Setting registers is more in line with VIM

[nnicandro]

This fixes the odd behavior of the yank register (0) and delete registers (1-9) described in #864 and makes their behavior identical to Vim. The following changes were made:

• Add a new variable, evil-delete-kill-ring, which stores the text for the delete registers.
• Remove evil-was-yanked-without-register
• Add a new variable evil-is-yank-and-delete which distinguishes between commands that yank
  and delete text vs commands that just yank text.
• Add the function evil-set-register-on-yank which replaces the register setting logic in the yank
  functions.
• Set the yank register when setting the unnamed register (") for behavior more in line with Vim.

I tried to figure out how I could keep the evil-was-yanked-without-register variable without changing its meaning since I know removing it will break other packages, but it couldn't be done. It was also being used in a confusing way, mainly only to prevent the setting of the yank register in evil-delete.

[wasamasa]

This should be put into a customizable variable, so that other packages can add their own things to it.

[wasamasa]

This looks pretty good, thanks! Regarding breakage, I've done a quick search and it seems that other than people vendoring their packages (which is totally fine), the rest seems to have copy-pasted the yank commands. They'll have to update them to make use of your code. Nothing much we can do about that.

[nnicandro]

An update on this, mainly I've tried to get the behavior as close to Vim's as possible. I think it is essentially identical now.

• I've renamed evil-set-register-on-yank to evil-kill-new. Essentially this function was kill-new, but also handled the behavior of registers.

• Handle the small delete register ?- in evil-kill-new instead of evil-delete, also make its behavior more in line with Vim's. That is, do not set register ?1 if register ?- is being set. Set both registers if the motion is one of evil-special-delete-motions.

• Add documentation describing the behavior of registers in evil-kill-new. Mention evil-kill-new in the documentation of evil-yank and evil-delete.

• Give an initial value to evil-delete-kill-ring and fix behavior when setting one of the delete registers. The previous behavior would set the wrong register if evil-delete-kill-ring was not at its maximum size.

• Removed the variable evil-is-yank-and-delete in favor of using this-command set to evil-delete to distinguish between a yank command or a yank followed by a delete, i.e. a delete command. We can make this distinction by simply assuming that the text passed to evil-kill-new is due to a yank command if this-command is not evil-delete and if this-command is evil-delete then the text is due to a delete command.

  • Actually it probably makes more sense to make the default be that the text is due to a delete command and make the check for evil-yank since killing text in Emacs is essentially the same as deleting text in Vim.

[nnicandro]

Force pushed to squash all of the WIP commits.

[syl20bnr]

@wasamasa do you think we can merge this ? This is a common issue with new Evil users coming from Vim.

[syl20bnr]

@DZoP Thank you for such a great PR. There is a core dump with Emacs 26.1 in Travis. Which version of Emacs are you using ?

[TheBB]

I think @wasamasa has withdrawn from active participation. I'll see if I can devote some time to this, and other PRs.

[nnicandro]

I'm using

GNU Emacs 26.1 (build 2, x86_64-apple-darwin18.0.0, Carbon Version 158 AppKit 1671) of 2018-12-27

but I think the issue is related to #945, running make test works fine. I triggered another Travis build to see if it goes through this time.

[nnicandro]

Friendly bump.

I've been using this change without issue for a long time now, but I think before it gets merged I should address the removal of the evil-was-yanked-without-register variable which breaks packages like lispyville and evil-collection.

Maybe it would be better to keep that variable around for compataility...

[leungbk]

@DZoP @TheBB What's the status on this? It would be nice to see this get merged at some point.

Friendly bump.

I've been using this change without issue for a long time now, but I think before it gets merged I should address the removal of the evil-was-yanked-without-register variable which breaks packages like lispyville and evil-collection.

Maybe it would be better to keep that variable around for compataility...

evil-collection uses it once, in reference to evil-mc, which has evil-was-yanked-without-register as part of a list of variables of interest, but does not otherwise record or make use of it.

The author of lispyville mentioned this change as something he would keep an eye on. I wouldn't mind making a PR to lispyville once this gets merged.

[axelf4]

I think these are good changes, and it would be nice to get it merged. (It is unfortunate that text deleted by delete or change commands cannot use the regular kill-ring instead of evil-delete-kill-ring, with "" being the special register. After all, it is kill-ring and not copy-ring, and you would then remain able to access text yanked from Emacs commands using the registers "1 through "9. But I understand that would complicate all paste commands.)

The docstrings could also perhaps be made a bit more understandable.

[nnicandro]

I've re-worked the documentation and addressed the issues that you mentioned. Please give this another look.

I did have to fix some tests that relied on the current behavior of the 1-9 registers. I'm thinking that people may already be used to how those registers worked before this change in that they stored yanked as well as deleted text. Perhaps there should be an option to revert back to the old behavior of those registers before this change? This would also address your concern about using the 1-9 registers to access text yanked by other Emacs commands.