Implement req_perform_spider()

Fixes#456

R/spider.R

@@ -0,0 +1,149 @@
+#' @param hash_key A function with argument `req` that returns the components
+#'    of the request that should be used for computing equality. By default,
+#'    `hash_key` inspects the `url`, `body`, and `headers`, which should be
+#'    adequate for most needs.
+#' @param progress_label A 1function with `req` that returns a string used to
+#'    label the progress bar. The default displays the URL which is most useful
+#'    for spidering HTML sites.
+#' @examples
+#' url <- "https://ggplot2.tidyverse.org/"
+#' req <- request(url)
+#' req_perform_spider(req, next_reqs = spider_descendents(url))
+req_perform_spider <- function(
+    req,
+    next_reqs,
+    path = NULL,
+    on_error = c("stop", "return", "continue"),
+    hash_key = NULL,
+    progress = TRUE,
+    progress_label = NULL
+) {
+
+  check_request(req)
+  check_function2(next_reqs, args = c("resp", "req"))
+  check_string(path, allow_empty = FALSE, allow_null = TRUE)
+  on_error <- match.arg(on_error)
+  check_function2(hash_key, args = "req", allow_null = TRUE)
+  check_function2(progress_label, args = "req", allow_null = TRUE)
+  check_bool(progress)
+
+  hash_key <- hash_key %||% function(req) req[c("url", "body", "headers")]
+  progress_label <- progress_label %||% function(req) req$url
+
+  get_path <- function(hash) {
+    if (is.null(path)) {
+      NULL
+    } else {
+      glue::glue(path)
+    }
+  }
+
+  todo <- fastmap::fastqueue()
+  done <- fastmap::fastmap()
+  seen <- fastmap::fastmap()
+
+  todo$add(req)
+
+  if (progress) {
+    cli::cli_progress_bar(
+      type = "custom",
+      total = NA,
+      format = "Spidering {done$size()}/{done$size() + todo$size()}: {progress_label(req)}"
+    )
+  }
+
+  while (todo$size() > 0) {
+    req <- todo$remove()
+    if (progress) cli::cli_progress_update()
+
+    req_hash <- hash(hash_key(req))
+    resp <- req_perform(req, path = get_path(req_hash))
+    done$set(req_hash, resp)
+    seen$set(req_hash, TRUE)
+
+    up_next <- next_reqs(req, resp)
+    for (req in up_next) {
+      req_hash <- hash(hash_key(req))
+      if (!seen$has(req_hash)) {
+        seen$set(req_hash, TRUE)
+        todo$add(req)
+      }
+    }
+  }
+
+  unname(done$as_list())
+}
+
+
+#' @export
+#' @rdname req_perform_spider
+spider_descendents <- function(home_url) {
+  force(home_url)
+  function(req, resp) {
+    html <- resp_body_html(resp)
+
+    a <- xml2::xml_find_all(html, "//a[@href]")
+    href <- xml2::xml_attr(a, "href")
+    href <- xml2::url_absolute(href, resp_url(resp))
+    href <- href[map_lgl(href, can_parse)]
+    href <- map_chr(href, strip_fragment)
+    href <- unique(href)
+
+    descendents <- href[map_lgl(href, url_is_child, home_url)]
+
+    map(descendents, function(path) req_url(req, path))
+  }
+}
+
+url_is_child <- function(child, parent) {
+  parent <- url_parse(parent)
+  child <- url_parse(child)
+
+  identical(child$scheme, parent$scheme) &&
+    identical(child$hostname, parent$hostname) &&
+    identical(child$port, parent$port) &&
+    path_is_child(child$path, parent$path)
+}
+
+# path_is_child("/foo2", "/foo")
+# path_is_child("/foo/bar", "/foo")
+path_is_child <- function(child, parent) {
+  parent <- normalize_path(parent)
+  child <- normalize_path(child)
+
+  if (startsWith(child, parent)) {
+    if (nchar(child) > nchar(parent)) {
+      i <- nchar(parent) + 1
+      substring(child, i, i) == "/"
+    } else {
+      FALSE
+    }
+  } else {
+    FALSE
+  }
+}
+
+normalize_path <- function(path) {
+  # strip index.html and friends
+  path <- sub("(index|default)\\.[a-z]+$", "", path, ignore.case = TRUE)
+  # strip trailing /
+  path <- sub("/$", "", path)
+  # url_parse ensures it always starts with /
+  path
+}
+
+strip_fragment <- function(url) {
+  url <- url_parse(url)
+  url$fragment <- NULL
+  url_build(url)
+}
+
+can_parse <- function(url) {
+  tryCatch(
+    {
+      url_parse(url)
+      TRUE
+    },
+    error = function(cnd) FALSE
+  )
+}