regression: java symbols no longer displayed as a tree

[fbricon]

With vscode-java/jdt.ls, a textDocument/documentSymbol call returns something like:

[Trace - 12:48:07] Received response 'textDocument/documentSymbol - (33)' in 6ms.
Result: [
    {
        "name": "main(String[])",
        "kind": 6,
        "location": {
            "uri": "file:///Users/fbricon/Desktop/Standalone.java",
            "range": {
                "start": {
                    "line": 12,
                    "character": 20
                },
                "end": {
                    "line": 12,
                    "character": 24
                }
            }
        },
        "containerName": "Standalone"
    },
    {
        "name": "Standalone",
        "kind": 5,
        "location": {
            "uri": "file:///Users/fbricon/Desktop/Standalone.java",
            "range": {
                "start": {
                    "line": 10,
                    "character": 13
                },
                "end": {
                    "line": 10,
                    "character": 23
                }
            }
        },
        "containerName": "Standalone.java"
    }
]

vscode-code-outline displays the main method at the root of the tree, same level as Standalone, instead of nesting it under Standalone. My guess is the latest algorithm gets confused by the root node having a containerName (Standalone.java) instead of a blank value. This used to work a while ago:

[patrys]

This library relies on location rather than containerName as there was no way to have it work correctly based on container names (there can be multiple objects with the same name).

[pvanek]

@patrys Patryk, what's expected format of json to get working outline tree, please? It looks like we are facing the same flat structure in our language server output...

[fbricon]

@patrys can you explain how the location is used to build the tree? Looking at the output for an html content, it's definitely not obvious.