finds.structured.signals

Signals implements structured dataset interface for derived signal values

Copyright 2022, Terence Lim

MIT License

class finds.structured.signals.Signals(sql: SQL, verbose=0)

Bases: Stocks

Provide structured stocks data interface to derived signal values

Parameters:

sql – Connection to SQL database

__call__(label: str, date: int, start: int = -1, rebaldate: str = 'rebaldate') → DataFrame

Return cross-section of signal values available as of a date

Parameters:

• label – Name of signal to retrieve

• date – As of this date back through (non-inclusive) start date

• start – Non-inclusive start of date range; -1 means exact date

• rebaldate – Name of rebalance date column

Returns:

DataFrame of signal values prevailing as of input date

__getitem__(label) → Table

Overrides parent class method to get Table schema of label

read(label: str, where: str = '') → DataFrame

Read signal values from sql and return as data frame

Parameters:

• label – Name of signal

• where – Where clause for sql select

Returns:

DataFrame of query with columns = [‘permno’, ‘rebaldate’, label]

summary(label: str) → DataFrame

Perform a ‘proc summary’ by rebaldate on a signal’s values

table_key(label: str) → str

Helper method generates a table key name for the input label

write(data: DataFrame, label: str, overwrite: bool = True, rebaldate: str = 'rebaldate', permno: str = 'permno') → int

Saves a new sql table from dataframe of signal values

Parameters:

• data – Signal values, with columns [‘permno’, ‘rebaldate’, label]

• label – Signal name of column and table (prefixed ‘__’)

• overwrite – If False, append to table ignoring dups. Else recreate

• rebaldate – Column name of rebalance dates in input dataframe

• permno – Column name of permno identifiers in input dataframe

Returns:

Number of rows saved

Notes:

• first removes dup keys, then drops null rows before saving to table

class finds.structured.signals.SignalsFrame(df: DataFrame, identifier: str = 'permno', rebaldate: str = 'rebaldate')

Bases: Signals

Cache dataframe of signals values, provide Signals-like interface

Parameters:

• df – DataFrame input with permno and rebaldate as columns

• identifer – column name of permno identifiers

• rebaldate – column name of rebalance dates

__call__(label: str, date: int, start: int = -1) → DataFrame

Select from rebaldates that fall between start and date, keep latest

Parameters:

• label – Name of column to return

• date – As of this date back through (non-inclusive) start date

• start – Non-inclusive start date. Set to 0 for all, -1 for exact