cachemeifyoucan.R

cachemeifyoucan-sketch.R

fn2 <- function(id, model_version, type = 'loan_id') {
  Reduce(rbind, lapply(id, function(id) {
    seed <- as.integer(paste0("0x", substr(digest::digest(paste(id, model_version, type)), 1, 6)))
    set.seed(seed)
    data.frame(id = id, x = runif(1), y = rnorm(1))
  }))
}

cache <- function(uncached_function, prefix, key, salt) {
  cached_function <- new("function")
  formals(cached_function)     <- formals(uncached_function)
  environment(cached_function) <-
    list2env(list(prefix = prefix, key = key, salt = salt, uncached_function = uncached_function),
      parent = environment(uncached_function))

  body(cached_function) <- quote({
    raw_call <- match.call()
    call <- as.list(raw_call[-1])

    true_salt <- call[intersect(names(call), salt)]
    for (name in names(true_salt)) {
      true_salt[[name]] <- eval.parent(true_salt[[name]])
    }
    true_salt <- digest::digest(true_salt)

    table_name <- paste0(prefix, '_', true_salt)

    cachemeifyoucan::execute(
      cached_function_call(uncached_function, raw_call, parent.frame(), table_name, key)
    )
  })

  cached_function


}

execute <- function(function_call) {
  key <- eval(function_call$call[[function_call$key]], envir = function_call$context)
  # Check which parts of key are already in function_call$table
  cached_keys   <- ...
  uncached_keys <- setdiff(key, cached_keys)

  cached_data <- grab from database using function_call$table and cached_keys
  function_call$call[[function_call$key]] <- uncached_keys
  uncached_data <- eval(function_call$call, envir = function_call$context)

  # Save to database

  # plyr::rbind.fill
}


cached_function_call <- function(fn, call, context, table, key) {
  structure(list(fn = fn, call = call, context = context, table = table, key = key),
            class = 'cached_function_call')
}

fn <- cache(fn2, prefix = 'blah_', key = 'id', salt = c('model_version', 'type'))

fn(1:3, 'default/en-US/1.0')
fn(1:3, 'default/en-US/1.0', type = 'customer_id') # should use a totally diff table

Can you provide More documentation when deployed, to provide greater readability

Agreed, this needs more documentation than there is code, plus heavy re-factoring

@FeiYeYe Can you try to make each method < 10 lines?

certainly, i need to spend some time refactoring the current cache function which is much less general than this gist!