quantlab.dataset.roster#
Roster dataset: another dataset read on a roster only.
A universe computes its factors and labels on its own Roster: the
undated list of symbols its panel may hold (cross-sectional ranks and
z-scores depend on who is in the panel). RosterDataset wraps the
vendor’s market-wide dataset and takes the roster dataset’s whole symbol
axis, so a universe reads the shared store instead of keeping a copy of it
cut to the roster. The roster is read on every call and follows its store:
a universe’s membership store gains a symbol the day it first enters.
The wrapper is a drop-in MarketDataset: a factor, a label or a backtest
given it in place of the wrapped dataset changes nothing else. Its calendar,
variable names, KunQuant export, tradable and delisting bars and ticker
lookup are the wrapped dataset’s. It has no store of its own and no build
path: its inputs are the wrapped and the roster datasets (inputs()), whose
stores are updated by their own vendor updates. Who is a member on a
given date is the membership a label or a predictor is masked with.
The wrapped dataset records the read in an open DataRecorder, asked
only for the roster’s symbols, so a run’s data fingerprint covers the
roster’s cells only.
Classes
|