remove soft deprecated dplyr-src funs

[pawelru]

close #999

I have found many other calls to those functions but I decided to leave them - here goes my explanation:

• dm_get_src()
  

  • R/ - only deprecated function itself
  • scratch - I have left your notes untouched
  • tests/ - as you soft-deprecated then that means that you still expect it to work - would be good to keep that check

• tbl()
  

  • I left demo/ directory as it is as I assume it's for demoes that already happened so this code is unlikely to be executed again. If yes - should be done on the current version at that point in time.
  • R/ - only comments and deprecated function itself
  • tests/ - same as above

• src_tbls()
  

  • demo/ - as above
  • I think that the last occurence is intended to call dplyr::src_tbls (would be good to prefix though) - please confirm

• copy_to()
  

  • R/ - same as above
  • tests/ - same as above
  • vignettes/ - here I probably need your help with assessment. Should it stay or not?

[krlmlr]

Thanks a lot! Is there a reason to prefer [[ over $ ?

[krlmlr]

Suggested change

	# Only need to compute `dm[[start]]`, `dm_apply_filters()` not necessary
	# Only need to compute `dm$start`, `dm_apply_filters()` not necessary

[pawelru]

Thanks a lot! Is there a reason to prefer [[ over $ ?

Yes. [[ works with variables storing a string wheras $ doesn't. E.g. here.

Is start a variable here (that stores the string) or it is a string (which then should be a "start")? Current code indicates that it's a variable.

If it's a variable then it needs to stay as dm[[start]] and if it a string then it could be changed into dm$start.

[krlmlr]

The demos are part of R CMD check, we will catch when things start to fail. But then they are only of historic interest, so if this happens we could prune all except the most recent one.

[krlmlr]

For the vignettes we'd want to look for idiomatic replacements. Happy to provide guidance as needed, pinging @TSchiefer and @moodymudskipper too.

[pawelru]

Thanks a lot! Is there a reason to prefer [[ over $ ?

Nothing really strong. To me, [[ deals better when list element name contains a "$". You would then have to add some back-ticks or quite which doesn't look good right after $. But this is rather my personal preference - you are the owner so I would do what you prefer

> x <- list("a$b" = 1, a = 2, b = 3)
> x$a$b
Error in x$a$b : $ operator is invalid for atomic vectors
> x$`a$b`
[1] 1
> x$"a$b"
[1] 1
> x[["a$b"]]
[1] 1

[krlmlr]

Thanks! Let's get this merged so that future PRs don't require confirmation to run workflows.

[krlmlr]

Thanks!