RC for initial CRAN release (#47)

* Update DESCRIPTION

* Make README shorter, mostly

* Tiny change to docs

* h2o function is now exported 🎉

* Update docs for torch. Related to #36.

* Update title

* Tiny doc changes

* Tiny changes to docs

* Redocument

* Remove the progress bar removal, skip loooooong examples

* Redocument

* Skip if no tensorflow

* Don't run examples + vignette on CRAN

* Need tensorflow to check if... tensorflow is installed 😫

* Redocument

* Better example checks

* Redocument

* Better title case

* Can't check for tensorflow unless R package is installed

* Don't use "native" to mean different things in the pkg

* Light refining of main vignette

* Redocument AGAIN

* Return values for internal functions

* Say "interface", not "API"

* Use single quotes for 'R'

* More carefully check for torch

* Redocument

* Ignore

Author: juliasilge

.Rbuildignore

@@ -10,3 +10,4 @@
 ^_pkgdown\.yml$
 ^docs$
 ^pkgdown$
+^CRAN-SUBMISSION$

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: bundle
-Title: Serialize Objects With A Consistent Interface
-Version: 0.0.0.9200
+Title: Serialize Model Objects with a Consistent Interface
+Version: 0.1.0
 Authors@R: c(
     person("Julia", "Silge", , "[email protected]", role = c("aut", "cre"),
            comment = c(ORCID = "0000-0002-3671-836X")),
@@ -9,11 +9,11 @@ Authors@R: c(
     person("Max", "Kuhn", , "[email protected]", role = "aut"),
     person("RStudio", role = c("cph", "fnd"))
   )
-Description: R holds most objects in memory. However, some models store
-    their data in locations that are not included when one uses `save()`
-    or `saveRDS()`.  bundle provides a common API to capture this
-    information, situate it within a portable object, and restore it for
-    use in new settings.
+Description: Typically, models in 'R' exist in memory and can be saved via
+    regular 'R' serialization. However, some models store information in 
+    locations that cannot be saved using 'R' serialization alone. The goal of 
+    'bundle' is to provide a common interface to capture this information, 
+    situate it within a portable object, and restore it for use in new settings.
 License: MIT + file LICENSE
 URL: https://github.com/rstudio/bundle, https://rstudio.github.io/bundle/
 BugReports: https://github.com/rstudio/bundle/issues
@@ -43,6 +43,7 @@ Suggests:
     rmarkdown,
     stacks,
     testthat (>= 3.0.0),
+    tensorflow,
     torch,
     torchvision,
     uwot,

R/bundle.R

@@ -4,9 +4,8 @@
 #'
 #' @description
 #' `bundle()` methods provide a consistent interface to serialization
-#' methods for statistical model objects. The outputted bundle can be saved,
-#' re-loaded into a new R session, and `unbundle()`d in a new R session for
-#' use in prediction.
+#' methods for statistical model objects. The created bundle can be saved,
+#' then re-loaded and `unbundle()`d in a new R session for use in prediction.
 #'
 #' @templateVar outclass referencing the modeling function
 #' @templateVar default . If a bundle method is not defined for the supplied object, `bundle.default` is the identity function.

R/bundle_caret.R

@@ -13,7 +13,7 @@
 #'   `train_model_object$finalModel`. See the class of the output of that
 #'   slot for more details on the bundling method for that object.
 #' @template butcher_details
-#' @examplesIf rlang::is_installed("caret")
+#' @examplesIf rlang::is_installed("caret") && identical(Sys.getenv("NOT_CRAN"), "true")
 #' # fit model and bundle ------------------------------------------------
 #' library(caret)
 #'

R/bundle_h2o.R

@@ -16,7 +16,7 @@
 #' @template param_unused_dots
 #' @seealso These methods wrap [h2o::h2o.save_mojo()] and
 #'   [h2o::h2o.saveModel()].
-#' @examplesIf rlang::is_installed("h2o") && rlang::is_installed("MASS")
+#' @examplesIf rlang::is_installed(c("h2o")) && identical(Sys.getenv("NOT_CRAN"), "true")
 #' # fit model and bundle ------------------------------------------------
 #' library(h2o)
 #'
@@ -80,9 +80,9 @@ bundle_h2o <- function(x, ...) {
   file_loc <- withr::local_tempdir(pattern = "bundle")
 
   if (x@have_mojo) {
-    file_loc <- with_no_progress(h2o::h2o.save_mojo(x, path = file_loc))
+    file_loc <- h2o::h2o.save_mojo(x, path = file_loc)
   } else {
-    file_loc <- with_no_progress(h2o::h2o.saveModel(x, path = file_loc))
+    file_loc <- h2o::h2o.saveModel(x, path = file_loc)
   }
   raw <- readBin(file_loc, "raw", file.size(file_loc), endian = "little")
 
