Files
shopdb-flask/shopdb/core/api/mappositions.py
cproudlock 3324dbd91e
Some checks failed
CI / backend (push) Failing after 9s
CI / naming (push) Successful in 2s
CI / frontend (push) Successful in 10s
CI / migrations-mysql (push) Failing after 7s
Buildings and levels for the floor map, and make every identifier searchable
The map was one picture of one floor. A second floor was added, the blueprint
changed size, and machines moved, so a position now records WHICH DRAWING its
coordinates belong to.

Buildings and levels (ADR-017). Each level owns its blueprint per theme and its
own native pixel size; assets.mapx/mapy are pixels of assets.levelid, not of the
site. A position whose level is unknown renders "level unknown" and is never
drawn on the default level, because a marker on the wrong floor plan looks
entirely correct while pointing at the wrong place.

Repositioning in bulk: filter by unplaced, needs-review or level, search, place,
confirm. Landmark recalibration solves the transform PER AXIS from landmark
pairs and never from image dimensions - the canvas grew taller without
rescaling, so a dimension-derived scale would stretch Y by 1.57 and be wrong
everywhere. It defaults to a dry run, reports what would land off the drawing,
snapshots before applying, and clears mapverifiedat because a transform is a
guess awaiting review. Snapshots restore, including the level and the review
state, and a restore snapshots first so an undo is undoable.

Search: gaugelabreference was matched only for measuring tools and
maintenancereference was matched nowhere at all, for any asset type, while
Settings happily offers both identifiers on machines and PCs. A tag an operator
is told to record has to be findable or it is a write-only field. USB devices
and printed items were unreachable from search entirely - neither is an asset,
so the generic asset search could not see them and no searcher existed; they
now match on serial, asset tag, label, bin code and gage-lab tag, honouring
isactive, with Settings toggles and result labels to match.

The retired-application rule was half a rule: GET /api/knowledgebase hid
articles whose topic application is retired while global search still returned
them and printed the retired application as the subject. A filter is only real
if every path that reaches the row applies it.

Contract to 0.20.0 (additive): Asset gained levelid and mapverifiedat, Location
gained levelid, and resolve_asset_position returns the levelid belonging to
whichever source supplied the coordinates. The five plugins that write a map
position are re-pinned. The install-list text format gained levelid as a NINTH
field, appended, because the shipped Pascal installer reads fields 0-7 by index.

That installer still compiles in one drawing's dimensions and bundles one
blueprint, so its map is accurate for the default level only; /api/maplevels is
deliberately unauthenticated so it can read both at runtime once rebuilt.
Recorded in PRINTER-INSTALLER.md section 6 along with the other known gaps.

Migration 7d33 converts an existing single-map site into one building and one
default level carrying the old map_* settings, then assigns every placed asset
and location to it. Nothing moves on screen. Old settings rows are kept so a
rollback still finds them. Verified end to end on MySQL 5.6 from a
production-shaped database.
2026-08-17 12:55:51 -04:00

