# Prostatectomy split archive transfer guide

This is a new transfer guide. It does not replace or modify the source dataset's
original `prostatectomy/README.md`.

## Archive layout

All archives are uncompressed tar files. Every archive retains the top-level
`prostatectomy/` path, so extracting every part into the same parent directory
reconstructs the original dataset tree.

First batch:

- `prostatectomy_non_video.tar`: original `README.md`, `data/`, and `meta/`
- `prostatectomy_videos_chunk-000.tar`
- `prostatectomy_videos_chunk-001.tar`
- `prostatectomy_videos_chunk-002.tar`
- `prostatectomy_videos_chunk-003.tar`
- `prostatectomy_videos_chunk-004.tar`

Remaining video chunks to package after the first batch has been uploaded and
local space has been reclaimed:

- `chunk-005` through `chunk-010`

Files ending in `.partial` are still being written and must not be uploaded.
Only upload files whose names end exactly in `.tar`.

## Reconstruct on the destination machine

Choose the parent directory that should contain `prostatectomy/`:

```bash
DEST_PARENT=/path/to/cmr_surgical
mkdir -p "${DEST_PARENT}"

tar -xf prostatectomy_non_video.tar -C "${DEST_PARENT}"
for archive in prostatectomy_videos_chunk-*.tar; do
    tar -xf "${archive}" -C "${DEST_PARENT}"
done
```

This produces:

```text
${DEST_PARENT}/prostatectomy/
├── README.md
├── data/
├── meta/
└── videos/
    ├── chunk-000/
    ├── chunk-001/
    └── ...
```

Parts may be extracted in any order. Extracting later batches into the same
`DEST_PARENT` adds the remaining chunks without changing existing chunks.

## Basic verification

Check an archive before extracting:

```bash
tar -tf prostatectomy_non_video.tar >/dev/null
tar -tf prostatectomy_videos_chunk-000.tar >/dev/null
```

After every batch has been extracted, the complete dataset should contain video
directories `chunk-000` through `chunk-010`.