|
| 1 | +from __future__ import annotations |
| 2 | + |
| 3 | +from pathlib import Path |
| 4 | + |
| 5 | + |
| 6 | +def replace_once(path: str, old: str, new: str) -> None: |
| 7 | + file = Path(path) |
| 8 | + text = file.read_text() |
| 9 | + count = text.count(old) |
| 10 | + if count != 1: |
| 11 | + raise RuntimeError( |
| 12 | + f"{path}: expected exactly one match, found {count}\nOLD:\n{old}" |
| 13 | + ) |
| 14 | + file.write_text(text.replace(old, new, 1)) |
| 15 | + |
| 16 | + |
| 17 | +replace_once( |
| 18 | + "src/nns/_reg_engine.py", |
| 19 | + "import math\nimport re\n", |
| 20 | + "import math\nimport re\nfrom itertools import pairwise\n", |
| 21 | +) |
| 22 | +replace_once( |
| 23 | + "src/nns/_reg_engine.py", |
| 24 | + "ord_value = max(1, int(math.floor(dep * 10 + 0.5)))", |
| 25 | + "ord_value = max(1, math.floor(dep * 10 + 0.5))", |
| 26 | +) |
| 27 | +replace_once( |
| 28 | + "src/nns/_reg_engine.py", |
| 29 | + "ord_value = max(1, int(math.floor(ord_value / 2)))", |
| 30 | + "ord_value = max(1, math.floor(ord_value / 2))", |
| 31 | +) |
| 32 | +replace_once( |
| 33 | + "src/nns/_reg_engine.py", |
| 34 | + "for a, b in zip(unique_knots[:-1], unique_knots[1:]):", |
| 35 | + "for a, b in pairwise(unique_knots):", |
| 36 | +) |
| 37 | +replace_once( |
| 38 | + "src/nns/_reg_engine.py", |
| 39 | + '"Variable": names + ["DENOMINATOR"],', |
| 40 | + '"Variable": [*names, "DENOMINATOR"],', |
| 41 | +) |
| 42 | +replace_once( |
| 43 | + "src/nns/_reg_engine.py", |
| 44 | + "k = max(1, int(math.floor((1.0 - dep) * math.sqrt(x.shape[0]))))", |
| 45 | + "k = max(1, math.floor((1.0 - dep) * math.sqrt(x.shape[0])))", |
| 46 | +) |
| 47 | + |
| 48 | +replace_once( |
| 49 | + "src/nns/_rrng.py", |
| 50 | + "v1 = int(math.floor(self.unif_rand() * 65536))", |
| 51 | + "v1 = math.floor(self.unif_rand() * 65536)", |
| 52 | +) |
| 53 | +replace_once( |
| 54 | + "src/nns/_rrng.py", |
| 55 | + "bits = int(math.ceil(math.log2(dn)))", |
| 56 | + "bits = math.ceil(math.log2(dn))", |
| 57 | +) |
| 58 | +replace_once( |
| 59 | + "src/nns/_rrng.py", |
| 60 | + " def sample_int(self, n: int, size: int | None = None, replace: bool = False) -> NDArray[np.int64]:\n", |
| 61 | + " def sample_int(\n" |
| 62 | + " self, n: int, size: int | None = None, replace: bool = False\n" |
| 63 | + " ) -> NDArray[np.int64]:\n", |
| 64 | +) |
| 65 | + |
| 66 | +replace_once( |
| 67 | + "src/nns/boost.py", |
| 68 | + "size = max(1, min(n_obs - 1, int(round(cv_fraction * n_obs))))", |
| 69 | + "size = max(1, min(n_obs - 1, round(cv_fraction * n_obs)))", |
| 70 | +) |
| 71 | +replace_once( |
| 72 | + "src/nns/boost.py", |
| 73 | + "k_small = max(1, int(math.floor(math.sqrt(minimum_train_size))))", |
| 74 | + "k_small = max(1, math.floor(math.sqrt(minimum_train_size)))", |
| 75 | +) |
| 76 | +replace_once( |
| 77 | + "src/nns/boost.py", |
| 78 | + "k_candidates = list(dict.fromkeys(list(range(1, k_small + 1)) + [minimum_train_size]))", |
| 79 | + "k_candidates = list(dict.fromkeys([*range(1, k_small + 1), minimum_train_size]))", |
| 80 | +) |
| 81 | + |
| 82 | +replace_once( |
| 83 | + "src/nns/stack.py", |
| 84 | + "def _scalar_integer(value: Any, name: str, minimum: int = 0, allow_null: bool = False) -> int | None:\n", |
| 85 | + "def _scalar_integer(\n" |
| 86 | + " value: Any, name: str, minimum: int = 0, allow_null: bool = False\n" |
| 87 | + ") -> int | None:\n", |
| 88 | +) |
| 89 | +replace_once( |
| 90 | + "src/nns/stack.py", |
| 91 | + " [class_values.index(v) + 1 for v in (dv.tolist() if response_categorical else dv.astype(np.float64).tolist())],\n", |
| 92 | + " [\n" |
| 93 | + " class_values.index(v) + 1\n" |
| 94 | + " for v in (\n" |
| 95 | + " dv.tolist()\n" |
| 96 | + " if response_categorical\n" |
| 97 | + " else dv.astype(np.float64).tolist()\n" |
| 98 | + " )\n" |
| 99 | + " ],\n", |
| 100 | +) |
| 101 | +replace_once( |
| 102 | + "src/nns/stack.py", |
| 103 | + ' "Method 2 was removed because dimension reduction requires more than one original predictor.",\n', |
| 104 | + ' "Method 2 was removed because dimension reduction requires more than "\n' |
| 105 | + ' "one original predictor.",\n', |
| 106 | +) |
| 107 | +replace_once( |
| 108 | + "src/nns/stack.py", |
| 109 | + " return np.clip(np.asarray(code, dtype=np.float64), 1, n_classes).astype(np.int64).astype(np.float64)\n", |
| 110 | + " return (\n" |
| 111 | + " np.clip(np.asarray(code, dtype=np.float64), 1, n_classes)\n" |
| 112 | + " .astype(np.int64)\n" |
| 113 | + " .astype(np.float64)\n" |
| 114 | + " )\n", |
| 115 | +) |
| 116 | +replace_once( |
| 117 | + "src/nns/stack.py", |
| 118 | + "size = min(g.size - 1, max(1, int(round(holdout_size * g.size))))", |
| 119 | + "size = min(g.size - 1, max(1, round(holdout_size * g.size)))", |
| 120 | +) |
| 121 | +replace_once( |
| 122 | + "src/nns/stack.py", |
| 123 | + "size = max(1, min(n_obs - 1, int(round(holdout_size * n_obs))))", |
| 124 | + "size = max(1, min(n_obs - 1, round(holdout_size * n_obs)))", |
| 125 | +) |
| 126 | +replace_once( |
| 127 | + "src/nns/stack.py", |
| 128 | + " def coefficient_vector(design: NDArray[np.float64], response: NDArray[np.float64]) -> NDArray[np.float64]:\n", |
| 129 | + " def coefficient_vector(\n" |
| 130 | + " design: NDArray[np.float64], response: NDArray[np.float64]\n" |
| 131 | + " ) -> NDArray[np.float64]:\n", |
| 132 | +) |
| 133 | +replace_once( |
| 134 | + "src/nns/stack.py", |
| 135 | + "l_small = max(1, int(math.floor(math.sqrt(n_obs))))", |
| 136 | + "l_small = max(1, math.floor(math.sqrt(n_obs)))", |
| 137 | +) |
| 138 | + |
| 139 | +replace_once( |
| 140 | + "src/nns/var.py", |
| 141 | + " _METHOD1_GUARD_MESSAGE = (\n", |
| 142 | + " method1_guard_message = (\n", |
| 143 | +) |
| 144 | +replace_once( |
| 145 | + "src/nns/var.py", |
| 146 | + ''' def run_var_stack(stack_obj_fn: Any, stack_objective: str) -> dict[str, Any]: |
| 147 | + return nns_stack( |
| 148 | + lagged_iv, |
| 149 | + lagged_dv, |
| 150 | + ivs_test=ivs_test, |
| 151 | + obj_fn=cast(Any, stack_obj_fn), |
| 152 | + objective=cast(Any, stack_objective), |
| 153 | + folds=1, |
| 154 | + method=(1, 2), |
| 155 | + order=None, |
| 156 | + stack=True, |
| 157 | + dim_red_method=cast(Any, dim_red_threshold_method), |
| 158 | + ts_test=ts_test, |
| 159 | + ) |
| 160 | +
|
| 161 | + try: |
| 162 | + result = run_var_stack(var_obj_fn, objective_value) |
| 163 | + except ValueError as error: |
| 164 | + if not (use_default_objective and str(error) == _METHOD1_GUARD_MESSAGE): |
| 165 | + raise |
| 166 | + result = run_var_stack(mse_obj_fn, "min") |
| 167 | +''', |
| 168 | + ''' def run_var_stack( |
| 169 | + stack_obj_fn: Any, |
| 170 | + stack_objective: str, |
| 171 | + stack_iv: np.ndarray, |
| 172 | + stack_dv: np.ndarray, |
| 173 | + stack_test: np.ndarray, |
| 174 | + stack_ts_test: int, |
| 175 | + ) -> dict[str, Any]: |
| 176 | + return nns_stack( |
| 177 | + stack_iv, |
| 178 | + stack_dv, |
| 179 | + ivs_test=stack_test, |
| 180 | + obj_fn=cast(Any, stack_obj_fn), |
| 181 | + objective=cast(Any, stack_objective), |
| 182 | + folds=1, |
| 183 | + method=(1, 2), |
| 184 | + order=None, |
| 185 | + stack=True, |
| 186 | + dim_red_method=cast(Any, dim_red_threshold_method), |
| 187 | + ts_test=stack_ts_test, |
| 188 | + ) |
| 189 | +
|
| 190 | + try: |
| 191 | + result = run_var_stack( |
| 192 | + var_obj_fn, objective_value, lagged_iv, lagged_dv, ivs_test, ts_test |
| 193 | + ) |
| 194 | + except ValueError as error: |
| 195 | + if not (use_default_objective and str(error) == method1_guard_message): |
| 196 | + raise |
| 197 | + result = run_var_stack( |
| 198 | + mse_obj_fn, "min", lagged_iv, lagged_dv, ivs_test, ts_test |
| 199 | + ) |
| 200 | +''', |
| 201 | +) |
| 202 | + |
| 203 | +replace_once( |
| 204 | + "tests/parity/test_regression.py", |
| 205 | + 'with pytest.raises(ValueError, match="noise.reduction"):', |
| 206 | + 'with pytest.raises(ValueError, match=r"noise\\.reduction"):', |
| 207 | +) |
0 commit comments