404 lines
16 KiB
Python

"""Bulk marker positions: transform, place, verify, undo (ADR-017).
A new blueprint invalidates every marker on a level at once, so the operations
here are deliberately bulk. Three rules shape all of them:
**A transform is derived from landmarks, never from image sizes.** The real case
that motivated this is 3300x2550 to 3308x4000, where the second level was added
below the first: the correct transform is identity scale with a Y offset, and a
scale derived from the dimension ratio would stretch every Y by 1.57 and be wrong
everywhere. Dimensions describe the canvas; landmarks describe the drawing.
**A transform is a guess, so it clears the review state.** The levels were
redrawn and machines moved. Nothing in a coordinate says whether the machine it
points at is still there, so every transformed marker is unreviewed until a human
confirms it.
**Nothing bulk happens without a snapshot.** Positions have no history, and an
operation that rewrites hundreds of rows needs a way back that is not a database
restore.
"""
from datetime import datetime, timezone
from flask import Blueprint, request
from flask_jwt_extended import jwt_required, get_jwt_identity
from shopdb.extensions import db
from shopdb.core.models import (Asset, AuditLog, MapLevel,
MapPositionSnapshot, User)
from shopdb.utils.responses import success_response, error_response, ErrorCodes
from shopdb.utils.authz import require_permission
mappositions_bp = Blueprint('mappositions', __name__)
def _now():
return datetime.now(timezone.utc).replace(tzinfo=None)
def _actor():
try:
user = db.session.get(User, int(get_jwt_identity()))
return user.username if user else None
except (TypeError, ValueError):
return None
# =============================================================================
# The transform
# =============================================================================
def solve_axis(pairs):
"""Least-squares scale and offset for one axis: new = scale * old + offset.
Two pairs solve it exactly; more are averaged, which is worth having because
a landmark picked by eye on a scanned drawing carries a few pixels of error
and three points let that cancel instead of accumulate.
Returns None when the landmarks cannot determine the axis - every old value
identical, so the denominator is zero. That is a real user error (two
landmarks on the same column) and it must be reported rather than papered
over with a scale of 1, which would look like it worked.
"""
count = len(pairs)
if count < 2:
return None
sumold = sum(old for old, _ in pairs)
sumnew = sum(new for _, new in pairs)
sumoldsq = sum(old * old for old, _ in pairs)
sumcross = sum(old * new for old, new in pairs)
denominator = count * sumoldsq - sumold * sumold
if abs(denominator) < 1e-9:
return None
scale = (count * sumcross - sumold * sumnew) / denominator
offset = (sumnew - scale * sumold) / count
return scale, offset
def derive_transform(landmarks):
"""Per-axis transform from [{'fromx','fromy','tox','toy'}, ...].
Per-axis rather than uniform on purpose. A level added below another changes
the canvas height without rescaling anything, so Y gets an offset and X gets
neither; forcing one scale onto both axes cannot express that.
"""
try:
xpairs = [(float(mark['fromx']), float(mark['tox'])) for mark in landmarks]
ypairs = [(float(mark['fromy']), float(mark['toy'])) for mark in landmarks]
except (KeyError, TypeError, ValueError):
return None, 'each landmark needs numeric fromx, fromy, tox and toy'
if len(landmarks) < 2:
return None, 'at least two landmarks are required'
xsolution = solve_axis(xpairs)
ysolution = solve_axis(ypairs)
if xsolution is None:
return None, ('the landmarks do not vary in X, so no horizontal scale '
'can be derived - pick points that differ across the drawing')
if ysolution is None:
return None, ('the landmarks do not vary in Y, so no vertical scale '
'can be derived - pick points that differ down the drawing')
return {
'scalex': xsolution[0], 'offsetx': xsolution[1],
'scaley': ysolution[0], 'offsety': ysolution[1],
}, None
def apply_transform(asset, transform):
return (
int(round(asset.mapx * transform['scalex'] + transform['offsetx'])),
int(round(asset.mapy * transform['scaley'] + transform['offsety'])),
)
@mappositions_bp.route('/transform', methods=['POST'])
@jwt_required()
@require_permission('assets.edit')
def transform_positions():
"""Move every placed marker on a level by a transform read off landmarks.
Body:
levelid the level whose markers are being moved (required)
landmarks [{fromx, fromy, tox, toy}, ...] - two or more points, each the
