cache on expression level, part 1

Not really active on expression level, just some infrastructure and code running. Must now process each block separately, i.e. parse_transform_serialize and parse_transform_serialize_block from branch caching@48a9a6

R/nest.R

@@ -5,10 +5,13 @@
 #' @param text A character vector to parse.
 #' @return A nested parse table. See [tokenize()] for details on the columns
 #'   of the parse table.
+#' @inheritParams tokenize
 #' @importFrom purrr when
 #' @keywords internal
-compute_parse_data_nested <- function(text) {
-  parse_data <- tokenize(text) %>%
+compute_parse_data_nested <- function(text,
+                                      transformers = NULL,
+                                      cache_dir = NULL) {
+  parse_data <- tokenize(text, transformers, cache_dir) %>%
     add_terminal_token_before() %>%
     add_terminal_token_after() %>%
     add_stylerignore()

R/parse.R

@@ -71,31 +71,56 @@ has_crlf_as_first_line_sep <- function(message, initial_text) {
 #'   * A column "child" that contains *nest*s.
 #'
 #' @param text A character vector.
+#' @inheritParams get_parse_data
 #' @return A flat parse table
 #' @importFrom rlang seq2
 #' @keywords internal
-tokenize <- function(text) {
-  get_parse_data(text, include_text = NA) %>%
+tokenize <- function(text,
+                     transformers = NULL,
+                     cache_dir = NULL) {
+  get_parse_data(text,
+    include_text = NA,
+    transformers = transformers,
+    cache_dir = cache_dir
+  ) %>%
     ensure_correct_str_txt(text) %>%
     enhance_mapping_special()
 }
 
 #' Obtain robust parse data
 #'
 #' Wrapper around `utils::getParseData(parse(text = text))` that returns a flat
-#' parse table.
+#' parse table. When caching information should be added, make sure that
+#' the cache is activated with `cache_activate()` and both `transformers` and
+#' `cache_dir` are non-`NULL`.
 #' @param text The text to parse.
 #' @param include_text Passed to [utils::getParseData()] as `includeText`.
 #' @param ... Other arguments passed to [utils::getParseData()].
+#' @param transformers A list of transformer functions. Used only to determine
+#'   if the code to style was previously cached.
+#' @param cache_dir The directory to look for the cache.
 #' @keywords internal
-get_parse_data <- function(text, include_text = TRUE, ...) {
+get_parse_data <- function(text,
+                           include_text = TRUE,
+                           transformers = NULL,
+                           cache_dir = NULL,
+                           ...) {
   # avoid https://bugs.r-project.org/bugzilla3/show_bug.cgi?id=16041
   parse_safely(text, keep.source = TRUE)
   parsed <- parse_safely(text, keep.source = TRUE)
   pd <- as_tibble(
     utils::getParseData(parsed, includeText = include_text),
     .name_repair = "minimal") %>%
     add_id_and_short()
+
+  pd$block <- seq_len(nrow(pd))
+  pd$is_cached <- NA
+  top_level_exprs <- which(pd$parent <= 0)
+  if (cache_is_activated() && !is.null(cache_dir) && !is.null(transformers)) {
+    for (idx in top_level_exprs) {
+      pd$is_cached[idx] <- is_cached(pd$text[idx], transformers, cache_dir)
+    }
+  }
   parser_version_set(parser_version_find(pd))
   pd
 }

R/token-create.R

@@ -31,7 +31,9 @@ create_tokens <- function(tokens,
                           indents = 0,
                           terminal = TRUE,
                           child = NULL,
-                          stylerignore = FALSE) {
+                          stylerignore = FALSE,
+                          block = NA,
+                          is_cached = NA) {
   len_text <- length(texts)
   new_tibble(
     list(
@@ -50,7 +52,9 @@ create_tokens <- function(tokens,
     indention_ref_pos_id = indention_ref_pos_ids,
     indent = indents,
     child = rep(list(child), len_text),
-    stylerignore = stylerignore
+    stylerignore = stylerignore,
+    block = block,
+    is_cached = is_cached
   ),
     nrow = len_text
   )

R/transform-files.R

@@ -87,11 +87,7 @@ make_transformer <- function(transformers,
     should_use_cache <- is_R.cache_installed && cache_is_activated()
 
     if (should_use_cache) {
-      use_cache <- R.cache::generateCache(
-        key = cache_make_key(text, transformers),
-        dirs = cache_dir
-      ) %>%
-        file.exists()
+      use_cache <- is_cached(text, transformers, cache_dir)
     } else {
       use_cache <- FALSE
     }
@@ -197,7 +193,7 @@ split_roxygen_segments <- function(text, roxygen_examples) {
 #' @keywords internal
 parse_transform_serialize_r <- function(text, transformers, warn_empty = TRUE) {
   text <- assert_text(text)
-  pd_nested <- compute_parse_data_nested(text)
+  pd_nested <- compute_parse_data_nested(text, transformers)
   start_line <- find_start_line(pd_nested)
   if (nrow(pd_nested) == 0) {
     if (warn_empty) {

R/utils-cache.R

@@ -13,6 +13,23 @@ hash_standardize <- function(text) {
     list()
 }
 
+#' Check if text is cached
+#'
+#' This boilds down to check if the hash exists at the caching dir as a file.
+#' @param text,transformers Passed to [cache_make_key()] to generate a key.
+#' @param cache_dir The caching directory relative to the `.Rcache` root to
+#'   look for a cached value.
+#' @keywords internal
+is_cached <- function(text, transformers, cache_dir) {
+  R.cache::generateCache(
+    key = cache_make_key(text, transformers),
+    dirs = cache_dir
+  ) %>%
+    file.exists()
+
+}
+
+
 #' Make a key for `R.cache`
 #'
 #' This is used to determine if caching already corresponds to a style guide.