@dforge-core/dforge-mcp 0.2.44 → 0.2.46-next

package/skills/dforge-1c-import/scripts/lib/seedstate.py

@@ -0,0 +1,488 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Rebuild a balance register's state from its journal, and check it against 1C's totals.
+
+A dForge `balance` register keeps a running total in a STATE entity and its movements in a
+JOURNAL. A port loads the movements - they are 1C's rows - but 1C has no state table to copy:
+it keeps per-period totals of its own (`_AccumRgT`) and sums on every read. The platform
+does not seed a state from a journal either: posting moves both together, and nothing here
+is posted. So this does it, as one idempotent statement per register:
+
+    state := SUM(resource * journalDirection) GROUP BY dimensions, the non-zero tuples only
+
+A zero tuple is left out because a zero row and no row answer `getBalance` alike, and 1C's
+own Остатки does not return one either. The state's key is minted from the tuple itself
+(md5 -> uuid), so a re-run lands on the same keys rather than reshuffling them.
+
+    python tools/1c.py seedstate run [module ...]      # rebuild every journalled balance state
+    python tools/1c.py seedstate check [module ...]    # state against its journal, no writes
+    python tools/1c.py seedstate verify [module ...]   # state against 1C's current totals
+    python tools/1c.py seedstate verify --asof 2026-08 # as-of read against 1C's month row
+    python tools/1c.py seedstate probe [module ...]    # the platform's getBalance against 1C
+
+Which registers: every `kind: "balance"` register in the tenant that names both an entity
+and a journal with a `journalDirection` - read from `"dForge".register`, so nothing here
+names a register. `run` refuses a journal with a row whose direction is not +1/-1: the
+loader stamps it (`migrate.record_kind`), and a NULL would drop the row from the sum.
+
+`verify` reads the journal's 1C table from the column map, finds its totals table through
+DBNames (the `AccumRgT` entry beside the `AccumRg` one), and reads the CURRENT totals - the
+rows at the table's last `_Period` - by a seek on the clustered index, summing on this side.
+No GROUP BY runs on the server, whose tempdb has no room for one.
+
+The writes run with `session_replication_role = 'replica'`, as the loader's do: the state
+table carries the audit trigger, and 5,000 rows of "inserted" history describe nothing.
+"""
+from __future__ import print_function
+
+import argparse
+import codecs
+import collections
+import datetime
+import decimal
+import json
+import os
+import sys
+
+import attrs
+import migrate
+
+Register = collections.namedtuple("Register", [
+	"module", "code", "state", "journal", "journal_entity", "direction", "date",
+	"dimensions", "resources", "key", "key_type", "resource_names"])
+
+
+def q(name):
+	return '"%s"' % name.replace('"', '""')
+
+
+def registers(cur, modules):
+	cur.execute("""
+		SELECT m.module_cd, r.register_cd, se.schema_name, se.db_object,
+		       je.schema_name, je.db_object, je.entity_cd,
+		       r.journal_direction, r.journal_date, r.dimensions, r.resources
+		FROM "dForge".register r
+		JOIN "dForge".module m ON m.module_id = r.module_id
+		JOIN "dForge".entity se ON se.entity_id = r.entity_id
+		JOIN "dForge".entity je ON je.entity_id = r.journal_entity_id
+		WHERE r.kind = 'balance' AND r.journal_direction IS NOT NULL
+		ORDER BY 1, 2""")
+	found = []
+	for (module, code, s_schema, s_table, j_schema, j_table, j_entity,
+	     direction, date, dims, resources) in cur.fetchall():
+		if modules and module not in modules:
+			continue
+		dims = dims if isinstance(dims, list) else json.loads(dims)
+		resources = resources if isinstance(resources, dict) else json.loads(resources)
+		fields = [spec["field"] for _, spec in sorted(resources.items())]
+		names = [name for name, _ in sorted(resources.items())]
+		cur.execute("""
+			SELECT a.attname, format_type(a.atttypid, a.atttypmod)
+			FROM pg_index i
+			JOIN pg_attribute a ON a.attrelid = i.indrelid AND a.attnum = ANY(i.indkey)
+			WHERE i.indrelid = %s::regclass AND i.indisprimary""",
+			("%s.%s" % (q(s_schema), q(s_table)),))
+		key = cur.fetchall()
+		if len(key) != 1:
+			raise SystemExit("seedstate: %s.%s - the state %s.%s is not keyed by one column"
+			                 % (module, code, s_schema, s_table))
+		found.append(Register(module, code, "%s.%s" % (q(s_schema), q(s_table)),
+		                      "%s.%s" % (q(j_schema), q(j_table)), j_entity,
+		                      direction, date, dims, fields, key[0][0], key[0][1], names))
+	return found
+
+
+def summed(reg):
+	"""The journal grouped by tuple: dims, then one signed sum per resource."""
