quantlab.backtest.live.input_stores

quantlab.backtest.live.input_stores(backtester)

Return the stores a run’s prediction and decision read, each with who reads it.

The "reads" uses of datasets (not of ranged stores) declared under INPUT_FIELDS (walk_store_uses; the model’s labels are left out, as a prediction reads none): a leaf dataset’s store, a view’s inputs’ stores, and not what a dataset only writes its store from (a mapped Trade bar store’s source view is a "source"). Every store is keyed by its absolute path, with the component path of the leaf dataset reading it, whose calendar answers on it, deepest first. A panel held in memory reads no store and is keyed <in memory: path>. These are the stores predict_live_bar checks to hold the bar it predicts.

Parameters:

backtesterBaseBacktester

The run’s backtester, rebuilt from its directory.

Returns:

dict

{store: (component path, leaf dataset)}.

Examples

With backtester rebuilt from a run whose price dataset is a RosterDataset of a SEP store on a membership store:

>>> list(input_stores(backtester))
['/data/quantlab/market/sharadar/sharadar_sep_1d/sharadar_sep_1d.zarr', '/data/quantlab/universe/sp500/membership/membership.zarr']