harden(deals): Cody gate — parseSize picks the EARLIEST-in-text sqft match, not by magnitude precedence, so a body 'N msf' comp can't override the title's real 'Nk sf' (branch-order bug: million scanned whole string first)

2930a2089739a902abc971d1a3102f64c876bd55 · 2026-08-06 07:48:56 -0700 · Steve

Files touched

M scripts/lib/deal-parse.mjs
M test/deals/parse.test.mjs

Diff

commit 2930a2089739a902abc971d1a3102f64c876bd55
Author: Steve <steve@designerwallcoverings.com>
Date:   Thu Aug 6 07:48:56 2026 -0700

    harden(deals): Cody gate — parseSize picks the EARLIEST-in-text sqft match, not by magnitude precedence, so a body 'N msf' comp can't override the title's real 'Nk sf' (branch-order bug: million scanned whole string first)
---
 scripts/lib/deal-parse.mjs | 19 +++++++++++++------
 test/deals/parse.test.mjs  |  3 +++
 2 files changed, 16 insertions(+), 6 deletions(-)

diff --git a/scripts/lib/deal-parse.mjs b/scripts/lib/deal-parse.mjs
index c06cb28e..add84472 100644
--- a/scripts/lib/deal-parse.mjs
+++ b/scripts/lib/deal-parse.mjs
@@ -59,12 +59,19 @@ export function parseSize(text) {
   // Square footage, MAGNITUDE-AWARE. CRE headlines write the asset size as "104k sf" (thousand) or
   // "1.27 msf" (million square feet); the old plain "N sf" regex couldn't match those (the number isn't
   // directly followed by "sf"), so it SKIPPED the title's real size and grabbed a stray small "N sf" from
-  // the body — a 104,000 sf building rendered as "26 SF". Try million, then thousand, then plain: the
-  // magnitude notations live in the title (front of the text) so matching them here wins over a body leak.
-  let sqft = null, m;
-  if ((m = text.match(/([\d,.]+)\s?(?:msf|m\s?sf|million\s?(?:sf|square[ -]feet))/i))) sqft = Math.round(parseFloat(m[1].replace(/,/g, '')) * 1e6);
-  else if ((m = text.match(/([\d,.]+)k\s?(?:sf|sq\.?\s?ft|square[ -]feet|square[ -]foot)/i))) sqft = Math.round(parseFloat(m[1].replace(/,/g, '')) * 1e3);
-  else if ((m = text.match(/([\d,]+)\s?(?:sf|sq\.?\s?ft|square[ -]feet|square[ -]foot)/i))) sqft = +m[1].replace(/,/g, '');
+  // the body — a 104,000 sf building rendered as "26 SF". We compute all three candidate matches and pick
+  // the one that occurs EARLIEST in the text (by .index), NOT by magnitude precedence: the title sits at
+  // the front of `hay`, so its size wins by position. Ordering by magnitude instead would let a BODY
+  // mention of a larger comp ("adjacent to a 2.5 msf campus") override the title's real "104k sf".
+  const mMil = text.match(/([\d,.]+)\s?(?:msf|m\s?sf|million\s?(?:sf|square[ -]feet))/i);
+  const mK = text.match(/([\d,.]+)k\s?(?:sf|sq\.?\s?ft|square[ -]feet|square[ -]foot)/i);
+  const mPl = text.match(/([\d,]+)\s?(?:sf|sq\.?\s?ft|square[ -]feet|square[ -]foot)/i);
+  const cands = [
+    mMil && { idx: mMil.index, val: Math.round(parseFloat(mMil[1].replace(/,/g, '')) * 1e6) },
+    mK && { idx: mK.index, val: Math.round(parseFloat(mK[1].replace(/,/g, '')) * 1e3) },
+    mPl && { idx: mPl.index, val: +mPl[1].replace(/,/g, '') },
+  ].filter(Boolean).sort((a, b) => a.idx - b.idx);
+  const sqft = cands.length ? cands[0].val : null;
   const parts = [];
   if (units) parts.push(`${units[1]} units`);
   if (sqft) parts.push(`${sqft.toLocaleString()} SF`);
diff --git a/test/deals/parse.test.mjs b/test/deals/parse.test.mjs
index 8c8c156c..7887b569 100644
--- a/test/deals/parse.test.mjs
+++ b/test/deals/parse.test.mjs
@@ -95,6 +95,9 @@ test('parseSize: magnitude-aware (k = thousand, msf = million) and title-first-w
   // the title's "104k sf" wins over a stray small "26 sf" leaked later in the body
   assert.equal(parseSize('SFF Buys 104k sf office. The lobby has 26 sf of signage.').sqft, 104000);
   assert.equal(parseSize('GO builds a 250,000 square feet facility').sqft, 250000);
+  // a BODY mention of a LARGER comp must NOT override the title's k-sf (earliest-in-text wins, not magnitude)
+  assert.equal(parseSize('104k sf Office Sells. Property is adjacent to 2.5 msf distribution hub.').sqft, 104000);
+  assert.equal(parseSize('400k sf campus is part of a 3.2 million square feet master plan').sqft, 400000);
 });
 
 // ── parseOccupancy / parseYearBuilt ───────────────────────────────────────────

← 28068930 fix(deals): magnitude-aware parseSize — parse '104k sf'/'1.2 · back to Rentv 2026 · auto-save: 2026-08-06T07:49:47 (7 files) — data/deals-regist fa57d846 →