Update basic typing (#362)

.pylintrc

@@ -3,7 +3,7 @@ extension-pkg-whitelist=numpy,nose,nose.tools,numpy.testing, pathlib.PurePath, p
 
 [MESSAGES CONTROL]
 # disabled messages
-disable=invalid-name,missing-docstring,too-few-public-methods, protected-access, import-error, no-self-use, fixme, no-else-return, no-member
+disable=invalid-name,missing-docstring,too-few-public-methods, protected-access, import-error, no-self-use, fixme, no-else-return, no-member, useless-import-alias
 
 [TYPECHECK]
 ignored-modules = numpy, numpy.testing

nevergrad/benchmark/exporttable.py

@@ -1,12 +1,19 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+#
+# This source code is licensed under the MIT license found in the
+# LICENSE file in the root directory of this source tree.
+#
 # Utils for exporting a data table in latex.
 # Used in competence maps.
+import typing as t
 
-def remove_parens(data):
+
+def remove_parens(data: t.List[t.List[str]]) -> t.List[t.List[str]]:
     # If data[i][j] contains a "(", we keep only the part before that "(".
     return [[d[:d.index("(")] if "(" in d else d for d in datarow] for datarow in data]
-
 
-def export_table(filename, rows, cols, data):
+
+def export_table(filename: str, rows: t.List[t.Any], cols: t.List[t.Any], data: t.List[t.List[str]]) -> None:
     """Exports data in filename with rows and cols as described.
     More precisely, rows specifies the row names, cols specifies the col names,
     and data[i][j] corresponds to the data in row rows[i] and col cols[j].
@@ -19,7 +26,7 @@ def export_table(filename, rows, cols, data):
     rows = [str(r) for r in rows]
     cols = [str(r) for r in cols]
     # Latex syntax.
-    data = [[d.replace("%", "\%").replace("_", "") for d in datarow] for datarow in data]
+    data = [[d.replace("%", r"\%").replace("_", "") for d in datarow] for datarow in data]
     data = remove_parens(data)
     print("filename=", filename)
     print("rows=", rows)
@@ -39,14 +46,14 @@ def export_table(filename, rows, cols, data):
         f.write("\\lccode`7=`7\n")
         f.write("\\lccode`8=`8\n")
         f.write("\\lccode`9=`9\n")
-        f.write("\\newcolumntype{P}[1]{>{\hspace{0pt}}p{#1}}\n")
+        f.write(r"\\newcolumntype{P}[1]{>{\hspace{0pt}}p{#1}}\n")
         f.write("\\begin{document}\n")
         f.write("\\scriptsize\n")
         f.write("\\renewcommand{\\arraystretch}{1.5}\n")
         f.write("\\sloppy\n")
-        p = str(1./(2+len(cols)))
+        p = str(1.0 / (2 + len(cols)))
         # f.write("\\begin{landscape}\n")
-        f.write("\\begin{tabular}{|P{" + p +"\\textwidth}|" + ("P{" + p + "\\textwidth}|") * len(cols) + "}\n")
+        f.write("\\begin{tabular}{|P{" + p + "\\textwidth}|" + ("P{" + p + "\\textwidth}|") * len(cols) + "}\n")
         f.write("\\hline\n")
         f.write(" & " + "&".join(cols) + "\\\\\n")
         f.write("\\hline\n")
@@ -62,4 +69,3 @@ def export_table(filename, rows, cols, data):
         f.write("\\end{tabular}\n")
         # f.write("\\end{landscape}\n")
         f.write("\\end{document}\n")
-    pass

nevergrad/benchmark/xpbase.py

@@ -217,7 +217,7 @@ def _run_with_error(self, callbacks: Optional[Dict[str, base._OptimCallBack]] =
             # Note: when resuming a job (if optimizer is not None), seeding is pointless (reproducibility is lost)
             np.random.seed(self.seed)  # seeds both functions and instrumentation (for which random state init is lazy)
             random.seed(self.seed)
-            torch.manual_seed(self.seed)
+            torch.manual_seed(self.seed)  # type: ignore
         # optimizer instantiation can be slow and is done only here to make xp iterators very fast
         if self._optimizer is None:
             self._optimizer = self.optimsettings.instantiate(instrumentation=instrumentation)

nevergrad/examples/powersystem.py

@@ -1,12 +1,17 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+#
+# This source code is licensed under the MIT license found in the
+# LICENSE file in the root directory of this source tree.
+#
 # run this with:
-# echo 'import nevergrad.examples.powersystem' | python 
+# echo 'import nevergrad.examples.powersystem' | python
 # or (e.g. MacOS):
-# echo 'import nevergrad.examples.powersystem' | pythonw 
+# echo 'import nevergrad.examples.powersystem' | pythonw
 
