release: 1.2.0

[MobyNL]

Two features, then the release commit.

Load Document and Insert Document From File

A fixture document read from an Extended JSON file rather than written into a suite, so
$oid and $date arrive as the BSON types MongoDB compares against. Two kinds of hole —
${name} from the calling suite, {name} from the keyword's arguments — plus dotted-path
overrides of fields the file already fills. A new document_path import argument names the
lookup directory and does nothing else.

Explain Query and Collection Should Have Index

A query that matches nothing and reports no error was documented for exactly one cause, a
string _id against an ObjectId. Explain Query covers the rest: it reports the plan the
server chose and, most usefully, index_bounds — the values it actually searched for.

${plan}    Explain Query    collection_name=readings    _id.deviceId=${device_id}    _id.date=${date}
Log    ${plan.index_bounds}

Explain of {'_id.deviceId': 'device-1', '_id.date': datetime.datetime(2026, 1, 8, 0, 0)} on 'readings':
  stage            FETCH
  index_name       _id.deviceId_1__id.date_1
  collection_scan  False
  keys_examined    1
  docs_examined    1
  returned         1
  index_bounds
    _id.deviceId  ["device-1", "device-1"]
    _id.date      [new Date(1767830400000), new Date(1767830400000)]

Design notes worth reviewing:

• The summary is derived, not copied, because the explain document's shape varies while
  the questions do not: SBE nests stages under queryPlan, sharded clusters report per
  shard, and the _id fast path is EXPRESS_IXSCAN on MongoDB 8 but IDHACK before it.
  A scan is detected by the presence of a COLLSCAN, so there is no version check.
• It runs through database.command rather than cursor.explain(), which pymongo pins to
  allPlansExecution; that is what makes executionStats the default and queryPlanner
  available.
• It is a diagnostic and asserts nothing, collection_scan included — MongoDB rightly
  scans small collections, so such an assertion would pass on production-shaped data and
  fail on a freshly seeded test collection with nothing wrong. Collection Should Have Index is the assertion, reading index definitions rather than a plan so it cannot flake,
  and asking by fields rather than by the derived name.

Verification

• 368 unit tests, MongoDBLibrary/keywords.py at 100% coverage; mypy, ruff and
  robocop check atest clean.
• New atest/explain_tests.robot, 11/11 against MongoDB 8, seeding 200 documents keyed by
  a compound _id. One test reproduces the field-order trap end to end: the same _id
  with its fields reversed explains as returned: 0 while Find Document With Query
  returns None.
• Explain unit tests use canned explain documents on the MagicMock fixture, since mongomock
  implements no explain at all.

🤖 Generated with Claude Code