̶T̶h̶i̶s̶ ̶s̶t̶r̶i̶n̶g̶ ̶i̶s̶ ̶t̶h̶e̶ ̶g̶r̶a̶m̶m̶a̶r̶ ̶t̶h̶a̶t̶ ̶l̶a̶r̶k̶ ̶u̶s̶e̶s̶ ̶t̶o̶ ̶p̶a̶r̶s̶e̶ ̶t̶h̶e̶ ̶u̶s̶e̶r̶ ̶s̶u̶b̶m̶i̶t̶t̶e̶d̶ ̶s̶q̶l̶ ̶i̶n̶t̶o̶ ̶a

̶T̶h̶i̶s̶ ̶s̶t̶r̶i̶n̶g̶ ̶i̶s̶ ̶t̶h̶e̶ ̶g̶r̶a̶m̶m̶a̶r̶ ̶t̶h̶a̶t̶ ̶l̶a̶r̶k̶ ̶u̶s̶e̶s̶ ̶t̶o̶ ̶p̶a̶r̶s̶e̶ ̶t̶h̶e̶ ̶u̶s̶e̶r̶ ̶s̶u̶b̶m̶i̶t̶t̶e̶d̶ ̶s̶q̶l̶ ̶i̶n̶t̶o̶ ̶a̶n̶ ̶A̶S̶T̶.̶ ̶T̶h̶i̶s̶ ̶p̶a̶r̶s̶i̶n̶g̶ ̶i̶s̶ ̶f̶a̶r̶ ̶m̶o̶r̶e̶ ̶c̶o̶m̶p̶l̶e̶t̶e̶ ̶a̶n̶d̶ ̶r̶o̶b̶u̶s̶t̶ ̶t̶h̶a̶n̶ ̶w̶h̶a̶t̶ ̶d̶o̶i̶n̶g̶ ̶t̶h̶i̶s̶ ̶i̶n̶ ̶p̶u̶r̶e̶ ̶p̶y̶t̶h̶o̶n̶ ̶w̶o̶u̶l̶d̶ ̶a̶l̶l̶o̶w̶ ̶(̶w̶i̶t̶h̶o̶u̶t̶ ̶o̶f̶ ̶c̶o̶u̶r̶s̶e̶ ̶i̶m̶p̶l̶e̶m̶e̶n̶t̶i̶n̶g̶ ̶t̶h̶e̶ ̶e̶n̶t̶i̶r̶e̶ ̶l̶e̶x̶e̶r̶ ̶a̶n̶d̶ ̶p̶a̶r̶s̶e̶r̶ ̶t̶h̶a̶t̶ ̶l̶a̶r̶k̶ ̶i̶m̶p̶l̶e̶m̶e̶n̶t̶s̶)̶.̶ T̶h̶e̶r̶e̶ ̶m̶a̶y̶ ̶b̶e̶ ̶s̶o̶m̶e̶ ̶p̶o̶s̶t̶ ̶p̶a̶r̶s̶i̶n̶g̶ ̶v̶a̶l̶i̶d̶a̶t̶i̶o̶n̶ ̶t̶h̶a̶t̶ ̶c̶a̶n̶ ̶b̶e̶ ̶d̶o̶n̶e̶ ̶h̶e̶r̶e̶-̶ ̶b̶u̶t̶ ̶t̶h̶a̶t̶ ̶w̶o̶u̶l̶d̶ ̶b̶e̶ ̶s̶o̶m̶e̶t̶h̶i̶n̶g̶ ̶t̶h̶a̶t̶'̶s̶ ̶b̶e̶y̶o̶n̶d̶ ̶t̶h̶e̶ ̶d̶o̶m̶a̶i̶n̶ ̶o̶f̶ ̶t̶h̶e̶ ̶p̶a̶r̶s̶e̶r̶.̶ Edit: I see what you mean. I surveyed a bunch a parser generator libraries, and they also seemed to use a text based DSL- rather than DSL based on python structures. What you're describing would have made the grammar development more ergonomic and simple.