From d7b18baf2e2e6dda107980c0750fc8ba01c73f92 Mon Sep 17 00:00:00 2001
From: Felix Delattre <felix@delattre.de>
Date: Thu, 11 Jun 2026 18:03:34 +0200
Subject: [PATCH] Improved single sites metrics calculation.

---
 5-metrics.py    | 15 ++++++++++++---
 run-pipeline.py | 13 ++++++-------
 2 files changed, 18 insertions(+), 10 deletions(-)

diff --git a/5-metrics.py b/5-metrics.py
index d4a833e..9658797 100644
--- a/5-metrics.py
+++ b/5-metrics.py
@@ -756,8 +756,6 @@ def main() -> None:
     # Export per-site data for the requested year
     year_sites = manifest["sites"].get(str(year), {})
     fusion_sites = {s: m for s, m in year_sites.items() if m["has_fusion"]}
-    if filter_site:
-        fusion_sites = {s: m for s, m in fusion_sites.items() if s == filter_site}
 
     print(f"Exporting {len(fusion_sites)} site(s) with fusion data for {year}")
     for site, meta in tqdm(fusion_sites.items(), desc="Sites"):
@@ -771,7 +769,18 @@ def main() -> None:
             print(f"  ✗ {site} — no fusion data found")
 
     manifest_path = out_base / "manifest.json"
-    manifest_path.write_text(json.dumps(manifest, separators=(",", ":")))
+    if filter_site and manifest_path.is_file():
+        # Merge: update only the filtered site(s); preserve all other entries.
+        existing: dict = json.loads(manifest_path.read_text())
+        for year_key, year_sites_new in manifest["sites"].items():
+            existing.setdefault("sites", {}).setdefault(year_key, {}).update(
+                year_sites_new
+            )
+        all_years = sorted(set(existing.get("years", [])) | set(manifest["years"]))
+        existing["years"] = all_years
+        manifest_path.write_text(json.dumps(existing, separators=(",", ":")))
+    else:
+        manifest_path.write_text(json.dumps(manifest, separators=(",", ":")))
     print(f"Manifest written → {manifest_path}")
 
 
diff --git a/run-pipeline.py b/run-pipeline.py
index 3c18dc9..2067396 100644
--- a/run-pipeline.py
+++ b/run-pipeline.py
@@ -1,11 +1,9 @@
 """Pipeline wrapper: run steps 1 → 2 → 3 → 4 → 5.
 
 Steps 1 and 2 run once for the whole year (skipped when their output already
-exists).  Steps 3 and 4 run site-by-site for every PASS site from
-``data/phenocam_screening/{year}.json``; a site is skipped when
-``data/metrics/{year}/{site}/metrics.json`` already exists.  Step 5 always
-runs once at the end without ``--site`` so that ``manifest.json`` is written
-with all processed sites (not just the last one).
+exists).  Steps 3–5 run site-by-site for every PASS site from
+``data/phenocam_screening/{year}.json``; a site is skipped entirely when
+``data/metrics/{year}/{site}/metrics.json`` already exists.
 
 Any failure stops the run immediately.  Fix the issue and re-run — completed
 steps and sites are skipped automatically.
@@ -13,7 +11,7 @@ steps and sites are skipped automatically.
 CLI:
 
 - ``--evaluation-year`` (default 2025)
-- ``--site``            single site to run steps 3–4 for (default: all PASS sites)
+- ``--site``            single site to run steps 3–5 for (default: all PASS sites)
 """
 
 from __future__ import annotations
@@ -38,6 +36,7 @@ GLOBAL_STEPS: list[tuple[str, Path]] = [
 PER_SITE_STEPS = [
     "3-sentinel-data.py",
     "4-fusion.py",
+    "5-metrics.py",
 ]
 
 
@@ -101,7 +100,7 @@ def main(argv: list[str] | None = None) -> int:
                     f"[pipeline] {script} failed (exit {rc}); cannot continue"
                 )
 
-    # --- Per-site steps (steps 3–5) ---
+    # --- Per-site steps (steps 3, 4, 5) ---
     sites = _load_pass_sites(year)
     if args.site:
         if args.site not in sites: