Queries and workloads

A workload is an ordered collection of query entries. Entries optionally carry an integer pattern_id, useful for relating generated query instances to a shared shape.

from bondi.mongodb import MongoCollectionRef, MongoFindQuery
from bondi.workload import Workload, WorkloadEntry

business = MongoCollectionRef("demo", "business")
workload = Workload([
    WorkloadEntry(MongoFindQuery(business, {"city": "Sydney"}), pattern_id=0),
    WorkloadEntry(MongoFindQuery(business, {"city": "Melbourne"}), pattern_id=0),
])
print(len(workload))
print(workload[0].query.filter)

All entries must use the same backend. MongoDB find and aggregation queries can coexist; MongoDB and Couchbase queries cannot. The shared Query protocol describes common behavior without translating query languages.

Save and load

workload.save("workload.jsonl")
restored = Workload.load("workload.jsonl")
assert restored[0].query.fingerprint() == workload[0].query.fingerprint()

Each JSONL line stores a version, a native query envelope and a pattern ID. Loading reconstructs concrete query objects and BSON values where applicable. It does not connect to a database. Pattern objects and generation statistics are not stored.

The workload copies the entry list but keeps references to entries and queries. Edits are visible through those references; validation occurs at explicit checkpoints, including saving and execution. Do not mutate a workload while it is running.

Next: execute the workload or generate patterns. See the reference for custom queries, validation and persistence errors.