+	dims = ", ".join(q(d) for d in reg.dimensions)
+	sums = ", ".join("COALESCE(SUM(%s * %s), 0) AS %s" % (q(f), q(reg.direction), q(f))
+	                 for f in reg.resources)
+	return "SELECT %s, %s FROM %s GROUP BY %s" % (dims, sums, reg.journal, dims)
+
+
+def nonzero(alias, reg):
+	return " OR ".join("%s.%s <> 0" % (alias, q(f)) for f in reg.resources)
+
+
+def unstamped(cur, reg):
+	cur.execute("SELECT count(*) FILTER (WHERE %s IS NULL OR %s NOT IN (1, -1)),"
+	            " count(*) FILTER (WHERE %s IS NULL), count(*) FROM %s"
+	            % (q(reg.direction), q(reg.direction), q(reg.date), reg.journal))
+	return cur.fetchone()
+
+
+def seed(pg, reg):
+	cur = pg.cursor()
+	bad, undated, rows = unstamped(cur, reg)
+	if bad:
+		raise SystemExit("seedstate: %s.%s - %d of %d journal rows carry no direction (+1/-1);"
+		                 " reload the journal with its direction transform first"
+		                 % (reg.module, reg.code, bad, rows))
+	if reg.key_type == "uuid":
+		# The tuple is the key: md5 over the quoted values, so NULL and '' stay apart.
+		parts = " || '|' || ".join("quote_nullable(g.%s)" % q(d) for d in reg.dimensions)
+		key = "md5('%s|' || %s)::uuid" % (reg.state.replace("'", "''"), parts)
+	elif reg.key_type == "bigint":
+		key = '"dForge".next_id()'
+	else:
+		raise SystemExit("seedstate: %s.%s - a state key of type %s cannot be minted here"
+		                 % (reg.module, reg.code, reg.key_type))
+	columns = [reg.key] + reg.dimensions + reg.resources
+	cur.execute("SET LOCAL session_replication_role = 'replica'")
+	cur.execute("DELETE FROM %s" % reg.state)
+	cur.execute("INSERT INTO %s (%s) SELECT %s, %s FROM (%s) g WHERE %s"
+	            % (reg.state, ", ".join(q(c) for c in columns), key,
+	               ", ".join("g.%s" % q(c) for c in reg.dimensions + reg.resources),
+	               summed(reg), nonzero("g", reg)))
+	written = cur.rowcount
+	pg.commit()
+	return rows, written, undated
+
+
+def drift(cur, reg):
+	"""Rows the state and the journal's sum do not share, counted from both sides.
+
+	A set difference rather than a FULL JOIN: the dimensions hold NULLs, which only
+	IS NOT DISTINCT FROM matches, and Postgres refuses that as a FULL JOIN condition.
+	EXCEPT treats two NULLs as the same value and compares numerics by value.
