dtplyr release?

[hadley]

Any concern with me doing a dtplyr release? (It's broken by dplyr 1.1.0)

Should this be a patch version or a minor version?

[markfairbanks]

I think it's ready for a minor version. There are a bunch of new functions/bug fixes as well as some performance improvements for select() and arrange().

Only possible negative is .by isn't implemented yet? I don't think that's necessarily a blocker though.

[hadley]

If someone could implement .by soon (maybe in the next week?) I think we could hold off; otherwise I think it's fine to include that in the next minor release.

[markfairbanks]

If someone could implement .by soon (maybe in the next week?) I think we could hold off;

Yeah that should be doable, I'll take a look.

[hadley]

@markfairbanks also would mind giving me a list of what you think the biggest new features are? (i.e. what are the things that I should highlight in the blog post?)

[markfairbanks]

Yep, will do. Also FYI I took an initial pass at .by and I'll have a PR by Friday.

[hadley]

Awesome, thanks!

[markfairbanks]

Here's the list. I grabbed this from the reorganized NEWS from #410. Not sure if all of these should be in the blog post but they're the most notable.

New features

• New translations:

  • add_count()
  • unite()

• .by/by has been implemented for mutate(), summarise(), filter(),
  and the slice() family (Implement .by #399)

• min_rank(), dense_rank(), percent_rank(), & cume_dist() are now mapped
  to their data.table equivalents (Translate min_rank, dense_rank, percent_rank, & cume_dist #396)

• pick() is now translated (Implement pick() and use across() output as a data frame #341)

• across() output can now be used as a data frame (Implement pick() and use across() output as a data frame #341)

Performance improvements

• select() now drops columns by reference when possible for improved performance (Fast select() #367)

• arrange() now utilizes setorder() when possible for improved performance (Faster arrange() #364)

• slice() uses an intermediate variable to reduce computation time of row selection (slice() unnecessarily evaluates ... twice for each group #377)

Minor improvements and bug fixes

• dplyr and tidyr verbs no longer dispatch to dtplyr translations when used
  directly on data.table objects. lazy_dt() must now explicitly be called by
  the user. (S3 method overloading on namespace load #312)

[hadley]

Thanks!

I'd suggest moving the last point to a new "breaking changes" heading at the top of the news for this release. I don't think it will affect many people, but it's good to advertise prominently.

[markfairbanks]

Sounds good

[hadley]

Closing this issue and about to open an official release issue.