Closed DavisVaughan closed 1 year ago
Merging #197 (53233fd) into master (bc8cf53) will decrease coverage by
0.10%
. The diff coverage is85.71%
.:exclamation: Current head 53233fd differs from pull request most recent head 14b2dc8. Consider uploading reports for the commit 14b2dc8 to get more accurate results
:mega: This organization is not using Codecov’s GitHub App Integration. We recommend you install it so Codecov can continue to function properly for your repositories. Learn more
@@ Coverage Diff @@
## master #197 +/- ##
===========================================
- Coverage 100.00% 99.90% -0.10%
===========================================
Files 15 15
Lines 1062 1057 -5
===========================================
- Hits 1062 1056 -6
- Misses 0 1 +1
Impacted Files | Coverage Δ | |
---|---|---|
R/utils.R | 100.00% <ø> (ø) |
|
R/spatial.R | 98.85% <66.66%> (-1.15%) |
:arrow_down: |
R/dates.R | 100.00% <100.00%> (ø) |
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.
Thank you 👍
This should be compatible with both dev and CRAN dplyr. It would help us out if you could go ahead and send a patch version of your package to CRAN ahead of time! Thanks!
We'll package a release with some other developments but I think it should work out timewise.
👍 looking like March 13 for the dplyr release
We are preparing to release dplyr 1.1.1 and this package popped up in the revdep checks.
We realized that we didn't get the
multiple
argument quite right the first time around 😞 . It was too aggressive since it warned on both one-to-many and many-to-many joins. In 1.1.1 we've made two improvements:multiple
now defaults to"all"
. The options ofNULL
,"error"
, and"warning"
are deprecated.relationship
is a new argument to add known constraints onto the join procedure, such as"many-to-one"
, which replacesmultiple = "error"
.The default of
relationship
checks to see if there is amany-to-many
relationship between the keys ofx
andy
and will warn if one is present. This should be much rarer than what we checked for before, and targets the most dangerous case that we were trying to warn the user about.You can read all about
relationship
here https://github.com/tidyverse/dplyr/pull/6753, along with the issues linked there.Unfortunately it does affect some code here, but I think it does so in a positive way!
set_dates_service()
which makes it faster and simplerinner_join()
inget_trip_geometry()
no longer needsmultiple
to be set. It seems like it was doing a one-to-many join before, which we no longer warn on. But we do have to branch on the dplyr version for this one to support CRAN and dev dplyr.We plan to submit dplyr 1.1.1 in 2-3 weeks.
This should be compatible with both dev and CRAN dplyr. It would help us out if you could go ahead and send a patch version of your package to CRAN ahead of time! Thanks!