Heavy rewrite to use Fabric's context managers as intended, but requiring a bugfix.

md5_to_vault.py

@@ -37,9 +37,21 @@ import pprint
 import time
 import click
 import textwrap
+from shlex import quote
 from fabric import Connection
 
 
+def escape_path(path):
+    """ Explicitly escape parentheses. This is required to work around a bug in Fabric's
+    Invoke module. See my question on Stackoverflow:
+    https://stackoverflow.com/q/63225018/607587
+    The recommended workaround, until Fabric fixes the bug, is to just "manually escape
+    the parentheses"
+
+    """
+    escaped_path = path.replace('(', '\(').replace(')', '\)')
+    return escaped_path
+
 @dataclass
 class Node:
     """A directory tree node"""
@@ -83,11 +95,10 @@ def send_checksum(node, remote_login, src_path):
     """
     # Check if there are any files in the node
     with Connection(remote_login) as c:
-        files = c.run(
-            rf"cd {node.src}; find -maxdepth 1 -type f -printf '%f\n'",
-            echo=True
-        )
-        files = files.stdout.strip()
+        with c.cd(escape_path(node.src)):
+            result = c.run(r"nice find -maxdepth 1 -type f -printf '%f\n'", echo=True)
+
+    files = result.stdout.strip()
     node.count = len(files.splitlines())
 
     print(f"Node:{node.src}, file count:{node.count}")
@@ -101,16 +112,18 @@ def send_checksum(node, remote_login, src_path):
         else:
             filename = node.src.replace(src_path+'/', '').replace('/', '_')
 
-        output = subprocess.run(
-            f"ssh {remote_login} 'cd {node.src}; nice md5sum *'"
-            f"| cat > {node.dest}/{filename}.md5",
-            shell=True,
-            check=True
-        )
-        print("stdout:", output.stdout)
-        print("stderr:", output.stderr)
+        with Connection(remote_login) as c:
+            with c.cd(escape_path(node.src)):
+                # checksum all files including hidden files
+                result = c.run("nice md5sum {*,.[^.],.??*}", echo=True)
+        md5s = result.stdout
+
+        md5_filename = pathlib.Path(node.dest) / (filename + '.md5')
+        with open(md5_filename, 'w') as f:
+            f.write(md5s)
+        print(f"Checksummed {node.count} files {node.src} -> {md5_filename}")
+
         # os.chmod(f"{node.dest}/{filename}.md5", 0o550)
-        print(f"Checksummed {node.count} files {node.src} -> {node.dest}")
 
     node.processed = True
 
@@ -155,7 +168,7 @@ def main(
     flag.
 
     """
-    assert 5 <= len(experiment_name) <= 6
+    assert 4 <= len(experiment_name) <= 6
     if pickle_filename is None:
         pickle_filename = experiment_name + "md5_state.pickle"
 
@@ -224,7 +237,7 @@ def main(
     else:
         # Get the directory tree from the remote server as a list.
         with Connection(remote_login) as c:
-            result = c.run(f"find {src_path} -type d")
+            result = c.run(f"nice find {quote(src_path)} -type d")
         remote_dirs = result.stdout.strip().splitlines()
 
         # Create a tree data structure that represents both source and