Pass information about the Job directories existence from Project to …

…Job.

This allows Job to avoid some `stat` calls which greatly improves performance on
networked file systems.

signac/job.py

@@ -249,7 +249,8 @@ class Job:
 
     Jobs can be opened by ``statepoint`` or ``id_``. If both values are
     provided, it is the user's responsibility to ensure that the values
-    correspond.
+    correspond. Set ``directory_known`` to ``True`` when the job directory
+    is known to exist - this skips some expensive isdir checks.
 
     Parameters
     ----------
@@ -259,6 +260,8 @@ class Job:
         State point for the job. (Default value = None)
     id_ : str, optional
         The job identifier. (Default value = None)
+    directory_known : bool, optional
+        Set to true when the job directory is known to exist. (Default value = False)
 
     """
 
@@ -275,10 +278,11 @@ class Job:
     KEY_DATA = "signac_data"
     "The job's datastore key."
 
-    def __init__(self, project, statepoint=None, id_=None):
+    def __init__(self, project, statepoint=None, id_=None, directory_known=False):
         self._project = project
         self._lock = RLock()
         self._initialize_lazy_properties()
+        self._directory_known = directory_known
 
         if statepoint is None and id_ is None:
             raise ValueError("Either statepoint or id_ must be provided.")
@@ -714,8 +718,13 @@ def init(self, force=False, validate_statepoint=True):
         with self._lock:
             try:
                 # Fast early exit when not validating.
-                if not validate_statepoint and os.path.isdir(self.path):
-                    return self
+                if not validate_statepoint:
+                    if self._directory_known:
+                        return self
+
+                    if os.path.isdir(self.path):
+                        self._directory_known = True
+                        return self
 
                 # Attempt early exit if the state point file exists and is valid.
                 try:
@@ -733,6 +742,8 @@ def init(self, force=False, validate_statepoint=True):
                         )
                         raise
 
+                    self._directory_known = True
+
                     # The state point save will not overwrite an existing file on
                     # disk unless force is True, so the subsequent load will catch
                     # when a preexisting invalid file was present.
@@ -806,6 +817,8 @@ def remove(self):
                     self._document = None
                 self._stores = None
 
+            self._directory_known = False
+
     def move(self, project):
         """Move this job to project.
 

signac/project.py

@@ -552,7 +552,7 @@ def open_job(self, statepoint=None, id=None):
             elif not self._contains_job_id(id):
                 # id does not exist in the project data space
                 raise KeyError(id)
-            return Job(project=self, id_=id)
+            return Job(project=self, id_=id, directory_known=True)
 
     def _job_dirs(self):
         """Generate ids of jobs in the workspace.
@@ -1631,7 +1631,7 @@ def get_job(cls, path=None):
         project = cls.get_project(os.path.join(job_path, os.pardir))
 
         # Return the matched job id from the found project
-        return Job(project=project, id_=job_id)
+        return Job(project=project, id_=job_id, directory_known=True)
 
     def __getstate__(self):
         state = dict(self.__dict__)
@@ -1688,7 +1688,9 @@ def __init__(self, project, ids):
         self._ids_iterator = iter(ids)
 
     def __next__(self):
-        return Job(project=self._project, id_=next(self._ids_iterator))
+        return Job(
+            project=self._project, id_=next(self._ids_iterator), directory_known=True
+        )
 
     def __iter__(self):
         return type(self)(self._project, self._ids)