From a571ac745d9a9644a060c46821042e9d5bb85509 Mon Sep 17 00:00:00 2001
From: Pablo Hernandez-Cerdan <pablo.hernandez.cerdan@outlook.com>
Date: Fri, 15 Dec 2023 18:41:16 +0100
Subject: [PATCH] Add merge_downloaded_parts script

md.ai splits dicoms into parts folders

This script merges all of them, optionally removing the zip files.
---
 mdai_utils/merge_downloaded_parts.py | 73 ++++++++++++++++++++++++++++
 1 file changed, 73 insertions(+)
 create mode 100644 mdai_utils/merge_downloaded_parts.py

diff --git a/mdai_utils/merge_downloaded_parts.py b/mdai_utils/merge_downloaded_parts.py
new file mode 100644
index 0000000..6b4a94b
--- /dev/null
+++ b/mdai_utils/merge_downloaded_parts.py
@@ -0,0 +1,73 @@
+import re
+import shutil
+from pathlib import Path
+
+
+def merge_downloaded_parts(part_folder, remove_zip_part_file=False):
+    """
+    md.ai splits the data into multiple parts (~50GB each). This function
+    merges all the parts into one directory. A simple mv does not work because
+    there are duplicated folders (with different content) in the parts.
+
+    part_folder is the path to the folder containing the parts of the data.
+    For example:
+    mdai_uab_project_7YNdkRbz_images_dataset_D_vVO1L2_2023-12-14-121547_part21of21
+
+    The output folder will be parallel to the part folder, without the _partXXofYY.
+    """
+
+    # Create a new directory with the dir string
+    part_folder_path = Path(part_folder)
+    out_folder_path = part_folder_path.parent
+    dir_string = part_folder_path.name
+    # Remove the part information from the dir string
+    re_pattern = re.compile("_part[0-9]+of[0-9]+")
+    dir_string = re.sub(re_pattern, "", dir_string)
+
+    # Create a new directory with the dir string
+    out_folder_path = part_folder_path.parent / dir_string
+    out_folder_path.mkdir(exist_ok=True, parents=True)
+
+    # Collect all parts folder, don't capture zip files
+    all_parts = out_folder_path.parent.glob(dir_string + "_part*")
+    all_parts = [part for part in all_parts if not part.name.endswith(".zip")]
+    all_parts = sorted(all_parts)
+    if len(all_parts) == 0:
+        raise ValueError(f"No part folders matching {out_folder_path}")
+
+    for part_folder in all_parts:
+        # Move all files from the part folder to the out folder.
+        print(f"Moving files from {part_folder}")
+        for study in part_folder.glob("*"):
+            for series in study.glob("*"):
+                destination = out_folder_path / study.name / series.name
+                shutil.move(str(series), str(destination))
+            assert (
+                len(list(study.glob("*"))) == 0
+            ), f"Study folder is not empty: {study}. Aborting, do it manually for this part folder."
+            shutil.rmtree(str(study))
+        assert (
+            len(list(part_folder.glob("*"))) == 0
+        ), f"Part folder is not empty: {part_folder}. Aborting, do it manually for this part folder."
+        shutil.rmtree(str(part_folder))
+        if remove_zip_part_file:
+            zip_file = part_folder.with_suffix(".zip")
+            zip_file.unlink()
+
+
+if __name__ == "__main__":
+    import argparse
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--part_folder", type=str, help="The path to one (any) part folder"
+    )
+    parser.add_argument(
+        "--remove_zip_files",
+        action="store_true",
+        help="Remove the zip file containing the part. Default: False",
+    )
+    args = parser.parse_args()
+    print(args)
+
+    merge_downloaded_parts(args.part_folder, args.remove_zip_files)