perf: load master data on install instead of every migrate

Move E-Taxes Item Group (~13k rows) and Classification code (~116 rows)
out of the `fixtures` hook. Re-syncing these on every `bench migrate` was
the dominant cost of the fixture sync step.

Add invoice_az.master_data_loader.load_master_data() that bulk-inserts
from the existing JSON files; it runs once via after_install and is
idempotent (skips rows that already exist by primary key) so it can be
re-invoked from the bench console to refresh master data.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

invoice_az/hooks.py

@@ -72,8 +72,10 @@ scheduler_events = {
 def after_install():
 	"""Run installation tasks"""
 	from invoice_az.auth import setup_token_renewal
+	from invoice_az.master_data_loader import load_master_data
 
 	setup_token_renewal()
+	load_master_data()
 
 
 def after_migrate():
@@ -84,16 +86,11 @@ def after_migrate():
 
 # Fixtures for master data
 # ------------------------
-fixtures = [
-	{
-		"doctype": "E-Taxes Item Group",
-		"filters": []
-	},
-	{
-		"doctype": "Classification code",
-		"filters": []
-	}
-]
+# Master data (E-Taxes Item Group ~13k rows, Classification code ~116 rows) is
+# loaded once on `after_install` via invoice_az.master_data_loader, not on every
+# `bench migrate`. To refresh from JSON on an existing site, call
+# `invoice_az.master_data_loader.load_master_data()` from `bench --site <site> console`.
+fixtures = []
 
 # Apps
 # ------------------

invoice_az/master_data_loader.py

@@ -0,0 +1,82 @@
+"""One-time master-data loader for E-Taxes reference doctypes.
+
+Replaces fixture sync for ~13k E-Taxes Item Group + ~116 Classification code
+rows. Loading via fixtures runs on every `bench migrate`; bulk SQL insert runs
+once on `after_install`. Idempotent: skips rows that already exist by primary
+key, so safe to re-run.
+"""
+
+import json
+import os
+
+import frappe
+
+_FIXTURES_DIR = os.path.join(os.path.dirname(__file__), "fixtures")
+
+_DATASETS = (
+    ("E-Taxes Item Group", "e_taxes_item_group.json"),
+    ("Classification code", "classification_code.json"),
+)
+
+
+def load_master_data():
+    """Bulk-load reference data on fresh installs."""
+    for doctype, filename in _DATASETS:
+        path = os.path.join(_FIXTURES_DIR, filename)
+        if not os.path.exists(path):
+            print(f"[invoice_az] master data: {filename} not found, skipping")
+            continue
+        try:
+            inserted = _bulk_insert_from_json(doctype, path)
+            print(f"[invoice_az] master data: {doctype}: inserted {inserted} rows")
+        except Exception as e:
+            print(f"[invoice_az] master data: failed to load {doctype}: {e}")
+            frappe.db.rollback()
+
+
+def _bulk_insert_from_json(doctype: str, path: str) -> int:
+    with open(path, encoding="utf-8") as f:
+        rows = json.load(f)
+    if not rows:
+        return 0
+
+    existing = {
+        r[0] for r in frappe.db.sql(f"SELECT name FROM `tab{doctype}`")
+    }
+
+    meta = frappe.get_meta(doctype)
+    field_names = {df.fieldname for df in meta.fields}
+
+    inserted = 0
+    chunk = []
+    CHUNK_SIZE = 500
+
+    for row in rows:
+        name = row.get("name")
+        if not name or name in existing:
+            continue
+        doc = frappe.new_doc(doctype)
+        doc.name = name
+        doc.flags.name_set = True
+        for k, v in row.items():
+            if k in ("doctype", "name", "modified", "docstatus"):
+                continue
+            if k in field_names:
+                doc.set(k, v)
+        chunk.append(doc)
+        existing.add(name)
+
+        if len(chunk) >= CHUNK_SIZE:
+            for d in chunk:
+                d.db_insert()
+            frappe.db.commit()
+            inserted += len(chunk)
+            chunk = []
+
+    if chunk:
+        for d in chunk:
+            d.db_insert()
+        frappe.db.commit()
+        inserted += len(chunk)
+
+    return inserted