Deprecate optimization.backend

src/everest/config/optimization_config.py

@@ -1,14 +1,16 @@
+import logging
 from typing import Any
 
 from pydantic import BaseModel, Field, model_validator
 
 from everest.config.cvar_config import CVaRConfig
 from everest.optimizer.utils import get_ropt_plugin_manager
+from everest.strings import EVEREST
 
 
 class OptimizationConfig(BaseModel, extra="forbid"):  # type: ignore
     algorithm: str | None = Field(
-        default="default",
+        default="optpp_q_newton",
         description="""Algorithm used by Everest.  Defaults to
 optpp_q_newton, a quasi-Newton algorithm in Dakota's OPT PP library.
 """,
@@ -33,8 +35,8 @@ class OptimizationConfig(BaseModel, extra="forbid"):  # type: ignore
 (From the Dakota Manual.)""",
     )
     backend: str | None = Field(
-        default="dakota",
-        description="""The optimization backend used.  Defaults to "dakota".
+        default=None,
+        description="""(deprecated) The optimization backend used.".
 
 Currently, backends are included to use Dakota or SciPy ("dakota" and "scipy").
 The Dakota backend is the default, and can be assumed to be installed. The SciPy
@@ -199,9 +201,17 @@ class OptimizationConfig(BaseModel, extra="forbid"):  # type: ignore
     @model_validator(mode="after")
     def validate_backend_and_algorithm(self):  # pylint: disable=E0213
         method = "default" if self.algorithm is None else self.algorithm
-        backend = "dakota" if self.backend is None else self.backend
-        if not get_ropt_plugin_manager().is_supported(
-            "optimizer", f"{backend}/{method}"
-        ):
-            raise ValueError(f"Optimizer algorithm '{backend}/{method}' not found")
+        if not get_ropt_plugin_manager().is_supported("optimizer", method):
+            raise ValueError(f"Optimizer algorithm '{method}' not found")
+
+        if self.backend is not None:
+            message = (
+                "optimization.backend is deprecated. "
+                "The correct backend will be inferred by the algorithm. "
+                "If several backends have an algorithm named A and you want to pick "
+                "a specific backend B, put B/A in optimization.algorithm."
+            )
+            print(message)
+            logging.getLogger(EVEREST).warning(message)
+
         return self

src/everest/optimizer/everest2ropt.py

@@ -258,9 +258,8 @@ def _parse_optimization(
     ropt_optimizer = ropt_config["optimizer"]
     ropt_gradient = ropt_config["gradient"]
 
-    backend = ever_opt.backend or "dakota"
-    algorithm = ever_opt.algorithm or "default"
-    ropt_optimizer["method"] = f"{backend}/{algorithm}"
+    algorithm = ever_opt.algorithm or "optpp_q_newton"
+    ropt_optimizer["method"] = f"{algorithm}"
 
     alg_max_iter = ever_opt.max_iterations
     if alg_max_iter:

tests/everest/test_config_validation.py

@@ -544,7 +544,7 @@ def test_that_missing_optimization_algorithm_errors():
     with pytest.raises(ValueError) as e:
         EverestConfig.with_defaults(**{"optimization": {"algorithm": "ddlygldt"}})
 
-    assert has_error(e.value, match="Optimizer algorithm 'dakota/ddlygldt' not found")
+    assert has_error(e.value, match="Optimizer algorithm 'ddlygldt' not found")
 
 
 @pytest.mark.parametrize(

tests/everest/test_ropt_initialization.py

@@ -132,7 +132,7 @@ def test_everest2ropt_controls_optimizer_setting():
     config = EverestConfig.load_file(config)
     ropt_config = everest2ropt(config)
     assert len(ropt_config.realizations.names) == 15
-    assert ropt_config.optimizer.method == "dakota/conmin_mfd"
+    assert ropt_config.optimizer.method == "conmin_mfd"
     assert ropt_config.gradient.number_of_perturbations == 20
     assert ropt_config.realizations.names == tuple(range(15))
 
@@ -244,4 +244,4 @@ def test_everest2ropt_no_algorithm_name(copy_test_data_to_tmp):
 
     config.optimization.algorithm = None
     ropt_config = everest2ropt(config)
-    assert ropt_config.optimizer.method == "dakota/default"
+    assert ropt_config.optimizer.method == "optpp_q_newton"