-
New functions
monkey_classify()
andmonkey_extract()
that:- Accept as input both a vector and a dataframe and named column
- Always return a tibble explicitly relating each input to its classification, allowing for the removal of the MD5 hash
- Have an
unnest
flag to unnest the output (turn 1 row per input into 1 row per output) - Have a
.keep_all
flag to retain other columns if input is a dataframe - Coerce
NULL
values and empty vectors returned from MonkeyLearn toNA
s - Include inputs that could not be processed as
NA
s in the output - Message the first 20 indices of inputs that are not sent to the API (these now include
NA
andNULL
values as well as empty strings) - Message the currently processing batch
-
Bug fixes and improvements to
monkeylearn_classify()
andmonkeylearn_extract()
monkeylearn_classify()
can now acceptparams
- Fix to messaging when unable to connect to MonkeyLearn API
- Default texts per request is set to 200 now (the recommended number), rather than 20
- Addition of message suggesting that users switch to newer functions
-
Implementation of
ratelimitr
. Creation and documentation of two environment variables allowing smarter rate handling when querying the MonkeyLearn API. -
Creation of
pkgdown
website -
Programmatic test coverage to re-use common tests for multiple circumstances.
-
Use of a
cowsay
monkey when verbose=TRUE.
-
Better states the dependency on tibble, it is tibble >= 1.2.
-
Better handles blank text in input, outputs an empty tibble and a warning if the request is only blank, and a message if only parts of the request are blank.
- Disables HTTP2 for now because of a bug for Windows users. Fix by Jeroen Ooms.
- Added a
NEWS.md
file to track changes to the package.