- Fixed a bug with vcr testing prompted by the rtweet update to v1.1 (#90)
- Fixed another bug in
get_upstream_tweets()where the columnpossibly_sensitivewas sometimes read in as a list, other times as a logical type. - Addressed a bug in vcr v1.1 that was causing tidytags tests to error in the CRAN logs (https://cran.r-project.org/web/checks/check_results_tidytags.html).
- The
setup-vcr.Rfile was renamed tohelper-vcr.Rso that it will not be loaded when someone simply installs tidytags, only when package developers compile and test. - In addition, the vcr bug has been addressed by the developers of that package, so vcr v1.2 is now the minimum suggested version for tidytags.
- Fixed a bug in
get_upstream_tweets()where column names were out of order and caused an error
- Fixed broken URLs and reduced tarball size in preparation for CRAN resubmission.
- In function documentation, \dontrun{} instances have been updated to \donttest{}.
- Cleaned up documentation in preparation for CRAN submission.
- Updated Twitter authentication process to align with breaking changes caused by the rtweet 1.0 release.
- Updated the process_tweets() function to align with changes in available metadata and new variable names used in rtweet 1.0.
- Removed the lookup_many_users() function. With the rtweet 1.0 update, user information can be accessed with the rtweet::users_data() function.
- Updated flag_unknown_upstream() and get_upstream_tweets() to align with new variable names used in rtweet 1.0.
- Updated filter_by_tweet_type(), create_edgelist(), and add_users_data() to align with new variable names used in rtweet 1.0.
- Removed the geocode_tags() function because rtweet 1.0 changed how location data is available and also added a new rtweet::lookup_coords() function. Note that at this time, rtweet::lookup_coords() requires a Google Maps API key rather than the OpenCage API we had recommended in earlier versions of tidytags. We still recommend the sf and mapview R packages for working with locations and geocoding.
- Updated the read_tags() function so that a Google API key is no longer needed to pull tweet data from publicly shared Google Sheets.
- The process_tweets() function now also adds user information associated with the creator of each status. process_tweets() also now returns a column for the tweet type of each status.
- Updated to most recent versions of CI tests for R-CMD-check and test coverage.
- Updated paper.md and paper.bib to coincide with submission for peer review at Journal of Open Source Software (JOSS).
- Added a new function filter_by_tweet_type() to filter a Twitter dataset to only include statuses of a particular type (e.g., replies, retweets, quote tweets, mentions).
- Updated the function create_edgelist() to take a "type" argument (e.g., "reply", "retweet", "quote", "mention", "all"). This replaces the need for specialized functions like create_mentions_edgelist().
- Added a new function lookup_many_users() to automatically iterate through the Twitter API limit of pulling metadata for only 90,000 users at one time
- Updated several function names so as not to mask newer functions imported from {rtweet}, for example, get_mentions() is now create_mentions_edgelist(), and similar updates have been made for function building edgelists from quotes, replies, and retweets
- Updated tests to work with latest version of {vcr}
- Made fixes so CI tests would again work with real requests in addition to pre-recorded {vcr} data
- Extensively updated the README doc and Setup vignette to help scaffold {tidytags} setup
- CI tests now work with real requests in addition to pre-recorded vcr data
- Added a Google API key for accessing a Google Sheet with
read_tags()
- Clarified process for obtaining an setting up API keys and tokens for Google, Twitter, and OpenCage
- Switched to OpenCage for geocoding (previously used Google Maps API)
- Switched to GitHub Actions (from Travis CI) for CI testing
- Initial release on GitHub.