FEA parallel_config context manager to allow more fine-grained control (

#1392)

CHANGES.rst

@@ -40,6 +40,12 @@ In development
   tracebacks and more efficient running time.
   https://github.com/joblib/joblib/pull/1393
 
+- Add the `parallel_config` context manager to allow for more fine-grained
+  control over the backend configuration. It should be used in place of the
+  `parallel_backend` context manager. In particular, it has the advantage
+  of not requiring to set a specific backend in the context manager.
+  https://github.com/joblib/joblib/pull/1392
+
 Release 1.2.0
 -------------
 

joblib/__init__.py

@@ -122,6 +122,7 @@
 from .parallel import cpu_count
 from .parallel import register_parallel_backend
 from .parallel import parallel_backend
+from .parallel import parallel_config
 from .parallel import effective_n_jobs
 from ._cloudpickle_wrapper import wrap_non_picklable_objects
 
@@ -130,7 +131,7 @@
            'load', 'Parallel', 'delayed', 'cpu_count', 'effective_n_jobs',
            'register_parallel_backend', 'parallel_backend',
            'register_store_backend', 'register_compressor',
-           'wrap_non_picklable_objects']
+           'wrap_non_picklable_objects', 'parallel_config']
 
 
 # Workaround issue discovered in intel-openmp 2019.5:

joblib/_utils.py

@@ -1,6 +1,7 @@
 # Adapted from https://stackoverflow.com/a/9558001/2536294
 
 import ast
+from dataclasses import dataclass
 import operator as op
 
 # supported operators
@@ -42,3 +43,12 @@ def eval_(node):
         return operators[type(node.op)](eval_(node.operand))
     else:
         raise TypeError(node)
+
+
+@dataclass(frozen=True)
+class _Sentinel:
+    """A sentinel to mark a parameter as not explicitly set"""
+    default_value: object
+
+    def __repr__(self):
+        return f"default({self.default_value!r})"