• zip_list() now has a type column, for the file type.

• unzip() now correctly creates symbolic links on Unix (#127).

• zip_list() now returns a tbl object, and loads the pillar package,
  if installed, to produce the nicer output for long data frames.

• The zip shared library now hides its symbols (on platforms that support
  this), to avoid name clashes with other libraries (#98).

• zip now handles large zip files on Windows (#65, #75, #79, @weshinsley).

• zip now behaves better for absolute paths in mirror mode, and when the
  paths contain a : character (#69, #70).

• zip::unzip() now uses the process's umask value (see umask(2)) on Unix
  if the zip file does not contain Unix permissions (#67).

• Fix segmentation fault when zip file can't be created (#91, @zeehio)

• Fix delayed evaluation error on zipfile when zip::zip()
  is used (#92, @zeehio)

• New deflate() and inflate() functions to compress and uncompress
  GZIP streams in memory.

• No user visible changes.

• No user visible changes.

• Header values (of version made by and external attributes) are now
  correctly read and written on big-endian systems (#68).

• zip_list() now also returns crc32 and offset (#74, @jefferis).

This version has no user visible changes.

• unzip_process() now does not fail randomly on Windows (#60).

• Now all functions handle Unicode paths correctly, on Windows
  as well (#42, #53).

• unzip_process() now works when R library is on different drive
  than exdir on Windows (#45)

• zip functions now have a mode argument to choose how files and
  directories are assembled into the archive. See the docs for
  details.

• zip functions now have a root argument, zip changes the working
  directory to this before creating the archive, so all files are
  relative to root.

• zip() and zip_append() are not deprecated any more, as it was
  hard to achieve the same functionality with the other zip functions.

• unzip_process() prints better error messages to the standard error,
  and exits with a non-zero status, on error.