Public API

bagelquant-data is operated through Python APIs. The package intentionally returns pandas objects and plain metadata instead of importing downstream research packages.

Top-Level Exports

from bagelquant_data import (
    DataContract,
    DataLakeManager,
    DataRequest,
    DataSource,
    DataSourceRegistry,
    DatasetSchema,
    FieldSchema,
    LoadedDataset,
    Loader,
    LocalDataLake,
    RetrievedPanel,
    Transform,
    TushareTableUpdateSpec,
    default_registry,
)

Provider Registry

• DataSource: base provider adapter protocol.
• DataRequest: dataset, fields, filters, date range, version, snapshot, and options.
• DataSourceRegistry.register(source): register an adapter.
• DataSourceRegistry.resolve(name): retrieve a registered adapter.
• default_registry: process-level convenience registry.

Data Lake

• LocalDataLake(root): local filesystem snapshot backend.
• DataLakeManager(lake, registry=None): high-level mutation and update API.
• DataLakeManager.add(source, dataset, frame): add custom data.
• DataLakeManager.edit(source, dataset, frame): replace a dataset snapshot.
• DataLakeManager.delete(source, dataset): delete a dataset pointer.
• LocalDataLake.read(...): read projected and date-filtered data.
• LocalDataLake.read_panel_field(...): shape a qualified field id into a panel frame.

Loader

• Loader(registry=None, lake=None): lake-first retrieval facade.
• Loader.source(name).load(...): return LoadedDataset.
• Loader.source(name).load_panel(...): return RetrievedPanel.
• LoadedDataset.data: pandas dataset payload.
• RetrievedPanel.data: date-by-asset frame for one field.

Metadata And Contracts

• DataContract: provider or dataset contract.
• DatasetSchema: dataset-level schema metadata.
• FieldSchema: field-level metadata.
• Transform: stateless DataFrame transformation pipeline.

Tushare Helpers

• TushareDataSource: provider adapter exposed from bagelquant_data.datasource.
• TushareTableUpdateSpec: production-style table update specification.
• DataLakeManager.scan_tushare_updates(...): build a dry-run update report.
• DataLakeManager.execute_tushare_update_report(report, workers=4): execute reviewed jobs.