NEWS

tibblify 0.4.1.9000

tib_*() functions now use .fill when a field is NULL or otherwise has length 0 (e.g., list() or character()), treating it the same as an absent field (#231).
.fill arguments now accept values that can be losslessly coerced to the expected type, such as "1" when an integer or double is expected (#337, #340).
tibblify() now accepts values that can be losslessly coerced to the expected type, such as "1" when an integer or double is expected, and "false lists" that trivially unlist() to the expected type (such as list(list("a"))) (#330).

All arguments of functions that accept meaningful named ... are now prefixed with . to minimize conflicts with column and object names in .... The un-dotted versions of the arguments are still accepted, but calling functions directly with un-dotted arguments will produce a warning once per session (see ?lifecycle::deprecate_soft()). Un-dotted arguments will be phased out in a future version of this package, so we recommend switching to the dot-prefixed versions. See ?tspec_df and ?tib_scalar() for details (#228).

All code has been refactored for maintainability. While we were careful to ensure that output is unchanged, it is possible that a corner case is no longer handled how it was in version 0.3.0. Please notify us (https://github.com/wranglezone/tibblify/issues) if something has changed for the worse in an unexpected way (#243).

The guess_tspec() variants guess_tspec_list() and guess_tspec_object_list() are now exported (along with guess_tspec_df() and guess_tspec_object(), which were already exported). guess_tspec() should correctly guess the format in most cases, but you can call the variant directly if you think guess_tspec() is dispatching incorrectly (#249).
untibblify() now automatically uses the tib_spec attribute when present, so tibblified objects can be round-tripped without explicitly passing the spec (#235).
parse_openapi_spec() supports many more fields and works for many more APIs (#170, #186, #190, #200, @jonthegeek and @mgirlich).

The underlying C implementation has been updated to comply with R's C API. We also fixed various bugs during this update (#203, #204, #222).

All vignettes and the documentation of all functions have been updated for clarity (#243).

New parse_openapi_spec() and parse_openapi_schema() to convert an OpenAPI specification to a tibblify specification.
Fix ptype of a tib_vector() inside a tib_df().
New unpack_tspec() to unpack the elements of tib_row() fields (#165).
Improved printing of lists parsed with tspec_object().
Improved performance of the tspec() family.
Improved guessing.

In column major format all fields are required.
Fixed a memory leak.
tib_vector() is now uses less memory and is faster.
tspec_*(), tib_df(), and tib_row() now discard NULL in .... This makes it easier to add a field conditionally with, for example tspec_df(if (x) tib_int("a")).
tib_variant() and tib_vector() give you more control for transforming:
- transform is now applied to the whole vector.
- There is a new elt_transform argument that is applied to every element.
New tspec_recursive() and tib_recursive() to parse tree like structure, e.g. a directory structure with its children.

Major rewrite of the tibblify package with lots of benefits:

tibblify() is now implemented in C and therefore way faster.
Support of column major format.
Support for vectors as scalar lists and objects.
Specification functions have been renamed
- lcols() to tspec_df()
- new specs tspec_object() and tspec_row()
- lcol_int() to tib_int() etc
tspec_df() gains an argument .names_to to store the names of a recordlist in a column.
Added untibblify() to turn a tibble into a nested list, i.e. to reverse the action of tibblify().
Added spec_combine() to combine multiple specifications.
Added argument unspecified to tibblify() to control how to handle unspecified fields.
Many bugfixes.