Merge pull request #63 from bcdev/forman-55-lock_file_issue

Report that target parent must exist

Author: thomasstorm

CHANGES.md

@@ -1,5 +1,11 @@
 ## Version 0.4.1 (in development)
 
+### Fixes
+
+* If the target _parent_ directory did not exist, an exception was raised 
+  reporting that the lock file to be written does not exist. Changed this to
+  report that the target parent directory does not exist. [#55]
+
 ### Enhancements
 
 * Added missing documentation of the `append_step` setting in the 

docs/config.md

@@ -146,7 +146,7 @@ Variable metadata.
 ## `target_dir`
 
 Type _string_.
-The URI or local path of the target Zarr dataset. Must be a directory.
+The URI or local path of the target Zarr dataset. Must specify a directory whose parent directory must exist.
 
 ## `target_storage_options`
 

docs/guide.md

@@ -16,7 +16,8 @@ zappend(os.listdir("inputs"), target_dir="output/mycube.zarr")
 
 Both invocations will create the Zarr dataset `output/mycube.zarr` by concatenating
 the "slice" datasets provided in the `inputs` directory along their `time` dimension. 
-Both the CLI command and the Python function can be run without any further 
+`target_dir` must specify a directory for the Zarr dataset. (Its parent directory must 
+exist.) Both the CLI command and the Python function can be run without any further 
 configuration provided the paths of the target dataset and the source slice datasets 
 are given. The target dataset path must point to a directory that will contain a Zarr 
 group to be created and updated. The slice dataset paths may be provided as Zarr as 
@@ -433,6 +434,7 @@ You can disable transaction management by specifying
 The `target_dir` setting is mandatory. If it is not specified in the configuration,
 it must be passed either as `--target` or `-t` option to the `zappend` command or as 
 `target_dir` keyword argument when using the `zappend` Python function.
+Note, the parent directory of `target_dir` must already exist.
 
 If the target path is given for another filesystem, additional storage options may be 
 passed using the optional `target_storage_options` setting. 

tests/test_api.py

@@ -55,6 +55,21 @@ def test_some_slices_local(self):
             for slice_dir in slices:
                 shutil.rmtree(slice_dir, ignore_errors=True)
 
+    def test_some_slices_local_output_to_non_existing_dir(self):
+        target_dir = "non_existent_dir/target.zarr"
+        slices = [
+            "slice-1.zarr",
+            "slice-2.zarr",
+            "slice-3.zarr",
+        ]
+        for uri in slices:
+            make_test_dataset(uri=uri)
+        with pytest.raises(
+            FileNotFoundError,
+            match="\\ATarget parent directory does not exist: .*/non_existent_dir\\Z",
+        ):
+            zappend(slices, target_dir=target_dir)
+
     def test_some_slices_with_class_slice_source(self):
         target_dir = "memory://target.zarr"
         slices = [make_test_dataset(), make_test_dataset(), make_test_dataset()]

zappend/config/schema.py

@@ -523,7 +523,7 @@
         target_dir={
             "description": (
                 "The URI or local path of the target Zarr dataset."
-                " Must be a directory."
+                " Must specify a directory whose parent directory must exist."
             ),
             "type": "string",
             "minLength": 1,

zappend/fsutil/transaction.py

@@ -115,6 +115,12 @@ def __enter__(self):
             )
         self._entered_ctx = True
 
+        if not self.target_dir.parent.exists():
+            raise FileNotFoundError(
+                f"Target parent directory does not exist:"
+                f" {self.target_dir.parent.path}"
+            )
+
         lock_file = self._lock_file
         if lock_file.exists():
             raise OSError(f"Target is locked: {lock_file.uri}")