@@ -92,9 +92,9 @@ bundle_h2o <- function(x, ...) {
       new_file <- withr::local_tempfile(pattern = "unbundle")
       writeBin(object, new_file, endian = "little")
       if (!!x@have_mojo) {
-        res <- h2o:::with_no_h2o_progress(h2o::h2o.import_mojo(new_file))
+        res <- h2o::h2o.import_mojo(new_file)
       } else {
-        res <- h2o:::with_no_h2o_progress(h2o::h2o.loadModel(new_file))
+        res <- h2o::h2o.loadModel(new_file)
       }
 
       res
@@ -115,7 +115,3 @@ select_from_automl <- function(x, id = NULL, n = NULL) {
   }
   x
 }
-
-with_no_progress <- function(expr) {
-  rlang::eval_tidy(h2o:::with_no_h2o_progress(expr))
-}

R/bundle_keras.R

@@ -15,7 +15,7 @@
 #' @details This bundler does not currently support custom keras extensions,
 #'   such as use of a [keras::new_layer_class()] or custom metric function.
 #'   In such situations, consider using [keras::with_custom_object_scope()].
-#' @examplesIf FALSE
+#' @examplesIf rlang::is_installed(c("keras")) && identical(Sys.getenv("NOT_CRAN"), "true")
 #' # fit model and bundle ------------------------------------------------
 #' library(keras)
 #'

R/bundle_stacks.R

@@ -13,7 +13,7 @@
 #' @details This bundler wraps [bundle.model_fit()] and [bundle.workflow()].
 #' Both the fitted members and the meta-learner (in `x$coefs`) are bundled.
 #'
-#' @examplesIf rlang::is_installed(c("stacks"))
+#' @examplesIf rlang::is_installed(c("stacks")) && identical(Sys.getenv("NOT_CRAN"), "true")
 #' # fit model and bundle ------------------------------------------------
 #' library(stacks)
 #'

R/bundle_torch.R

@@ -15,15 +15,11 @@
 #' via the luz package, "a higher level API for torch providing
 #' abstractions to allow for much less verbose training loops."
 #'
-#' These bundlers rely on serialization methods from luz and torch,
-#' which are [described by the package authors][torch::torch_save]
-#' as "experimental" and not for "use for long term storage."
-#'
 #' @method bundle luz_module_fitted
 #' @rdname bundle_torch
-#' @seealso This method adapts [luz::luz_save()] and the internal luz function
-#'   `model_to_raw()`, as well as [torch::torch_save()].
-#' @examplesIf FALSE
+#' @seealso This method wraps [luz::luz_save()] and [luz::luz_load()].
+#' @examplesIf rlang::is_installed(c("torch")) && identical(Sys.getenv("NOT_CRAN"), "true")
+#' if (torch::torch_is_installed()) {
 #' # fit model and bundle ------------------------------------------------
 #' library(torch)
 #' library(torchvision)
@@ -99,6 +95,7 @@
 #' mod_unbundled <- unbundle(mod_bundle)
 #'
 #' mod_unbundled_preds <- predict(mod_unbundled, test_dl)
+#' }
 #'
 #' @aliases bundle.luz_module_fitted
 #' @export

R/package.R

@@ -1,4 +1,4 @@
-#' bundle: Serialize Objects With A Consistent Interface
+#' bundle: Serialize Model Objects With A Consistent Interface
 #'
 #' @docType package
 #' @name bundle_description

R/utils.R

@@ -5,6 +5,10 @@
 #' These functions are not user-facing and are only exported for developer
 #' extensions.
 #'
+#' @return The two `_constr()` functions are constructors that return a bundle
+#' and a situater, respectively. `swap_element()` returns `x` after swapping
+#' out the specified element.
+#'
 #' @rdname internal_functions
 #' @keywords internal
 #' @export