Make the legacy import safe to resume, and aimable without editing it

Three defects broke the loader's own documented resume guarantee. All three
were silent: nothing raised, the run finished cheerfully, and fewer rows
arrived than anyone counted.

THE PAGE SIZE WAS CLAMPED, SO THE 409-RESOLVE SEARCHED THE FIRST 100 ROWS

_upsert asked for ?per_page=10000 and read one response. The server clamps any
page size to MAX_PAGE_SIZE, which is 100 - not an error, not a warning, just
100 rows. Anything past that resolved to None, so the id never reached the
crosswalk and the run quietly dropped a model's link on every machine that used
it, or an application's installs.

The target already holds more than 100 models and more than 100 applications,
so this fired on any re-run or resume - exactly the case the loader documents
as safe.

_resolve_existing now tries the exact natural-key filter the API has exposed
all along (?modelnumber= on models, ?appname= on applications, ?vendor= on
vendors) and then pages properly, bounded. An endpoint without such a filter
ignores the unknown parameter and the paged walk finds the row.

stage_metrology repeated the same mistake in its own way: it skipped 409s
entirely, so a resumed run left the tool out of the crosswalk and the Controls
edge unmade - metrology PCs driving nothing. stage_printers already had the
resolve branch; that stage was written later and missed it.

THE CROSSWALK WAS NOT BOUND TO THE DATABASE IT DESCRIBED

idmap.json maps this site's legacy ids to assetids in ONE database, and those
assetids mean nothing anywhere else. Keeping it across a drop-and-rebuild left
warranties and printer assignments attaching to whatever asset now holds that
id: wrong data, no error. The runbook's two "drop and rebuild" steps never
mentioned the file.

The loader now writes a uuid marker into the TARGET's own settings table, so it
dies with the data, and stamps it into the crosswalk. A crosswalk built against
a different database is refused before a row is written. The database name
alone would not do - a rebuild keeps the name. A file written before this guard
is adopted rather than refused, so no existing crosswalk is stranded.

Also adds --idmap, so one checkout can drive more than one target.

THE SOURCE DATABASES ARE NAMED, NOT HARDCODED

'shopdb_src' appeared at 29 call sites, so aiming the loader at the prodscratch
restore meant editing run.py. Those names are now LOGICAL and resolved in
Source.db() from SOURCE_DB_MAIN / SOURCE_DB_USB / SOURCE_DB_EMPLOYEES, with
host, port, user and password overridable too. The call sites are untouched -
resolving centrally avoids missing one, and the literals now read as
documentation of which source a stage draws from. The run prints every source
it redirected, so "it imported nothing" can be told apart from "it read the
wrong database".

This is what stood between the loader and the prodscratch run.

TWO SMALLER ONES

_source_password() raised IndexError rather than its intended message whenever
the DATABASE_URL user is not root, which it will not be on a real server.

main() returned 1 on failure into a caller that discarded it. A run that logged
200 endpoint errors exited 0 and read as a success in a deploy log; it now
exits non-zero.

TESTS AND DOCS

The loader had no tests at all. Sixteen now cover the resolve, the crosswalk
binding and the source aliasing - the logic that can run without a MySQL
source, which is the part that was wrong. Verified they are not vacuous by
running the old algorithm against the same fixture: it returns None where the
new one returns the row.

The README claimed 15 stages "complete" with counts from 2026-07-13 that two
later routing changes invalidated, and the runbook told operators to verify a
fresh import against them. Both corrected, with the three gaps NOT fixed here
recorded plainly: warranties, notifications and knowledgebase POST without a
duplicate check so resuming across them doubles rows; there is no --dry-run or
target guard; and the verify stage counts target assets against source machines,
which is apples to oranges.

Full suite 2178 passed.

scripts/site_imports/wjf/run.py

@@ -9,16 +9,19 @@ Stages are idempotent (re-running resolves existing rows by natural key), so a
 crashed run resumes. See scripts/site_imports/wjf/harness.py for the contract
 this drives, and scratchpad/IMPORT-PLAN.md for the full mapping + decisions.
 
-Implemented: reference (vendors, businessunits, operatingsystems), employees.
-TODO stages (assets hub + dependents + network + usb) are stubbed - they need
-the machineid->assetid crosswalk this harness persists.
+All 17 stages are implemented; STAGES at the bottom of this file is the list
+and main()'s --stages default is the order they must run in. The assets hub
+comes before everything that references an asset, because it is what builds the
+machineid->assetid crosswalk the harness persists.
 """
 
 import argparse
+import sys
+from urllib.parse import quote
 
 import pymysql
 
-from .harness import Harness
+from .harness import Harness, Source, StaleIdMapError
 
 
 def _id_of(row, idfield):
@@ -33,6 +36,76 @@ def _id_of(row, idfield):
     return None
 
