perf: avoid O(N^2) accumulator spread in initHiddenPaths

[ibreathebsb]

Problem

initHiddenPaths (Tree/index.tsx) builds the hidden-paths map with a reduce that spreads the accumulator on every iteration:

return originFlatData.value.reduce((acc, item) => {
  ...
  return {
    ...acc,
    [item.path]: 1,
  };
}, {})

For N flat nodes this allocates a fresh object per iteration and copies every previously-added key — O(N²) total, with N(N-1)/2 property copies.

initHiddenPaths is called synchronously inside setup (line 112) when constructing state.hiddenPaths, so this cost lands on the cold-mount path.

For small JSON it's invisible. For large payloads it dominates: on a real-world 1.9MB payload (~10K flat container nodes) this single line blocks the main thread for ~3–4 seconds during initial render. Profiler confirms the time is spent inside the reduce, not in jsonFlatten or DOM work.

This appears to be the same root cause behind the long-standing #138.

Fix

reduce accumulators are private — mutating in place is safe. One-line change from spread to assignment makes it O(N):

acc[item.path] = 1;
return acc;

Behavior is identical. The final as Record<string, 1> cast is preserved.

Measurements

Same 1.9MB payload, same deep=1, same machine:

	Before	After
initHiddenPaths time	~3.8s	~12ms
Cold mount of <VueJsonPretty>	~4.0s	~50ms

Smaller payloads see no perceptible change (already fast).

Risk

• No API change.
• No behavior change — same output object, same key set.
• No new allocations introduced; one fewer per iteration.
• Type cast unchanged.

Closes #138 (or contributes substantially toward it).

[ibreathebsb]

@leezng Can you review this PR please?

[leezng]

@ibreathebsb Thanks!