Skip to contents

Load the return values of targets into the current environment (or the environment of your choosing). For a typical target, the return value lives in a file in _targets/objects/. For dynamic files (i.e. format = "file") the paths loaded in place of the values. tar_load_everything() is shorthand for tar_load(everything()) to load all targets.


  branches = NULL,
  meta = tar_meta(targets_only = TRUE, store = store),
  strict = TRUE,
  silent = FALSE,
  envir = parent.frame(),
  store = targets::tar_config_get("store")



Names of the targets to load. The object supplied to names should be a tidyselect expression like any_of() or starts_with() from tidyselect itself, or tar_described_as() to select target names based on their descriptions.


Integer of indices of the branches to load for any targets that are patterns.


Data frame of metadata from tar_meta(). tar_read() with the default arguments can be inefficient for large pipelines because all the metadata is stored in a single file. However, if you call tar_meta() beforehand and supply it to the meta argument, then successive calls to tar_read() may run much faster.


Logical of length 1, whether to error out if one of the selected targets is in the metadata but cannot be loaded. Set to FALSE to just load the targets in the metadata that can be loaded and skip the others.


Logical of length 1. Only relevant when strict is FALSE. If silent is FALSE and strict is FALSE, then a message will be printed if a target is in the metadata but cannot be loaded. However, load failures will not stop other targets from being loaded.


Environment to put the loaded targets.


Character of length 1, path to the targets data store. Defaults to tar_config_get("store"), which in turn defaults to _targets/. When you set this argument, the value of tar_config_get("store") is temporarily changed for the current function call. See tar_config_get() and tar_config_set() for details about how to set the data store path persistently for a project.



Storage access

Several functions like tar_make(), tar_read(), tar_load(), tar_meta(), and tar_progress() read or modify the local data store of the pipeline. The local data store is in flux while a pipeline is running, and depending on how distributed computing or cloud computing is set up, not all targets can even reach it. So please do not call these functions from inside a target as part of a running pipeline. The only exception is literate programming target factories in the tarchetypes package such as tar_render() and tar_quarto().

Cloud target data versioning

Some buckets in Amazon S3 or Google Cloud Storage are "versioned", which means they track historical versions of each data object. If you use targets with cloud storage ( and versioning is turned on, then targets will record each version of each target in its metadata.

Functions like tar_read() and tar_load() load the version recorded in the local metadata, which may not be the same as the "current" version of the object in the bucket. Likewise, functions tar_delete() and tar_destroy() only remove the version ID of each target as recorded in the local metadata.

If you want to interact with the latest version of an object instead of the version ID recorded in the local metadata, then you will need to delete the object from the metadata.

  1. Make sure your local copy of the metadata is current and up to date. You may need to run tar_meta_download() or tar_meta_sync() first.

  2. Run tar_unversion() to remove the recorded version IDs of your targets in the local metadata.

  3. With the version IDs gone from the local metadata, functions like tar_read() and tar_destroy() will use the latest version of each target data object.

  4. Optional: to back up the local metadata file with the version IDs deleted, use tar_meta_upload().


if (identical(Sys.getenv("TAR_EXAMPLES"), "true")) { # for CRAN
tar_dir({ # tar_dir() runs code from a temp dir for CRAN.
    tar_target(y1, 1 + 1),
    tar_target(y2, 1 + 1),
    tar_target(z, y1 + y2)
}, ask = FALSE)
ls() # Does not have "y1", "y2", or "z".
ls() # Has "y1" and "y2" but not "z".
ls() # Has "y1", "y2", and "z".