 
+# The server clamps any page size to MAX_PAGE_SIZE (shopdb/config.py), which
+# is 100. Asking for more is not an error and not a warning - it just silently
+# returns 100 rows, which is why the 409-resolve below pages instead.
+PAGE_SIZE = 100
+
+# A resolve that walks more pages than this is a bug in the caller, not a big
+# table. Bounded so a paging mistake cannot spin forever against a live server.
+MAX_RESOLVE_PAGES = 200
+
+
+def _rows_of(payload):
+    """The list out of a list-endpoint response, whichever shape it took."""
+    if isinstance(payload, list):
+        return payload
+    if isinstance(payload, dict):
+        items = payload.get('items')
+        if isinstance(items, list):
+            return items
+    return []
+
+
+def _resolve_existing(h, lookup, unique_field, wanted, idfield):
+    """Find an existing row by its natural key. Returns its id, or None.
+
+    This used to ask for ?per_page=10000 and read the single response. The
+    server clamps per_page to 100, so it only ever searched the first 100 rows
+    and returned None for everything after that. The failure was silent and
+    downstream: _upsert returned None, the id never reached the crosswalk, and
+    the run finished "successfully" having quietly dropped a model's link on
+    every machine that used it. The target already holds more than 100 models
+    and more than 100 applications, so this fired on any re-run or resume - the
+    exact case the loader documents as safe.
+
+    Two passes. Some endpoints expose an exact natural-key filter for this
+    (?modelnumber= on models, ?appname= on applications, ?vendor= on vendors);
+    an endpoint without one ignores the unknown query parameter and returns its
+    normal first page, so the paged walk that follows is what finds the row.
+    """
+    wanted_key = str(wanted).strip().lower()
+    separator = '&' if '?' in lookup else '?'
+
+    def match(rows):
+        for row in rows:
+            if str(row.get(unique_field, '')).strip().lower() == wanted_key:
+                return _id_of(row, idfield)
+        return None
+
+    # Pass 1: the exact filter, where the endpoint has one.
+    exact = '%s%s%s=%s&per_page=%d' % (
+        lookup, separator, unique_field, quote(str(wanted)), PAGE_SIZE)
+    _, data = h.get(exact)
+    found = match(_rows_of(data))
+    if found is not None:
+        return found
+
+    # Pass 2: every page, in order.
+    for page in range(1, MAX_RESOLVE_PAGES + 1):
+        _, data = h.get('%s%sper_page=%d&page=%d'
+                        % (lookup, separator, PAGE_SIZE, page))
+        rows = _rows_of(data)
+        if not rows:
+            return None
+        found = match(rows)
+        if found is not None:
+            return found
+        if len(rows) < PAGE_SIZE:
+            return None
+    return None
+
+
 def _upsert(h, path, payload, unique_field, idfield, list_path=None):
     """POST payload; on 409 resolve the existing row by its unique field.
     Returns the new/existing id, or None."""
@@ -40,14 +113,8 @@ def _upsert(h, path, payload, unique_field, idfield, list_path=None):
     if status in (200, 201):
         return _id_of(data, idfield)
     if status == 409:
-        lookup = list_path or path
-        lookup += ('&' if '?' in lookup else '?') + 'per_page=10000'
-        _, rows = h.get(lookup)
-        items = rows if isinstance(rows, list) else rows.get('items', rows)
-        for row in (items or []):
-            if str(row.get(unique_field, '')).strip().lower() == \
-                    str(payload[unique_field]).strip().lower():
-                return _id_of(row, idfield)
+        return _resolve_existing(h, list_path or path, unique_field,
+                                 payload[unique_field], idfield)
     return None
 
 
@@ -418,10 +485,22 @@ def stage_metrology(h):
             'businessunitid': h.ids.get('businessunit', m['businessunitid']),
             'mapx': m['mapleft'], 'mapy': m['maptop'],
             'levelid': h.defaultlevelid})
-        if status not in (200, 201):
+        if status in (200, 201):
+            tool_assetid = _id_of(data, 'assetid')
+            made += 1
+        elif status == 409:
+            # Already imported by an earlier run. Resolve it rather than
+            # skipping: `continue` here left the crosswalk without this tool,
+            # so the Controls edge below was never made and a resumed run
+            # produced metrology PCs driving nothing. stage_printers already
+            # had this branch; this stage was written later and missed it.
+            tool_assetid = _resolve_existing(
+                h, '/api/measuringtools', 'assetnumber',
+                f"MT-{m['machineid']}", 'assetid')
+            if not tool_assetid:
+                continue
+        else:
             continue
-        tool_assetid = _id_of(data, 'assetid')
-        made += 1
         if tool_assetid and controls_id:
             h.ids.put('metrologytool', m['machineid'], tool_assetid)
             st, _ = h.post('/api/assets/relationships', {
@@ -906,10 +985,27 @@ def main():
                 'metrology,communications,applications,warranties,notifications,'
                 'knowledgebase,relationships,defaultprinters,subnets,usb,verify',
         help='comma list of stages to run')
+    parser.add_argument(
+        '--idmap', default=None,
+        help='path to the legacy-id crosswalk (default: idmap.json beside '
+             'this script). One per target database - a crosswalk from '
+             'another database maps to the wrong assets.')
     args = parser.parse_args()
 
-    h = Harness()
-    print(f'== WJ import loader (target: {h.app.config["SQLALCHEMY_DATABASE_URI"].split("/")[-1]}) ==')
+    try:
+        h = Harness(idmap_path=args.idmap)
+    except StaleIdMapError as error:
+        print('REFUSED: %s' % error)
+        return 1
+
+    target = h.app.config['SQLALCHEMY_DATABASE_URI'].split('/')[-1]
+    print(f'== WJ import loader (target: {target}) ==')
+    # Say which sources this run reads. They are overridable now, so "it
+    # imported nothing" is usually "it read the wrong database".
+    for logical in ('shopdb_src', 'cmmc_usb_src', 'wjf_employees_src'):
+        real = Source.resolve(logical)
+        if real != logical:
+            print(f'   source {logical} -> {real}')
     try:
         for name in args.stages.split(','):
             name = name.strip()
@@ -922,9 +1018,15 @@ def main():
             print(f'\n  {len(h.errors)} endpoint error(s):')
             for path, code, _payload, data in h.errors[:10]:
                 print(f'    {code} {path}: {data}')
+        # An endpoint error is not a crash, but it did leave rows unimported,
+        # and a run that exits 0 after 200 of them looks like a success in a
+        # deploy log.
+        if h.errors:
+            return 1
     finally:
         h.close()
+    return 0
 
 
 if __name__ == '__main__':
-    main()
+    sys.exit(main())