✨ Find duplicates across directories.

The new CLI flag --across now lets users list duplicates only if they
are found in multiple of the given directories.

Closes gh-5.

This also removes support for arbitrary arguments to `git annex find`,
as we need to know which of the CLI arguments are directories to search
in. This might be re-introduced later.

Closes gh-4.

git_annex_duplicates.py

@@ -1,21 +1,23 @@
 import subprocess
 import collections
+import os.path
 import re
+
+from pathlib import Path
+
 import click
 
 
-@click.command()
-@click.option("--exclude", multiple=True, default=[])
-@click.argument("arguments", nargs=-1)
-def main(exclude, arguments):
+def collect_files(exclude, directories):
     # build command
-    find_cmd = ["git", "annex", "find", "--include", "*", "--format=${key}:${file}\n", *arguments]
+    find_cmd = ["git", "annex", "find", "--include", "*", "--format=${key}:${file}\n", *directories]
     for e in exclude:
         find_cmd += ["--exclude", e]
 
     # run
     result = subprocess.run(find_cmd, capture_output=True)
 
+    # collect output
     files = collections.defaultdict(list)
     for l in result.stdout.decode("utf-8").split("\n"):
         if l == "":
@@ -24,11 +26,51 @@ def main(exclude, arguments):
 
         files[key].append(file)
 
+    return files
+
+
+def print_files(fs):
+    for f in fs:
+        print(f)
+
+
+def resolve_parent(p):
+    return p.parent.resolve() / p.stem
+
+
+@click.command()
+@click.option("--exclude", multiple=True, default=[])
+@click.option("--across", is_flag=True, default=False)
+@click.argument("directories", nargs=-1, default=["."], type=click.Path(exists=True, file_okay=False, path_type=Path))
+def main(exclude, across, directories):
+
+    if across and len(directories) < 2:
+        raise click.UsageError("--across requires at least two directories")
+
+    directories = [d.resolve() for d in directories]  # for is_relative_to()
+
+    files = collect_files(exclude, directories)
+
     for k in files:
         if len(files[k]) > 1:
-            for f in files[k]:
-                print(f)
-            print()
+            if not across:
+                print_files(files[k])
+                print()
+            else:
+                # A file is a duplicate across the directories iff it's in more than one
+                # directory.
+                count = 0
+                found_across = False
+                for directory in directories:
+                    if any(resolve_parent(Path(f)).is_relative_to(directory) for f in files[k]):
+                       count += 1
+
+                    if count > 1:
+                        found_across = True
+                        break
+                if found_across:
+                    print_files(files[k])
+                    print()
 
 
 if __name__ == "__main__":