+	"""
+	columns = ", ".join(q(c) for c in reg.dimensions + reg.resources)
+	journal = "SELECT %s FROM (%s) g WHERE %s" % (columns, summed(reg), nonzero("g", reg))
+	# A zero state row is left out as the seed leaves it out: posting keeps the row it created
+	# when a tuple returns to zero, and a zero row and no row answer `getBalance` alike.
+	state = "SELECT %s FROM %s s WHERE %s" % (columns, reg.state, nonzero("s", reg))
+	cur.execute("SELECT (SELECT count(*) FROM (%s EXCEPT ALL %s) a)"
+	            " + (SELECT count(*) FROM (%s EXCEPT ALL %s) b)"
+	            % (journal, state, state, journal))
+	return cur.fetchone()[0]
+
+
+# --------------------------------------------------------------------------- 1C side
+
+def totals_table(table):
+	"""`_AccumRg16999` -> `_AccumRgT17010`: the same object's AccumRgT entry in DBNames."""
+	with codecs.open(os.path.join(attrs.OUT, "dbnames.json"), "r", "utf-8") as f:
+		entries = json.load(f)["entries"]
+	guid = next((e["guid"] for e in entries if e["physical"] == table), None)
+	found = [e["physical"] for e in entries
+	         if e["guid"] == guid and e["type"] == "AccumRgT"]
+	return found[0] if found else None
+
+
+def one_c_balance(mssql, pg, reg, mapping, source, period=None):
+	"""(totals table, period, {tuple: resources}) - 1C's totals at a period, non-zero only.
+
+	No period means the table's last one: 5999-11-01, where 1C keeps its CURRENT totals. A
+	month row `P` holds the balance at the START of P - measured, not assumed: it equals the
+	platform's as-of read at the end of the month before, on every month compared.
+	"""
+	mapped = mapping.get(reg.module, {}).get(reg.journal_entity)
+	if not mapped:
+		raise SystemExit("seedstate: %s.%s - the column map has no source for %s"
+		                 % (reg.module, reg.code, reg.journal_entity))
+	table = totals_table(mapped["table"])
+	if not table:
+		raise SystemExit("seedstate: %s has no totals table in DBNames" % mapped["table"])
+	schema, name = [p.strip('"') for p in reg.journal.split(".")]
+	types = dict((c[0], c[1]) for c in migrate.table_columns(pg, schema, name))
+	held = source[mapped["table"]]["columns"]
+	transforms = mapped.get("transforms") or {}
+
+	columns, converters = [], []
+	for ours in reg.dimensions + reg.resources:
+		physical = mapped["columns"][ours]
+		columns.append(physical)
+		if ours in transforms:
+			converters.append(migrate.TRANSFORMS[transforms[ours]])
+		else:
+			converters.append(migrate.converter(types[ours], held[physical]["type"]))
+
+	if period is None:
+		mssql.execute("SELECT MAX(_Period) FROM [%s]" % table)
+		period = mssql.fetchone()[0]
+	mssql.execute("SELECT %s FROM [%s] WHERE _Period = ?"
+	              % (", ".join("[%s]" % c for c in columns), table), period)
+	width = len(reg.dimensions)
+	balance = collections.defaultdict(lambda: [decimal.Decimal(0)] * len(reg.resources))
+	for row in mssql.fetchall():
+		values = [normal(f(v), types[c])
+		          for f, v, c in zip(converters, row, reg.dimensions + reg.resources)]
+		held_now = balance[tuple(values[:width])]
+		for i, value in enumerate(values[width:]):
+			held_now[i] += decimal.Decimal(value or 0)
+	return table, period, dict((k, v) for k, v in balance.items() if any(v))
+
+
+def normal(value, pg_type):
+	"""One value as psycopg2 returns it for a column of that type."""
