From 75e47f10e52a2eb8d921c9bd0b0648378aa3e083 Mon Sep 17 00:00:00 2001
From: Federico Battini <fbattini1@gmail.com>
Date: Mon, 22 Dec 2025 14:22:20 +0100
Subject: [PATCH] Fix Windows path compatibility for data loading

  ---                                                                                                                                                                                                                                      ## Problem

  The `metoffice load` command failed on Windows with two issues:

  1. Incorrect path structure: The archive path was missing year/month/day
     subdirectories, causing 404 errors when downloading from HuggingFace.
     - Expected: `data/2023/12/01/2023-12-01-00.zarr.zip`
     - Actual: `data/2023-12-01-00.zarr.zip`

  2. Windows path separators: On Windows, `Path` objects use backslashes (`\`)
     which broke HuggingFace URLs when converted to strings.
     - Expected: `data/2023/12/01/...`
     - Actual: `data\2023\12\01\...`

  ## Changes

  - `main.py`: Added year/month/day subdirectories to archive_path (line 115)
  - `data_downloader.py`: Convert backslashes to forward slashes in download_from_hf()

  ## Testing

  - Tested on Windows 11
  - Successfully downloads and loads Met Office UK data
  - All 53 existing tests pass

  ---
---
 src/open_data_pvnet/main.py                  | 10 ++++++++--
 src/open_data_pvnet/utils/data_downloader.py |  1 +
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/src/open_data_pvnet/main.py b/src/open_data_pvnet/main.py
index f8d733e..c24fe85 100644
--- a/src/open_data_pvnet/main.py
+++ b/src/open_data_pvnet/main.py
@@ -107,12 +107,18 @@ def handle_load(provider: str, year: int, month: int, day: int, **kwargs):
     hour = kwargs.get("hour")
 
     # Base path for the data
-    base_path = Path("data") / str(year) / f"{month:02d}" / f"{day:02d}"
+    base_path = Path("data")
 
     try:
         if hour is not None:
             # Load specific hour
-            archive_path = base_path / f"{year}-{month:02d}-{day:02d}-{hour:02d}.zarr.zip"
+            archive_path = (
+                base_path
+                / str(year)
+                / f"{month:02d}"
+                / f"{day:02d}"
+                / f"{year}-{month:02d}-{day:02d}-{hour:02d}.zarr.zip"
+            )
             dataset = load_zarr_data(
                 archive_path,
                 chunks=chunks,
diff --git a/src/open_data_pvnet/utils/data_downloader.py b/src/open_data_pvnet/utils/data_downloader.py
index 9fe77f8..b023dbf 100644
--- a/src/open_data_pvnet/utils/data_downloader.py
+++ b/src/open_data_pvnet/utils/data_downloader.py
@@ -14,6 +14,7 @@
 
 def download_from_hf(repo_path: str, local_path: Path) -> None:
     """Download a file from HuggingFace."""
+    repo_path = repo_path.replace("\\", "/")
     logger.info(f"Downloading {repo_path} from Hugging Face...")
     local_path.parent.mkdir(parents=True, exist_ok=True)
     hf_hub_download(