Maybe I am missing what they meant here, but isn’t the entire point of tree-sitter that you can reuse old trees to make edit updates faster? It is seems quite probable that all the performance gains came from optimising for fixed files with no error recovery, at the cost of how ts is actually used.
Maybe I am missing what they meant here, but isn’t the entire point of tree-sitter that you can reuse old trees to make edit updates faster? It is seems quite probable that all the performance gains came from optimising for fixed files with no error recovery, at the cost of how ts is actually used.