+	if value is None:
+		return None
+	if pg_type == "date" and isinstance(value, datetime.datetime):
+		return value.date()
+	if pg_type in ("numeric", "integer", "bigint") and not isinstance(value, bool):
+		return decimal.Decimal(str(value))
+	return value
+
+
+def state_balance(pg, reg):
+	cur = pg.cursor()
+	cur.execute("SELECT %s, %s FROM %s"
+	            % (", ".join(q(d) for d in reg.dimensions),
+	               ", ".join(q(f) for f in reg.resources), reg.state))
+	width = len(reg.dimensions)
+	return dict((tuple(normal(v, None) if not isinstance(v, decimal.Decimal) else v
+	                   for v in row[:width]),
+	             [decimal.Decimal(v or 0) for v in row[width:]])
+	            for row in cur.fetchall())
+
+
+def as_of_balance(pg, reg, until):
+	"""What `getBalance(asOf)` answers for the day before `until`: the state minus every
+	journal row dated on or after it (RegisterReader.GetBalance), per tuple, non-zero only."""
+	dims = ", ".join(q(d) for d in reg.dimensions)
+	cur = pg.cursor()
+	cur.execute("""SELECT %s, %s FROM (
+		SELECT %s, %s FROM %s
+		UNION ALL
+		SELECT %s, %s FROM %s WHERE %s >= %%s) u GROUP BY %s""" % (
+		dims, ", ".join("SUM(%s)" % q(f) for f in reg.resources),
+		dims, ", ".join(q(f) for f in reg.resources), reg.state,
+		dims, ", ".join("-%s * %s" % (q(f), q(reg.direction)) for f in reg.resources),
+		reg.journal, q(reg.date), dims), (until,))
+	width = len(reg.dimensions)
+	out = {}
+	for row in cur.fetchall():
+		values = [decimal.Decimal(v or 0) for v in row[width:]]
+		if any(values):
+			out[tuple(row[:width])] = values
+	return out
+
+
+def verify(pg, reg, mapping, source, mssql, month=None):
+	"""The state against 1C's current totals, or - given a month - the as-of read at its
+	end against 1C's totals row for the month after."""
+	if month:
+		year, number = [int(p) for p in month.split("-")]
+		until = datetime.date(year + number // 12, number % 12 + 1, 1)
+		period = datetime.datetime(until.year + migrate.YEAR_OFFSET, until.month, 1)
+		table, period, theirs = one_c_balance(mssql, pg, reg, mapping, source, period)
+		ours = as_of_balance(pg, reg, until)
+	else:
+		table, period, theirs = one_c_balance(mssql, pg, reg, mapping, source)
+		ours = state_balance(pg, reg)
+	both = set(theirs) & set(ours)
+	differ = [k for k in both if theirs[k] != ours[k]]
+	only_1c = set(theirs) - set(ours)
+	only_ours = set(ours) - set(theirs)
+	print("%-28s %-15s %s  1C %6d  state %6d  equal %6d  differ %d  only 1C %d  only state %d"
+	      % ("%s.%s" % (reg.module, reg.code), table, period.strftime("%Y-%m-%d"),
+	         len(theirs), len(ours), len(both) - len(differ), len(differ),
+	         len(only_1c), len(only_ours)))
+	for k in sorted(differ, key=str)[:5]:
+		print("    differ  %s  1C %s  state %s" % (k, theirs[k], ours[k]))
+	for k in sorted(only_1c, key=str)[:5]:
+		print("    only 1C %s  %s" % (k, theirs[k]))
+	for k in sorted(only_ours, key=str)[:5]:
+		print("    only us %s  %s" % (k, ours[k]))
+	return not (differ or only_1c or only_ours)
+
+
+# --------------------------------------------------------------------------- the platform
+
+# `verify` checks the state by a formula written to imitate RegisterReader. `probe` asks
+# RegisterReader itself, through `modules/register-probe` (one batch action calling
+# getBalance per request), and compares each answer with 1C's totals summed client-side.
