Clarify differences between count() and tally() #5349
Conversation
And revert count() to 0.8.5 behaviour. Fixes #5163
hadley
changed the title
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
|
@lionel- could you check my reasoning here please? |
There was a problem hiding this comment.
count() and tally() have always behaved differently because they use different default values (NULL vs missing): dc92c24
The historical behaviour is:
tally()guesses the weight variablecount()doesn't guess.
This behaviour was preserved when we switched to tidy eval. However it was inadvertently changed in 0.8.2 with #4408. Since that version (almost one year old) neither count() nor tally() try to guess the weighting column.
I think not guessing is the expected behaviour for most people. I also like that count and tally are consistent. Maybe we should sanction the 0.8.2 behaviour that the weight column is never guessed? In that case, we don't need the guess_wt() sentinel.
|
Ok, if we accidentally broke the autoguessing in 0.8.2, then lets remove it all together. |
|
@yutannihilation since you have an eye for detail, would you mind taking a look at this PR. Most importantly, is the reasoning in the NEWS clear, and does it make sense to you? Thanks! |
|
Looks good! Just one thing to confirm, is it safe to keep using |
|
Good point, probably worth keeping that check around just as a precaution. |
| (#5324). | ||
| * `count()` and `tally()` no longer automatically weights by column `n` if | ||
| present (#5298). dplyr 1.0.0 introduced this behaviour because of Hadley's | ||
| faulty memory. Historically `tally()` automatically weighted and `count()` |
And revert count() to 0.8.5 behaviour. Fixes #5298