What Lean code is generating the AST for the parser module?

[zaporter]

Hi! Big fan of this repo and this project. This is a bit of a basic question, but I am stumped.

What lean code provides the AST? lake exe repl (repo) doesn't seem to support ast-retrieval. (even the quick example shows an empty ast object:

DeepSeek-Prover-V1.5/quick_start.py

Line 69 in 2c4ba91

'''{'sorries': [], 'tactics': [], 'errors': [], 'warnings': [{'severity': 'warning', 'pos': {'line': 14, 'column': 7}, 'endPos': {'line': 14, 'column': 10}, 'data': "unused variable `h₁'`\nnote: this linter can be disabled with `set_option linter.unusedVariables false`"}, {'severity': 'warning', 'pos': {'line': 15, 'column': 7}, 'endPos': {'line': 15, 'column': 10}, 'data': "unused variable `h₂'`\nnote: this linter can be disabled with `set_option linter.unusedVariables false`"}, {'severity': 'warning', 'pos': {'line': 19, 'column': 35}, 'endPos': {'line': 19, 'column': 38}, 'data': 'Used `tac1 <;> tac2` where `(tac1; tac2)` would suffice\nnote: this linter can be disabled with `set_option linter.unnecessarySeqFocus false`'}, {'severity': 'warning', 'pos': {'line': 20, 'column': 15}, 'endPos': {'line': 20, 'column': 18}, 'data': 'Used `tac1 <;> tac2` where `(tac1; tac2)` would suffice\nnote: this linter can be disabled with `set_option linter.unnecessarySeqFocus false`'}], 'infos': [], 'system_messages': '', 'system_errors': None, 'ast': {}, 'verified_code': "import Mathlib\nimport Aesop\n\nset_option maxHeartbeats 0\n\nopen BigOperators Real Nat Topology Rat\n\n/-- The second and fourth terms of a geometric sequence are $2$ and $6$. Which of the following is a possible first term?\nShow that it is $\x0crac{2\\sqrt{3}}{3}$.-/\ntheorem amc12b_2003_p6 (a r : ℝ) (u : ℕ → ℝ) (h₀ : ∀ k, u k = a * r ^ k) (h₁ : u 1 = 2)\n (h₂ : u 3 = 6) : u 0 = 2 / Real.sqrt 3 ∨ u 0 = -(2 / Real.sqrt 3) := by\n simp_all only [Nat.one_eq_succ_zero, Nat.zero_eq, zero_add, Nat.add_succ, Nat.add_zero,\n Nat.succ_add]\n have h₁' : a * r = 2 := by simpa [h₀] using h₁\n have h₂' : a * r ^ 3 = 6 := by simpa [h₀] using h₂\n have h₃ : r ^ 2 = 3 := by\n nlinarith\n have h₄ : a = 2 / Real.sqrt 3 ∨ a = -(2 / Real.sqrt 3) := by\n apply eq_or_eq_neg_of_sq_eq_sq <;>\n field_simp <;>\n nlinarith\n simpa [h₀] using h₄", 'pass': True, 'complete': True, 'verify_time': 23.28123140335083}'''

)

Did the codebase originally use LeanDojo's ExtractData.lean to get the AST? (If it still does, where?? I can't find any references to LeanDojo or the required code to generate an AST (does mathlib4 have it builtin now?)).

Thanks for your time! (& for this cool paper / project!)

[xinhjBrant]

Thank you for your attention! The AST extraction code is originally based on LeanDojo's ExtractData.lean and references the following repository for the usage of this project: ExtractData.lean.