same physical feature on the old drawing and the new one
tolevelid optional: write the results to a different level, for splitting
one drawing into two
assetids optional: restrict to these assets
dryrun default TRUE. Nothing is written unless this is explicitly false.
Dry run returns every marker's old and new position and whether it lands
outside the target level, which is the only way to see a bad landmark pair
before it has moved 300 markers.
"""
data = request.get_json() or {}
levelid = data.get('levelid')
level = db.session.get(MapLevel, levelid) if levelid else None
if not level:
return error_response(ErrorCodes.VALIDATION_ERROR,
'levelid is required and must exist')
target = level
if data.get('tolevelid'):
target = db.session.get(MapLevel, data['tolevelid'])
if not target:
return error_response(ErrorCodes.NOT_FOUND, 'No such tolevelid',
http_code=404)
transform, problem = derive_transform(data.get('landmarks') or [])
if problem:
return error_response(ErrorCodes.VALIDATION_ERROR, problem)
query = Asset.query.filter(
Asset.levelid == level.levelid,
Asset.mapx.isnot(None), Asset.mapy.isnot(None),
Asset.isactive.is_(True))
if data.get('assetids'):
query = query.filter(Asset.assetid.in_(data['assetids']))
assets = query.order_by(Asset.assetid).all()
moves = []
outofbounds = 0
for asset in assets:
newx, newy = apply_transform(asset, transform)
outside = not (0 <= newx <= target.mapwidth and
0 <= newy <= target.mapheight)
if outside:
outofbounds += 1
moves.append({
'assetid': asset.assetid,
'assetnumber': asset.assetnumber,
'name': asset.name,
'fromx': asset.mapx, 'fromy': asset.mapy,
'tox': newx, 'toy': newy,
'outofbounds': outside,
})
# The derived transform is reported back whichever mode this is, because it
# is the number a human can sanity-check: a Y scale of 1.57 on a level that
# only grew taller is the mistake this endpoint exists to avoid, and it is
# obvious in the response and invisible in the result.
payload = {
'transform': transform,
'levelid': level.levelid,
'tolevelid': target.levelid,
'assetcount': len(moves),
'outofboundscount': outofbounds,
'moves': moves,
'dryrun': True,
}
if data.get('dryrun', True):
return success_response(payload)
snapshot = MapPositionSnapshot.capture(
assets,
reason='transform of %s (%d markers)' % (level.levelname, len(moves)),
levelid=level.levelid, createdby=_actor())
for asset, move in zip(assets, moves):
asset.mapx = move['tox']
asset.mapy = move['toy']
asset.levelid = target.levelid
# Cleared, not preserved: a transformed position is a guess, and the
# whole point of the review pass is to tell guesses from confirmations.
asset.mapverifiedat = None
db.session.commit()
AuditLog.log('updated', 'Asset', entityname='%d marker(s) transformed'
% len(moves),
changes={'transform': transform,
'snapshotid': snapshot.snapshotid})
db.session.commit()
payload['dryrun'] = False
payload['snapshotid'] = snapshot.snapshotid
return success_response(payload, message='%d marker(s) moved; snapshot %d '
'can restore them' % (len(moves),
snapshot.snapshotid))
# =============================================================================
# Bulk place and verify
# =============================================================================
@mappositions_bp.route('/positions', methods=['POST'])
@jwt_required()
@require_permission('assets.edit')
def set_positions():
"""Set positions for many assets at once.
Body: {positions: [{assetid, mapx, mapy, levelid}], verified: bool}
`levelid` is required per position rather than taken from a single body-level
value, because a bulk save from the editor can legitimately span levels, and
inferring it would be the guess this whole feature exists to remove.
"""
data = request.get_json() or {}
rows = data.get('positions') or []
if not rows:
return error_response(ErrorCodes.VALIDATION_ERROR,
'positions is required and must not be empty')
wanted = {}
for row in rows:
assetid = row.get('assetid')
levelid = row.get('levelid')
if not assetid or not levelid:
return error_response(
ErrorCodes.VALIDATION_ERROR,
'every position needs an assetid and a levelid; a position '
'without a level cannot be rendered')
if row.get('mapx') is None or row.get('mapy') is None:
return error_response(ErrorCodes.VALIDATION_ERROR,
'every position needs mapx and mapy')
wanted[int(assetid)] = row
levelids = {int(row['levelid']) for row in wanted.values()}
known = {level.levelid for level in
MapLevel.query.filter(MapLevel.levelid.in_(levelids)).all()}
missing = levelids - known
if missing:
return error_response(ErrorCodes.NOT_FOUND,
'unknown levelid(s): %s'
