In HikaGenji/fql: Functional Query Language

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(fql)

This package is a collection of verbs that implement a DSL to work with dataframes. It is broadly inspired by [https://code.kx.com/v2/ref/]. FQL stands for Functional Query Language. This library has two dependencies:

functional
magrittr

Meaty Dataframe queries

Let us create a dataset of daily price and volume for ORA and VOD.

d <- data.frame(symbol=c(rep("ORA"  , 10), rep("VOD", 10)),
                date=as.character(rep(seq(as.Date("2019-01-01"),
                                          9+as.Date("2019-01-01"),by=1), 2)),
                price=c(14*cumprod(1+0.05*rnorm(10)), 150*cumprod(1+0.05*rnorm(10))),
                volume=c(cumsum(runif(10)), cumsum(runif(10)))
                  )
summary(d)

Now let us say that we want to compute the 10-day vwap and the average turnover by symbol:

d %>%
update(what=alist(turnover=price*volume)) %>%
select(by=alist(symbol=symbol),
       what=alist(turnover=mean(turnover),
                  vwap=sum(turnover) / sum(volume)))

Now let us say that we have the exchange in a separate table and we want to join it into the result:

e <- data.frame(symbol=c("ORA", "VOD"), exchangeid=c(1, 34))
d %>%
update(what=alist(turnover=price*volume)) %>%
select(by=alist(symbol=symbol),
       what=alist(turnover=mean(turnover),
                  vwap=sum(turnover) / sum(volume))) %>%
lj(e %>% xkey("symbol"))

Now let us say we want to compute returns at various horizons. We can achieve it with metaprogramming:

n <- 1:5
r <- n %>% each(function(k) parse(text=paste("log(price/xprev(price, ", k, "))"))) %>%
     dict(paste0("r_", n), .)
head(r, 2)

Once we have prepared the returns expressions, we can then update d with it:

d %>% 
update(by=alist(symbol=symbol), what=r) %>%
tail(5)

Function Decorators

onCall

onCall allows decorating a function and execute a function call right at the start of its evaluation. The default handler is to display its arguments.

f <- function(x) x
g <- onCall(f)
g(1)

onError

onError allows decorating a function and execute a handler whenever an error is thrown (tryCatch block). Its default handler is to print the error and return NULL.