Tree shake parameter defaults

[lukastaegert]

This PR contains:

• bugfix
• feature
• refactor
• documentation
• other

Are tests included?

• yes (bugfixes and features will not be merged without tests)
• no

Breaking Changes?

• yes (breaking changes will not be merged unless absolutely necessary)
• no

List any relevant issue numbers:
Resolves #4466

Description

This allows tree-shaking for unnecessary default parameters in functions. This includes arrow functions but does not include nested defaults in destructuring patterns REPL

// input
const foo = (a = 'removed', { b = 'unused but included' } = 'removed') =>
  console.log(a,b);
foo('a', 'b');

// output
const foo = (a, { b = 'unused but included' }) =>
  console.log(a,b);
foo('a', 'b');

While the feature may sound simple at first glance, it was actually quite complex. The main problem is that JavaScript does not make a difference between not passing a parameter to a function and passing the value undefined: In both cases, the default value will be used.

Therefore, there are quite some subtleties involved which in the end I resolved by extending the mechanism I created for omitting unneeded parameters. This mechanism was removing arguments from function calls if the corresponding parameter was unused. Now, it will also see if parameter defaults need to be included. Care was taken to ensure no side effects are omitted REPL

// input
const foo = (a = 'removed because unused') => effect();
foo();

const bar = (a = effect('included because triggered')) => {};
// It does not make a difference if you pass nothing or undefined
bar(undefined);

const baz = (a = effect('removed')) => {};
baz('value');

// output
const foo = (a) => effect();
foo();

const bar = (a = effect('included because triggered')) => {};
// It does not make a difference if you pass nothing or undefined
bar();

const baz = (a) => {};
baz();

It is very important to track the instance where we "lose track" of how a function is used, e.g. because it is exported, passed to an unknown function etc. In those cases, all necessary defaults are included REPL

// input
export const foo = (a = 'included') => console.log(a);

// output
const foo = (a = 'included') => console.log(a);

export { foo };

[github-actions]

Thank you for your contribution! ❤️

You can try out this pull request locally by installing Rollup via

npm install rollup/rollup#tree-shake-parameter-defaults

or load it into the REPL:
https://rollupjs.org/repl/?pr=4498

[codecov]

Codecov Report

Merging #4498 (d035821) into master (05d294f) will increase coverage by 0.01%.
The diff coverage is 98.40%.

@@            Coverage Diff             @@
##           master    #4498      +/-   ##
==========================================
+ Coverage   98.74%   98.76%   +0.01%     
==========================================
  Files         207      208       +1     
  Lines        7357     7433      +76     
  Branches     2084     2118      +34     
==========================================
+ Hits         7265     7341      +76     
- Misses         33       34       +1     
+ Partials       59       58       -1     

Impacted Files	Coverage Δ
src/ast/nodes/ArrowFunctionExpression.ts	100.00% <ø> (ø)
src/ast/nodes/AssignmentExpression.ts	100.00% <ø> (ø)
src/ast/nodes/AwaitExpression.ts	100.00% <ø> (ø)
src/ast/nodes/ExpressionStatement.ts	87.50% <ø> (ø)
src/ast/nodes/Property.ts	100.00% <ø> (ø)
src/ast/nodes/SequenceExpression.ts	97.14% <ø> (ø)
src/ast/nodes/SpreadElement.ts	100.00% <ø> (ø)
src/ast/nodes/shared/MethodTypes.ts	89.65% <ø> (+6.89%)	⬆️
src/ast/nodes/shared/ObjectMember.ts	100.00% <ø> (+10.00%)	⬆️
src/ast/scopes/FunctionScope.ts	100.00% <ø> (ø)
... and 42 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 05d294f...d035821. Read the comment docs.

[lukastaegert]

This is ready for review now

[bitjson]

For anyone else following – this seems to be the issue behind the revert: #4505

[lukastaegert]

There were actually two more issues. So far so I identified the causes but could not find time to fix stuff yet du to family things, so I reverted to get more time without people reporting the same issues again.

[lukastaegert]

There is a new attempt that does this more conservatively at #4510

[lukastaegert]

To everyone following this, I decided to roll back the feature for good as it turned out to be far too brittle with far too much overhead for my liking at the moment, see #4520