tarchetypes 0.6.0

tarchetypes 0.4.1

Select list elements from command1 using [[ and not [ in tar_map2() functions.

Implement tar_map_rep() and tar_map_rep_raw() for dynamic batched replication within static branching for data frames (#78).
Implement tar_map2_count(), tar_map2_count_raw(), tar_map2_size(), and tar_map2_size_raw() for batched dynamic-within-static branching for data frames (#78).
Deprecate tar_rep_map() in favor of tar_rep2() to avoid name confusion. Likewise with tar_rep_map_raw() to tar_rep2_raw() (#78).

Allow empty / NULL target list in tar_map() (@kkami1115).
Do not claim to support "aws_file" format in tar_files() or related target factories.

Relax assertion on language objects.
Explain targets timestamps correctly in the help files of tar_age() and tar_cue_age().

When names = NULL in tar_map(), use hashes instead of numeric indexes for generated target names (#67). That way, target names are no longer sensitive to the order of values, and so targets will incorrectly invalidate less often. Unfortunately, this is an invalidating change: some targets will automatically rerun after you install this version of tarchetypes. I apologize for the inconvenience this causes. However, we do need this patch in order to solve #67, and targets will incorrectly invalidate less frequently in the future.

Migrate to utilities for error handling and metaprogramming exported from targets (#59).

Make the *_raw() target factories process command the same way whether it is an expression or ordinary language object.
Ensure compatibility with targets 0.5.0.9000, which logs skipped targets.

Add tar_rep_map() and tar_rep_map_raw() to perform batched computation downstream of tar_rep() (#50).
Add tar_select_names() and tar_select_targets() to make certain metaprogramming tasks easier.
In tar_map(), attempt to convert the elements of values into lists of language objects.

Allow trailing commas in tar_plan() (#40, @kendonB).
Implement tar_age() based on tar_cue_age() (#39, @petrbouchal).
Implement new cue factories tar_cue_age(), tar_cue_age_raw(), tar_cue_force(), and tar_cue_skip() (#39).
Implement tar_download() (#38, @noamross, @petrbouchal)
Set intermediate temporary directory to remove race condition in tar_render_rep() (#36, @gorgitko).
Prefix internal condition classes with “tar_”.
Add new format helpers such as tar_aws_rds() and tar_parquet().
Support hooks tar_hook_before(), tar_hook_inner(), and tar_hook_outer() (#44).
Deep-copy the cue in tar_map().

Add new functions for easier grouping of data frames for dynamic branching: tar_group_by(), tar_group_select(), tar_group_size(), tar_group_count() (#32, @liutiming).
In tar_render() and related functions, track the *_files/ output directory if it exists (#30).
Implement an external walk_ast() function to make it easier for other developers to extend the static code analysis of tarchetypes (@MilesMcBain).

Skip literate programming tests if pandoc is missing or has an insufficient version.
Use explicit temp files in examples even when running inside targets::tar_dir(). (targets::tar_dir() and targets::tar_test() already run code in a temporary directory.)
Add comments in the examples to emphasize that targets::tar_dir() runs code in a temporary directory, which means all ostensibly files created in the enclosed expression will actually be written to temporary storage and not the user’s file space.

Make sure every function with a help file in man/ has Rd-tags \value and \arguments.
For every function with a help file in man/, describe the return value in the \value Rd tag. For each function that returns a target object or list of target objects, the \value tag now links to https://books.ropensci.org/targets/, the user manual where the purpose of target objects is explained, and https://books.ropensci.org/targets-design/, the design specification which documents the structure and composition of target objects.
Ensure that examples, vignettes, and test do not write to the home file space of the user.
Ensure that no function defined in the tarchetypes package writes by default to the home file space of the user. The paths of all output files are controlled by non-tarchetypes functions that invoke tarchetypes.

tar_plan() now returns a list of target objects rather than a pipeline object. Related: https://github.com/ropensci/targets/issues/253.

First version.
Implement tar_knitr_deps() and tar_knitr_deps_expr() to accommodate custom multi-file literate programming projects like R Markdown sites and bookdown projects (#23, @tjmahr).