From 41ff08742cc3ac4da87c7ae6a6cff43da66fc359 Mon Sep 17 00:00:00 2001
From: DEls <davide.els1@outlook.com>
Date: Tue, 10 Nov 2026 10:12:19 +0200
Subject: [PATCH] fix(po-email): recognise a site named with an instance number

MEGCPO9566's subject reads "MEGCREQ11976 (Equestria 215)". A site name with
a bare instance number was not recognised as a site reference, so the PO
fell through the whole name chain and came back as "no MEGA site name found"
rather than naming the site it wanted.

A bracketed name with an instance number is now a site reference, and a
text whose bracketed names all resolve is not flagged at all.
---
 docs/n8n/po-email-intake.md     |  9 +++++++++
 scripts/po_email_intake.py      | 24 +++++++++++++++---------
 scripts/test_po_email_intake.py | 17 +++++++++++++++++
 3 files changed, 41 insertions(+), 9 deletions(-)

diff --git a/docs/n8n/po-email-intake.md b/docs/n8n/po-email-intake.md
index 544ecbb..fada147 100644
--- a/docs/n8n/po-email-intake.md
+++ b/docs/n8n/po-email-intake.md
@@ -234,6 +234,15 @@ bracket that is a code or a spec — `(C26/27)`, `(110mm PVC)` — is not a site
 trigger it. **Willow Park Ext 98** (`C25/40`) was added to the app on 2026-11-09 with its
 commercial details still blank, and MEGCPO9564 was moved onto it.
 
+A site name may carry an instance number instead of an "EXT" qualifier — MEGCPO9566's subject
+reads `MEGCREQ11976 (Equestria 215)`, and MEGA's own plant register calls that site *"Equestria
+215"* too (buildsite lists it as *Equestria*, code *Ext 215*). **Equestria 215** (`C25/31`) was
+added to the app on 2026-11-10 with its commercial details still blank.
+
+> **A message flagged for review has already been marked as seen.** Once the site it names has
+> been added, re-run the script with `--force` so the message is checked again — a plain re-run
+> skips it as already seen and the PO is never filed.
+
 **The project code on the PO's lines is deliberately not used.** It is a cost allocation, not a
 statement of where the work is going. MEGCPO9550 is a **Cedar Gate** delivery — its subject says
 `(Cedar Gate)` and its email subject says *"RE: Cedar Gate Paving & Kerbs PO"* — yet four of its
diff --git a/scripts/po_email_intake.py b/scripts/po_email_intake.py
index 4560f2c..4b1cb27 100755
--- a/scripts/po_email_intake.py
+++ b/scripts/po_email_intake.py
@@ -330,8 +330,9 @@ def site_vocabulary(sites: list[dict]) -> set[str]:
 def names_a_site(candidate: str, vocabulary: set[str]) -> bool:
     """Is this bracketed phrase MEGA writing the site, rather than a code?
 
-    "EXT98" is MEGA's own way of marking an instance of a site, and a phrase
-    whose words are all ours is a site name. "(DN150)", "(T16)" and
+    "EXT98" is MEGA's own way of marking an instance of a site, a name with an
+    instance number -- "(Equestria 215)" -- is one too, and a phrase whose
+    words are all ours is a site name. "(DN150)", "(T16)" and
     "(Union fitting)" are none of those, so they are not site names.
     """
     tokens = [t for t in re.split(r"[^A-Za-z0-9]+", candidate.lower()) if t]
@@ -340,6 +341,8 @@ def names_a_site(candidate: str, vocabulary: set[str]) -> bool:
     if any(re.fullmatch(r"ext\d+", t) for t in tokens):
         return True
     words = [t for t in tokens if len(t) >= 4 and t.isalpha()]
+    if any(re.fullmatch(r"\d{2,}", t) for t in tokens):
+        return bool(words)
     return bool(words) and all(word in vocabulary for word in words)
 
 
@@ -353,13 +356,16 @@ def unknown_site_name(text: str, sites: list[dict]) -> str | None:
     now left for a person instead.
     """
     vocabulary = site_vocabulary(sites)
-    for match in BRACKETED.finditer(text or ""):
-        candidate = match.group(1).strip()
-        if not names_a_site(candidate, vocabulary):
-            continue
-        if resolve_site_by_name(candidate, sites) is None:
-            return candidate
-    return None
+    candidates = [
+        match.group(1).strip()
+        for match in BRACKETED.finditer(text or "")
+        if names_a_site(match.group(1).strip(), vocabulary)
+    ]
+    if not candidates:
+        return None
+    if any(resolve_site_by_name(candidate, sites) for candidate in candidates):
+        return None
+    return candidates[0]
 
 
 def site_for_po(
diff --git a/scripts/test_po_email_intake.py b/scripts/test_po_email_intake.py
index a78bdfc..1cdcf43 100644
--- a/scripts/test_po_email_intake.py
+++ b/scripts/test_po_email_intake.py
@@ -250,6 +250,23 @@ class UnknownSiteName(unittest.TestCase):
     def test_a_number_in_brackets_is_not_a_site(self):
         self.assertIsNone(unknown_site_name("Call (012) 803 4456", SITES))
 
+    def test_a_site_named_with_an_instance_number_is_named(self):
+        # MEGCPO9566: "MEGCREQ11976 (Equestria 215)".
+        self.assertEqual(
+            unknown_site_name("MEGCREQ11976 (Equestria 215)", SITES),
+            "Equestria 215",
+        )
+
+    def test_a_site_named_with_an_instance_number_resolves_once_it_exists(self):
+        sites = SITES + [{"id": "equestria", "name": "Equestria 215", "code": "C25/31"}]
+        self.assertIsNone(unknown_site_name("MEGCREQ11976 (Equestria 215)", sites))
+
+    def test_a_known_site_in_the_same_text_settles_it(self):
+        # One bracketed name we know means the text names a site we have.
+        self.assertIsNone(
+            unknown_site_name("Tools (Nellmapius EXT25) (Grade 40)", SITES)
+        )
+
 
 if __name__ == "__main__":
     unittest.main()