+import nevergrad as ng
 from nevergrad.functions.powersystems.core import PowerSystem
-from nevergrad.optimization import optimizerlib
 
-budget=3500
+budget = 3500
 width = 6
 depth = 6
 num_dams = 6
@@ -23,12 +28,11 @@
 #     num_years: int = 1,  # Number of years.
 #     failure_cost: float = 500.,  # Cost of not satisfying the demand. Equivalent to an expensive infinite capacity thermal plant.
 
-                    power_system_loss = PowerSystem(num_dams=num_dams, depth=depth, width=width, year_to_day_ratio=year_to_day_ratio, back_to_normal=back_to_normal, num_thermal_plants=num_thermal_plants, constant_to_year_ratio=constant_to_year_ratio) 
-                    optimizer = optimizerlib.SplitOptimizer9(instrumentation=power_system_loss.dimension, budget=budget, num_workers=10) 
-                    optimizer.minimize(power_system_loss)
-                    power_system_loss(optimizer.provide_recommendation().data)
-                    power_system_loss.make_plots(f"ps_{num_dams}dams_{depth}_{width}_ytdr{year_to_day_ratio}_btn{back_to_normal}_num_thermal_plants{num_thermal_plants}_ctyr{constant_to_year_ratio}_budget{budget}.png")
-
-
-
-
+power_system_loss = PowerSystem(num_dams=num_dams, depth=depth, width=width, year_to_day_ratio=year_to_day_ratio,
+                                back_to_normal=back_to_normal, num_thermal_plants=num_thermal_plants,
+                                constant_to_year_ratio=constant_to_year_ratio)
+optimizer = ng.optimizers.SplitOptimizer(instrumentation=power_system_loss.dimension, budget=budget, num_workers=10)
+optimizer.minimize(power_system_loss)
+power_system_loss(optimizer.provide_recommendation().data)
+power_system_loss.make_plots(f"ps_{num_dams}dams_{depth}_{width}_ytdr{year_to_day_ratio}_btn{back_to_normal}"
+                             f"_num_thermal_plants{num_thermal_plants}_ctyr{constant_to_year_ratio}_budget{budget}.png")

nevergrad/functions/multiobjective/__init__.py

@@ -3,4 +3,4 @@
 # This source code is licensed under the MIT license found in the
 # LICENSE file in the root directory of this source tree.
 
-from .core import MultiobjectiveFunction
+from .core import MultiobjectiveFunction as MultiobjectiveFunction

nevergrad/functions/multiobjective/core.py

@@ -30,7 +30,7 @@ class MultiobjectiveFunction:
     - The minimum value obtained for this objective function is -h,
       where h is the hypervolume of the Pareto front obtained, given upper_bounds as a reference point.
     - The callable keeps track of the pareto_front (see attribute paretor_front) and is therefor stateful.
-      For this reasonm it cannot be distributed. A user can however call the multiobjective_function 
+      For this reason it cannot be distributed. A user can however call the multiobjective_function
       remotely, and aggregate locally. This is what happens in the "minimize" method of optimizers.
     """
 
@@ -47,7 +47,7 @@ def compute_aggregate_loss(self, losses: ArrayLike, *args: Any, **kwargs: Any) -
         """
         # We compute the hypervolume
         if (losses - self._upper_bounds > 0).any():
-            return np.max(losses - self._upper_bounds)
+            return np.max(losses - self._upper_bounds)  # type: ignore
         arr_losses = np.minimum(np.array(losses, copy=False), self._upper_bounds)
         new_volume: float = self._hypervolume.compute([y for _, y in self._points] + [arr_losses])
         if new_volume > self._best_volume:  # This point is good! Let us give him a great mono-fitness value.
@@ -60,7 +60,7 @@ def compute_aggregate_loss(self, losses: ArrayLike, *args: Any, **kwargs: Any) -
         else:
             # Now we compute for each axis
             # First we prune.
-            self.pareto_front
+            self.pareto_front  # pylint: disable=pointless-statement
             distance_to_pareto = float("Inf")
             for _, stored_losses in self._points:
                 if (stored_losses <= arr_losses).all():