% ', '.join(str(one) for one in sorted(missing)),
http_code=404)
assets = Asset.query.filter(Asset.assetid.in_(wanted)).all()
found = {asset.assetid for asset in assets}
unknown = set(wanted) - found
if unknown:
return error_response(ErrorCodes.NOT_FOUND,
'unknown assetid(s): %s'
% ', '.join(str(one) for one in sorted(unknown)),
http_code=404)
snapshot = MapPositionSnapshot.capture(
assets, reason='bulk position set (%d markers)' % len(assets),
createdby=_actor())
verified = _now() if data.get('verified') else None
for asset in assets:
row = wanted[asset.assetid]
asset.mapx = int(row['mapx'])
asset.mapy = int(row['mapy'])
asset.levelid = int(row['levelid'])
# Placing a marker by hand IS the confirmation, so this stamps it. A
# caller that is only nudging things about can pass verified=false.
if verified or data.get('verified') is not False:
asset.mapverifiedat = verified or _now()
db.session.commit()
AuditLog.log('updated', 'Asset',
entityname='%d marker position(s) set' % len(assets),
changes={'snapshotid': snapshot.snapshotid})
db.session.commit()
return success_response({'updated': len(assets),
'snapshotid': snapshot.snapshotid},
message='%d position(s) saved' % len(assets))
@mappositions_bp.route('/verify', methods=['POST'])
@jwt_required()
@require_permission('assets.edit')
def verify_positions():
"""Mark markers as reviewed against the current drawing without moving them.
This is the common case in a review pass: the transform put it in the right
place and a human agrees. No snapshot, because nothing about the position
changes - only the statement that somebody looked.
"""
data = request.get_json() or {}
assetids = data.get('assetids') or []
if not assetids:
return error_response(ErrorCodes.VALIDATION_ERROR,
'assetids is required')
stamp = None if data.get('unverify') else _now()
updated = (Asset.query.filter(Asset.assetid.in_(assetids))
.update({'mapverifiedat': stamp}, synchronize_session=False))
db.session.commit()
return success_response({'updated': updated},
message='%d marker(s) %s' %
(updated, 'unverified' if stamp is None
else 'marked reviewed'))
# =============================================================================
# Undo
# =============================================================================
@mappositions_bp.route('/snapshots', methods=['GET'])
@jwt_required()
@require_permission('assets.view')
def list_snapshots():
"""Snapshots, newest first. Metadata only - the positions are large."""
snapshots = (MapPositionSnapshot.query
.order_by(MapPositionSnapshot.snapshotid.desc())
.limit(50).all())
return success_response([one.to_dict() for one in snapshots])
@mappositions_bp.route('/snapshots/<int:snapshotid>/restore', methods=['POST'])
@jwt_required()
@require_permission('assets.edit')
def restore_snapshot(snapshotid):
"""Put the positions in this snapshot back.
Takes its own snapshot first, so an undo is itself undoable - which matters
because the most likely reason to restore is a transform that looked right in
the preview and wrong on the drawing, and the second attempt is rarely the
last one either.
"""
snapshot = db.session.get(MapPositionSnapshot, snapshotid)
if not snapshot:
return error_response(ErrorCodes.NOT_FOUND, 'No such snapshot',
http_code=404)
rows = snapshot.positions
if not rows:
return error_response(ErrorCodes.VALIDATION_ERROR,
'this snapshot holds no positions')
assetids = [row['assetid'] for row in rows]
assets = {asset.assetid: asset for asset in
Asset.query.filter(Asset.assetid.in_(assetids)).all()}
MapPositionSnapshot.capture(
list(assets.values()),
reason='before restoring snapshot %d' % snapshotid,
levelid=snapshot.levelid, createdby=_actor())
restored = 0
skipped = []
for row in rows:
asset = assets.get(row['assetid'])
if asset is None:
# The asset was deleted since the snapshot. Say so rather than
# failing the whole restore: the other 299 markers still want
# putting back.
skipped.append(row['assetid'])
continue
asset.mapx = row.get('mapx')
asset.mapy = row.get('mapy')
asset.levelid = row.get('levelid')
stamp = row.get('mapverifiedat')
asset.mapverifiedat = datetime.fromisoformat(stamp) if stamp else None
restored += 1
snapshot.restoredat = _now()
db.session.commit()
AuditLog.log('updated', 'Asset',
entityname='%d marker(s) restored from snapshot %d'
% (restored, snapshotid),
changes={'snapshotid': snapshotid, 'skipped': skipped})
db.session.commit()
return success_response(
{'restored': restored, 'skippedassetids': skipped},
message='%d marker(s) restored%s' %
(restored, '; %d asset(s) no longer exist' % len(skipped)
if skipped else ''))