError is true, so stopping the tree walk early should be more efficient.

sandersn · 2019-05-06T23:16:21Z

src/services/refactors/extractType.ts

+                }
+            }
+            else if (isInferTypeNode(node)) {
+                const conditionalTypeNode = findAncestor(node, n => isConditionalTypeNode(n) && rangeContainsSkipTrivia(n.extendsType, node, file));


if the selection contains only a syntactically-correct infer type, then I think this code should (1) not set hasError but also (2) not push any results into result. What happens if you get an empty result list?

Test cases 31-33 show that this works, but I don't understand why.

That caused:

find the selections by largest and topmost type node

Infer type node contains only a TypeParameterDeclaration

in the following case:

type A = /*a*/Promise/*b*/

The selection is the TypeReference which have the same range as the Identifier``Promise.

In cases 31 and 32 were set the hasError flag.
And the case33:

type Item<T> = T extends (infer /*a*/P/*b*/)[] ? P : never

The selection is the TypeParameter(maybe, or an Identifier), that is a declaration rather a TypeNode. and we are already escaped before collectTypeParameters.

sandersn · 2019-05-06T23:20:42Z

src/services/refactors/extractType.ts

+                    }
+                }
+                else {
+                    if (isThisIdentifier(node.exprName.left) && !rangeContainsSkipTrivia(selection, node.parent, file)) {


you should skip this elsewhere too:

class C { m<T>(): /*a*/T | this | number/*b*/ { return {} as any } }

currently produces type NewType<T> = T | this | number

sandersn · 2019-05-07T15:27:34Z

@DanielRosenwasser you'll probably want to highlight this refactor in the release notes since it is the other half of the named-parameters refactor that @gabritto did.

wclr · 2019-07-22T13:06:13Z

@Kingwl
Rename refactoring doesn't work for linked node_modules imports #31825

add basically implement

0b66cfc

Kingwl changed the title ~~add basically implement~~ add basically extract type implement Mar 25, 2019

Kingwl added 6 commits March 25, 2019 14:00

add rename location and add testcase

8a6f1ed

collection type arguments

3948938

disallow infer type

379de4b

add support for typedef convert

0f414b4

refactor info to make type safe

9e405fc

disallow type pred

ead7b05

Kingwl marked this pull request as ready for review March 25, 2019 11:26

Kingwl changed the title ~~add basically extract type implement~~ add refactor of extract type Mar 25, 2019

Kingwl added 2 commits March 25, 2019 19:34

avoid unnecessary branch

5749f7f

disallow type query

0e37661

gabritto requested changes Mar 28, 2019

View reviewed changes

gabritto and others added 8 commits March 28, 2019 19:17

haha😂

0c1866c

Co-Authored-By: Kingwl <kingwenlu@gmail.com>

Update src/services/refactors/extractType.ts

65ecee5

Co-Authored-By: Kingwl <kingwenlu@gmail.com>

Update src/services/refactors/extractType.ts

26ff926

Co-Authored-By: Kingwl <kingwenlu@gmail.com>

add more tests

566da68

add template tag support in jsdoc

73e8297

add support of type parameters constraint

5c28b60

add more tests

a861ec4

merge branch

788d0d2

RyanCavanaugh assigned andrewbranch Apr 4, 2019

andrewbranch reviewed Apr 5, 2019

View reviewed changes

Kingwl added 2 commits April 7, 2019 21:44

add more tests

ee09157

Merge branch 'master' into extrace_types

af994fc

andrewbranch approved these changes Apr 8, 2019

View reviewed changes

andrewbranch requested review from sandersn and sheetalkamat April 14, 2019 20:02

Merge branch 'master' into extrace_types

2a8a7e4

andrewbranch requested review from rbuckton and weswigham April 23, 2019 16:25

sandersn requested changes May 6, 2019

View reviewed changes

Kingwl added 2 commits May 7, 2019 16:02

Merge branch 'master' into extrace_types

1dd1c68

refactor and update function name

bff56b7

sandersn approved these changes May 7, 2019

View reviewed changes

sandersn merged commit 714821f into microsoft:master May 7, 2019

Kingwl deleted the extrace_types branch May 7, 2019 16:18

microsoft locked as resolved and limited conversation to collaborators Oct 21, 2025

Conversation

Kingwl commented Mar 24, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

gabritto Mar 28, 2019 • 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!

Choose a reason for hiding this comment

Uh oh!

gabritto Mar 29, 2019 • 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!

gabritto Mar 28, 2019 • 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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Kingwl commented Apr 1, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

andrewbranch left a comment

Choose a reason for hiding this comment

Uh oh!

Kingwl commented Apr 23, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Kingwl commented May 5, 2019

Uh oh!

sandersn left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Kingwl May 7, 2019 • 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!

Choose a reason for hiding this comment

Uh oh!

sandersn commented May 7, 2019

Uh oh!

wclr commented Jul 22, 2019

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Kingwl commented Mar 24, 2019 •

edited

Loading

gabritto Mar 28, 2019 •

edited

Loading

gabritto Mar 29, 2019 •

edited

Loading

gabritto Mar 28, 2019 •

edited

Loading

Kingwl commented Apr 1, 2019 •

edited

Loading

Kingwl commented Apr 23, 2019 •

edited

Loading

Kingwl May 7, 2019 •

edited

Loading