List position based formatting

[saschanaz]

Revives #13574

Fixes #5830, fixes #5890, fixes #6252, fixes #6320, fixes #6451, fixes #10681, fixes #26906

[ghost]

function getListIfVisualStartEndIsInListRange(list: NodeArray<Node> | undefined, start: number, end: number, node: Node, sourceFile: SourceFile): NodeArray<Node> | undefined {
    return list && rangeContainsStartEnd(getListRange(node, list, sourceFile), start, end) ? list : undefined;
}
function getListRange(node: Node, list: TextRange, sourceFile: SourceFile): TextRange {
    const children = node.getChildren(sourceFile);
    for (let i = 1; i < children.length - 1; i++) {
        if (children[i].pos === list.pos && children[i].end === list.end) {
            return { pos: children[i - 1].end, end: children[i + 1].getStart(sourceFile) };
        }
    }
    return list;
}

• Should pass in sourceFile to some methods for performance.
• width = end - start, so end - width = start
• What does "visual" mean here?

[saschanaz]

NodeArray starts at the previous token's end positions and ends at the last list item's end position. So think about the following code:

func(
  // visually the list starts at line 0 and ends in line 2, but actually it ends at line 0
);

func(
  1,
  // visually the list starts at line 0 and ends in line 3, but actually it ends at line 1
  // which means the position of these comments does not belong to the list range.
);

[ghost]

You could create a closure so you don't have to repeat the arguments start, end, node, sourceFile everywhere.

[DanielRosenwasser]

I'm pulling this in, and if we'd like to address any changes related to code cleanliness or perf we can do this after. Thank you so much for sticking it out and sending out these changes again.

[DanielRosenwasser]

⭐️ ⭐️ ⭐️ ⭐️ ⭐️ ❤️