+
+PROBE_MONTHS = ("2019-03", "2021-06", "2023-12", "2025-12", "2026-07", "2026-08")
+PROBE_BATCH = 100
+
+
+def month_end(month):
+	"""'2026-07' -> (2026-07-31, the 1C totals period holding the balance at that end)."""
+	year, number = [int(p) for p in month.split("-")]
+	start = datetime.date(year + number // 12, number % 12 + 1, 1)
+	return (start - datetime.timedelta(days=1),
+	        datetime.datetime(start.year + migrate.YEAR_OFFSET, start.month, 1))
+
+
+def sample(reg, tuples, size):
+	"""Tuples worth asking about, deterministically: the largest, the negative, the ones
+	holding a NULL dimension, and an even spread over the rest."""
+	ordered = sorted(tuples, key=str)
+	picked = []
+
+	def take(items, n):
+		for t in items:
+			if n <= 0:
+				break
+			if t not in picked:
+				picked.append(t)
+				n -= 1
+
+	take(sorted(ordered, key=lambda t: -abs(tuples[t][0])), size // 4)
+	take([t for t in ordered if any(v < 0 for v in tuples[t])], size // 4)
+	take([t for t in ordered if None in t], size // 4)
+	step = max(1, len(ordered) // max(1, size - len(picked)))
+	take(ordered[::step], size - len(picked))
+	return picked
+
+
+def jsonable(value):
+	if isinstance(value, decimal.Decimal):
+		return float(value)
+	if isinstance(value, (datetime.date, datetime.datetime)):
+		return value.isoformat()
+	return value
+
+
+def requests_for(reg, picked):
+	"""Dimension filters: each full tuple (a NULL dimension left out, since a read cannot
+	ask for NULL by equality), the same tuple with each dimension in turn left out, the
+	first dimension alone, and nothing at all - the register's grand total."""
+	seen, out = set(), []
+
+	def add(dims):
+		key = tuple(sorted((k, str(v)) for k, v in dims.items()))
+		if key not in seen:
+			seen.add(key)
+			out.append(dims)
+
+	add({})
+	for t in picked:
+		full = dict((d, v) for d, v in zip(reg.dimensions, t) if v is not None)
+		add(full)
+		for d in list(full):
+			add(dict((k, v) for k, v in full.items() if k != d))
+		first = reg.dimensions[0]
+		if first in full:
+			add({first: full[first]})
+	return out
+
+
+def expected(reg, theirs, dims):
+	"""1C's answer to a dimension filter: every tuple it matches, summed."""
+	index = [(reg.dimensions.index(d), v) for d, v in dims.items()]
+	total = [decimal.Decimal(0)] * len(reg.resources)
+	for t, values in theirs.items():
+		if all(t[i] == v for i, v in index):
+			total = [a + b for a, b in zip(total, values)]
+	return total
+
+
+def probe(pg, reg, mapping, source, mssql, size, months):
+	import posting
+	cur = pg.cursor()
+	cur.execute("""SELECT a.action_id FROM "dForge".action a
+		JOIN "dForge".module m ON m.module_id = a.module_id
+		WHERE m.module_cd = 'register-probe' AND a.action_cd = 'balances'""")
+	found = cur.fetchone()
+	if not found:
+		raise SystemExit("seedstate probe: install modules/register-probe first")
+	action = found[0]
+	cur.execute("""SELECT f.folder_id FROM "dForge".folder f
+		JOIN "dForge".module m ON m.module_id = f.module_id
+		WHERE m.module_cd = 'production' ORDER BY 1 LIMIT 1""")
+	folder = str(cur.fetchone()[0])
+	# Batch mode runs only over records, so one movement is handed in and not touched.
