About a year ago, I decided to stop using dplyr for data processing and switched to using data.table. One important reason is its more concise syntax in most situations and faster performance. Another reason is that using tidy-like code often involves loading many packages, leading to a miscellaneous workspace. Tidy-like code typically involves repeating some behaviors in base R but provides a unified interface for performing tasks more conveniently and rigorously.

In certain cases, we might want to generate an R expression for later use. A concrete example is producing a linear expression from a named vector:

fit <- c(x = 1, y = 2)

coef_sym <- syms(names(fit))
coefs <- unname(fit)

summands <- map2(coef_sym, coefs, ~ expr((!!.x * !!.y)))
eq <- purrr::reduce(summands, ~ expr(!!.x + !!.y))

This solution is from the Advanced R book. The rlang and purrr packages provide functions related to non-standard evaluation (NSE) and functional programming tools. However, achieving the same result is possible using functions in base R. Some functions related to metaprogramming are listed below:

• parse and deparse: Convert between character string and expression.
• quote: Quote an R expression.
• expression: Quote multiple R expressions.
• bquote: Quoting and unquoting within a function call.
• substitute: Substitutes the values of variables into an R expression.
• call and as.call: Construct a function call.
• eval: Evaluate an R expression in a specified environment.

bquote:

An analogue of the LISP backquote macro. bquote quotes its argument, except that terms wrapped in .() are evaluated in the specified environment. If splice = TRUE, then terms wrapped in ..() are evaluated and spliced into a call.

substitute:

substitute returns the parse tree for the (unevaluated) expression expr, substituting any variables bound in env.

Thus, we can do it like this:

# Assuming 'fit' is a named numeric vector
fit <- c(x = 1, y = 2)

coef_sym <- names(fit)
coefs <- unname(fit)

summands <- mapply(function(x, y) bquote(.(as.name(x)) * .(y)), coef_sym, coefs, SIMPLIFY = FALSE)
eq <- Reduce(function(x, y) bquote(.(x) + .(y)), summands)

In data.table, we might want to construct different models for various variables, such as performing univariate Cox regression for multiple variables. We can achieve this as follows(This solution was originally from stackoverflow):

candidate <- c('ARG1', 'CASP9')
dt <- data.table(exp_vars = candidate)
formulas <- sprintf("Surv(OS, vital_status) ~ %s", candidate)

dt[, let(model = lapply(formulas, function(f) coxph(as.formula(f), data = train)) ), by = exp_vars]
dt[, let( # get p value
  p.cox = sapply(model, function(mod) summary(mod)$coefficients[1, "Pr(>|z|)"]),
  p.cox.zph = sapply(model, function(model) cox.zph(model)$table[1, 3])
)]