Skip to content

Universes from Python

What a universe is, and how to write one as an expression, is in Universes. From Python, everywhere qex takes a universe (cross_section_universe, BookSettings(universe=...), explain_missing(universe=...)) it takes any of three forms:

Form Example Member when
a mask data['member'].notna() True
a data field's name 'member' the field has a value (1 in, NaN out)
a universe expression "field('mcap') >= 300e6", or a parsed qex(...) the expression is nonzero

Text that names a field of the data is that field ('member' stays the field it always was); anything else is parsed as qex.

from qex import qex, cross_section_universe, universe_mask
from evaluation import BookSettings, build_book

LARGE = "(field('price') >= 2) & (field('price') * field('shares') >= 2e9)"

with cross_section_universe(LARGE, data):                       # an expression needs the data
    r = qex("rank(returns(5))").evaluate(data)                  # ranked among the large stocks only

book = build_book(qex("rank(-returns(5))"), data, BookSettings(delay=1, universe=LARGE))
members = universe_mask(LARGE, data)                            # the True/False frame itself

Inside one compute_context, a universe expression's membership is computed once and shared by every signal evaluated in it. BookSettings(mask_cross_section=False) applies the universe only to the finished signal, as qex did before 0.5.

cross_section_universe

cross_section_universe(mask, data=None)

Restrict cross-sectional operators to the members of a universe, day by day.

mask is a date x ticker DataFrame of booleans, a universe field (1 in, NaN out), or, with data, a field's name or a universe expression (see the module docstring). Dates or tickers it does not cover are outside the universe. None leaves operators unrestricted.

universe_mask

universe_mask(universe, data=None) -> Optional[pd.DataFrame]

The members, date x ticker True/False, of a universe given as None, a mask, a data field's name, or a universe expression (qex text or a Signal) evaluated on data. Text that names a field of data is that field.