+	cur.execute("SELECT output_shop_id::text FROM production.output_shop LIMIT 1")
+	record = {"output_shop_id": cur.fetchone()[0]}
+	jwt = posting.token()
+
+	dates = [(None, None)] + [month_end(m) for m in months]
+	theirs = {}
+	for as_of, period in dates:
+		theirs[as_of] = one_c_balance(mssql, pg, reg, mapping, source, period)[2]
+	pool = {}
+	for got in theirs.values():
+		for t, v in got.items():
+			pool.setdefault(t, v)
+	asked = requests_for(reg, sample(reg, pool, size))
+
+	register = "%s.%s" % (reg.module, reg.code)
+	compared = equal = nonzero = 0
+	failures = []
+	for as_of, _ in dates:
+		batch = [{"register": register, "dims": dict((k, jsonable(v)) for k, v in d.items()),
+		          "asOf": as_of.isoformat() if as_of else None} for d in asked]
+		answers = []
+		for i in range(0, len(batch), PROBE_BATCH):
+			reply = posting.call("action.execute", {
+				"actionId": action, "entity": "production.output_shop", "keys": [record],
+				"params": {"requests": json.dumps(batch[i:i + PROBE_BATCH])}}, folder, jwt)
+			messages = (reply.get("result") or {}).get("messages") or []
+			if len(messages) != 1 or messages[0].get("level") != "info":
+				raise SystemExit("seedstate probe: %s" % json.dumps(reply)[:2000])
+			answers.extend(json.loads(messages[0]["text"]))
+		for dims, answer in zip(asked, answers):
+			compared += 1
+			want = expected(reg, theirs[as_of], dims)
+			nonzero += any(want)
+			if "error" in answer:
+				failures.append((as_of, dims, want, answer["error"]))
+				continue
+			got = [decimal.Decimal(str(answer["ok"].get(n, 0))) for n in reg.resource_names]
+			# The action hands numbers back through JS doubles; compare to their precision.
+			if all(abs(g - w) <= abs(w) * decimal.Decimal("1e-12") + decimal.Decimal("1e-6")
+			       for g, w in zip(got, want)):
+				equal += 1
+			else:
+				failures.append((as_of, dims, want, got))
+	# 1C's empty date is held as 0001-01-01 (`since_beginning`), and the DSL turns that
+	# string into DateTime.MinValue, which Npgsql binds as -infinity: such a tuple reads as
+	# zero. The platform's defect, not the seed's - counted apart so it cannot hide another.
+	empty_date = [f for f in failures if datetime.date(1, 1, 1) in f[1].values()]
+	print("%-28s %3d filters x %d dates  %5d compared (%d non-zero)  %5d equal  %d differ%s"
+	      % (register, len(asked), len(dates), compared, nonzero, equal, len(failures),
+	         " (%d on a 0001-01-01 dimension)" % len(empty_date) if empty_date else ""))
+	for as_of, dims, want, got in [f for f in failures if f not in empty_date][:10]:
+		print("    %s %s  1C %s  platform %s" % (as_of or "now", dims, want, got))
+	return not failures
+
+
+def main():
+	migrate.load_env()
+	ap = argparse.ArgumentParser(description=__doc__.strip().splitlines()[0])
+	ap.add_argument("command", choices=("run", "check", "verify", "probe"))
+	ap.add_argument("module", nargs="*", help="limit to these modules")
+	ap.add_argument("--asof", metavar="YYYY-MM",
+	                help="verify: the as-of read at the end of this month instead of now")
+	ap.add_argument("--sample", type=int, default=24, metavar="N",
+	                help="probe: tuples per register (each also asked with dimensions left out)")
+	args = ap.parse_args()
+
+	pg = migrate.pg_connect()
+	found = registers(pg.cursor(), set(args.module))
