TK-10467: bounded tranche driver (<=460/tranche, public-sample gate after each, halt on error)

e60cec41c14c210701bc6658903d37d6aaac964f · 2026-09-23 12:36:25 -0700 · Steve Abrams

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

Files touched

A run_tranches.sh

Diff

commit e60cec41c14c210701bc6658903d37d6aaac964f
Author: Steve Abrams <steve@designerwallcoverings.com>
Date:   Wed Sep 23 12:36:25 2026 -0700

    TK-10467: bounded tranche driver (<=460/tranche, public-sample gate after each, halt on error)
    
    Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
---
 run_tranches.sh | 49 +++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)

diff --git a/run_tranches.sh b/run_tranches.sh
new file mode 100755
index 0000000..24bdd61
--- /dev/null
+++ b/run_tranches.sh
@@ -0,0 +1,49 @@
+#!/usr/bin/env bash
+# TK-10467 — bounded tranche driver for the Steve-approved (2026-09-23) full-pool PJ hero write.
+# Tier-1 rows 26..2303 in tranches of <=460, then Tier-2 (32). After every tranche: sample 10
+# public-storefront PDPs (anonymous, not Admin API) and STOP on any error or any sample failure.
+# Requires CONFIRM_PJ_WRITE=1 and the lock at /tmp/pj-canary-TK10467.lock held by the caller.
+set -uo pipefail
+cd "$(dirname "$0")"
+LOG=data/tranche_run_$(date -u +%Y%m%dT%H%M%SZ).log
+: "${CONFIRM_PJ_WRITE:?set CONFIRM_PJ_WRITE=1}"
+[ -f /tmp/pj-canary-TK10467.lock ] || { echo "no lock — refusing"; exit 2; }
+
+sample_public() {  # $1=map $2=offset $3=count -> exit 1 on any non-real pos1
+  python3 - "$1" "$2" "$3" <<'EOF'
+import csv, json, sys, urllib.request, random
+m, off, n = sys.argv[1], int(sys.argv[2]), int(sys.argv[3])
+rows = list(csv.DictReader(open(m), delimiter="\t"))[off:off+n]
+random.seed(off); pick = random.sample(rows, min(10, len(rows)))
+bad = 0
+for r in pick:
+    tok = r["real_image"].split("?")[0].rsplit("/",1)[-1].rsplit(".",1)[0]
+    try:
+        d = json.load(urllib.request.urlopen(f"https://designerwallcoverings.com/products/{r['handle']}.json", timeout=30))
+        src = d["product"]["images"][0]["src"]
+        ok = tok in src and "Logo" not in src
+    except Exception as e:
+        ok = False; src = f"ERR {e}"
+    print(("  ok  " if ok else "  BAD ") + r["dw_sku"] + " " + src.rsplit("/",1)[-1][:60])
+    bad += (not ok)
+print(f"sample: {len(pick)-bad}/{len(pick)} real@pos1")
+sys.exit(1 if bad else 0)
+EOF
+}
+
+run_tranche() {  # $1=map $2=offset $3=limit $4=label
+  echo "=== $4: map=$1 offset=$2 limit=$3 $(date -u +%FT%TZ) ===" | tee -a "$LOG"
+  PJ_MAP="$1" python3 write_shopify_images.py --apply --offset="$2" --limit="$3" 2>&1 | tee -a "$LOG" | grep -E "^(\[[0-9]+\] (ERR|WARN|HTTP|STOP)|done )"
+  local summary; summary=$(tail -1 "$LOG")
+  echo "$summary" | grep -q " err=0 " || { echo "!!! $4 had errors — HALTING: $summary" | tee -a "$LOG"; exit 1; }
+  echo "--- public sample after $4 ---" | tee -a "$LOG"
+  sample_public "$1" "$2" "$3" 2>&1 | tee -a "$LOG" || { echo "!!! $4 public sample FAILED — HALTING" | tee -a "$LOG"; exit 1; }
+}
+
+run_tranche data/pj_writable.tsv   26 460 "T1-a"
+run_tranche data/pj_writable.tsv  486 460 "T1-b"
+run_tranche data/pj_writable.tsv  946 460 "T1-c"
+run_tranche data/pj_writable.tsv 1406 460 "T1-d"
+run_tranche data/pj_writable.tsv 1866 438 "T1-e"
+run_tranche data/pj_tier2_writable.tsv 0 32 "T2"
+echo "=== ALL TRANCHES COMPLETE $(date -u +%FT%TZ) ===" | tee -a "$LOG"

← 5cd75f3 TK-10467: writer gains PJ_MAP env override + --offset for bo · back to Pj Image Repair · auto-data-snapshot: 2026-09-23T12:42:16 (1 data files) — dat 795ecd0 →