+	if not found:
+		print("no journalled balance register%s" % (" in " + ", ".join(args.module)
+		                                             if args.module else ""))
+		return 0
+	clean = True
+	if args.command == "run":
+		print("%-28s %10s %8s %s" % ("register", "journal", "state", ""))
+		for reg in found:
+			rows, written, undated = seed(pg, reg)
+			print("%-28s %10d %8d %s" % ("%s.%s" % (reg.module, reg.code), rows, written,
+			                             "%d undated" % undated if undated else ""))
+	elif args.command == "check":
+		cur = pg.cursor()
+		for reg in found:
+			bad, undated, rows = unstamped(cur, reg)
+			off = drift(cur, reg)
+			clean = clean and not (bad or off)
+			print("%-28s journal %8d  unstamped %d  undated %d  state off by %d tuple(s)"
+			      % ("%s.%s" % (reg.module, reg.code), rows, bad, undated, off))
+	else:
+		mapping, source = migrate.column_map(), migrate.source_schema()
+		mssql = migrate.mssql_connect().cursor()
+		for reg in found:
+			if args.command == "probe":
+				clean = probe(pg, reg, mapping, source, mssql, args.sample,
+				              [args.asof] if args.asof else PROBE_MONTHS) and clean
+			else:
+				clean = verify(pg, reg, mapping, source, mssql, args.asof) and clean
+	return 0 if clean else 1
+
+
+if __name__ == "__main__":
+	sys.exit(main())

package/skills/dforge-1c-import/scripts/lib/sizing.py

@@ -0,0 +1,184 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+How big is the rest of the port?
+
+Counts fields, tabular sections and references across the client-specific objects that
+hold data, so the remaining effort can be estimated from the same numbers the transport
+contour was estimated from rather than guessed at.
+
+    python tools/sizing.py
+"""
+from __future__ import print_function
+
+import codecs
+import collections
+import json
+import os
+import re
+
+import attrs
+import binding
+
+contours = binding.contours()
+import unpack
+
+LOWER_PREFIX = re.compile(r"^([a-zа-яёіїє]{2,6})(?=[A-ZА-Я])")
+
+# The data model is ported for every object in scope, so "done" no longer means the
+# object - it means its logic. `contours.DONE` is the set whose logic was looked at while
+# the transport contour was designed.
+DONE = contours.DONE
+
+
+_INDEX = None
+
+
+def code_index():
+	"""out/code/index.json, written once by extract_code.py. Falls back to unpacking on
+	demand if the code has not been extracted yet."""
+	global _INDEX
+	if _INDEX is None:
+		path = os.path.join(attrs.OUT, "code", "index.json")
+		if os.path.isfile(path):
+			with codecs.open(path, "r", "utf-8") as f:
+				_INDEX = json.load(f)
+		else:
+			_INDEX = []
+	return _INDEX
+
+
+def code_lines(guid):
+	"""Non-blank, non-comment lines of the object's own modules (not its forms)."""
+	index = code_index()
+	if index:
+		return sum(r["lines"] for r in index
+		           if r.get("guid") == guid and r.get("entry") == "text")
+	total = 0
+	for _, blob in unpack.containers_of(guid):
+		for name, content in unpack.entries(blob):
+			if name != "text":
+				continue
+			text = unpack.as_text(content) or ""
+			total += sum(1 for line in text.splitlines()
+			             if line.strip() and not line.strip().startswith("//"))
+	return total
+
+
+def code_breakdown():
+	"""Whose code is in the infobase, and is it business logic or UI."""
+	index = code_index()
+	if not index:
+		return
+	owner = collections.Counter()
+	kinds = collections.Counter()
+	for r in index:
+		if "lines" not in r:
+			continue
+		m = LOWER_PREFIX.match(r["object"] or "")
+		tag = m.group(1) if m else None
+		if tag in binding.client_prefixes():
+			bucket = "client"
+			kinds["UI (form modules)" if r["entry"] == "module"
+			      else "business logic"] += r["lines"]
+		elif tag in binding.toolkit_prefixes():
+			bucket = "third-party dev toolkit"
+		else:
+			bucket = "vendor (UTP)"
+		owner[bucket] += r["lines"]
+
+	print("\nextracted code, by owner")
+	print("-" * 46)
+	for bucket, n in owner.most_common():
+		print("   %-26s %9d lines" % (bucket, n))
+	print("\nthe client's own code splits into")
+	print("-" * 46)
+	for kind, n in kinds.most_common():
+		print("   %-26s %9d lines" % (kind, n))
+
+
+def main():
+	objects = attrs.load_json("objects.json")
+	types = attrs.load_json("typemap.json")
+	rows = attrs.load_json("report.json")
+	by_name = {o["name"]: o for o in objects if o.get("name")}
+
+	def is_custom(name):
+		m = LOWER_PREFIX.match(name or "")
+		return bool(m and m.group(1) in binding.client_prefixes() + binding.toolkit_prefixes())
+
+	scope = [r for r in rows
+	         if is_custom(r["name"]) and ((r["rows"] or 0) or r["childRows"])]
+
+	totals = collections.defaultdict(
+		lambda: {"objects": 0, "fields": 0, "sections": 0, "refs": 0, "done": 0,
+		         "code": 0})
+	grand = {"fields": 0, "sections": 0, "refs": 0, "remaining_fields": 0,
+	         "code": 0, "remaining_code": 0}
+
+	for row in scope:
+		name = row["name"]
+		obj = by_name.get(name)
+		if not obj:
+			continue
+		_, fields = attrs.fields_of(name, objects, types)
+		fields = fields or []
+		sections = len({f["section"] for f in fields if f["section"]})
+		refs = sum(1 for f in fields if "->" in f["type"])
+		code = code_lines(obj["guid"])
+		bucket = totals[row["family"]]
+		bucket["objects"] += 1
+		bucket["fields"] += len(fields)
+		bucket["sections"] += sections
+		bucket["refs"] += refs
+		bucket["code"] += code
+		grand["fields"] += len(fields)
+		grand["sections"] += sections
+		grand["refs"] += refs
+		grand["code"] += code
+		if name in DONE:
+			bucket["done"] += 1
+		else:
+			grand["remaining_fields"] += len(fields)
+			grand["remaining_code"] += code
+
+	print("%-12s %8s %6s %8s %9s %6s %9s" %
+	      ("family", "objects", "done", "fields", "sections", "refs", "code loc"))
+	print("-" * 68)
+	for family, b in sorted(totals.items(), key=lambda kv: -kv[1]["fields"]):
+		print("%-12s %8d %6d %8d %9d %6d %9d"
+		      % (family, b["objects"], b["done"], b["fields"], b["sections"],
+		         b["refs"], b["code"]))
+	print("-" * 68)
+	print("%-12s %8d %6d %8d %9d %6d %9d"
+	      % ("total", sum(b["objects"] for b in totals.values()),
+	         sum(b["done"] for b in totals.values()),
+	         grand["fields"], grand["sections"], grand["refs"], grand["code"]))
+	print("\nthe data model is generated for all of these; what remains is the logic:")
+	print("%d lines of 1C object-module code outside the contour already studied"
+	      % grand["remaining_code"])
+	by_module()
+	code_breakdown()
+
+
+def by_module():
+	"""How much 1C business logic sits behind each generated dForge module."""
+	objects = attrs.load_json("objects.json")
+	guid_of = {}
+	for obj in objects:
+		if obj.get("name"):
+			guid_of.setdefault(obj["name"], obj["guid"])
+
+	print("\n1C object-module code, by the module it was assigned to")
+	print("-" * 46)
+	totals = []
+	for module, contour in contours.CONTOURS.items():
+		lines = sum(code_lines(guid_of[name])
+		            for name in contours.names(contour) if name in guid_of)
+		totals.append((module, lines))
+	for module, lines in sorted(totals, key=lambda kv: -kv[1]):
+		print("   %-26s %9d lines" % (module, lines))
+
+
+if __name__ == "__main__":
+	main()