fix(api): adapt to upstream deprecations

[clason]

• fix(api): replace deprecated Nvim API
• fix: add shim for deprecated vim.F.if_nil

Closes #1436

[clason]

@L3MON4D3 I don't understand the test failure. How old is the Nvim version you're testing here?

[e-kwsm]

How old is the Nvim version you're testing here?

seems 0.7.0 (Apr 15, 2022).

LuaSnip/Makefile

Lines 6 to 18 in a62e108

	NVIM_0.7_PATH_REL=worktree_0.7
	NVIM_0.9_PATH_REL=worktree_0.9
	NVIM_MASTER_PATH=${NVIM_PATH}/${NVIM_MASTER_PATH_REL}
	NVIM_0.7_PATH=${NVIM_PATH}/${NVIM_0.7_PATH_REL}
	NVIM_0.9_PATH=${NVIM_PATH}/${NVIM_0.9_PATH_REL}
	# directory as target.
	${NVIM_PATH}:
	# fetch current master and 0.7.0 (the minimum version we support) and 0.9.0
	# (the minimum version for treesitter-postfix to work).
	git clone --bare --depth 1 https://github.com/neovim/neovim ${NVIM_PATH};
	git -C ${NVIM_PATH} fetch --depth 1 origin tag v0.7.0;
	git -C ${NVIM_PATH} fetch --depth 1 origin tag v0.9.0;

[clason]

That's way too old... It doesn't make sense to try to support anything older than 0.11, IMO.

(I would make a PR, but the whole Nix CI pipeline is too much for me.)

[L3MON4D3]

Hey @clason, TY for the PR :)
I think it's reasonable to drop support for older versions, especially 0.7.0 since it's... four years old 👀
0.9, luckily, still works, so I'm happy with keeping it until it becomes annoying to maintain.
I'll probably add test for 0.12 soon-ish, testing the latest stable version seems very sensible.

I've added another small fix for the vimscript eval, let me know if you have any
concerns about this PR, otherwise I'll merge this tomorrow or so :)

[clason]

Thank you! That was the one thing I was not sure about, so I appreciate your fixing my (too) aggressive change!

(As a data point, I only include "stable" and "nightly" tags in CI; I try not to break the previous (stable-1) versions by intentionally dropping support, but I only guarantee coverage for those two.)

[L3MON4D3]

Thank you! That was the one thing I was not sure about, so I appreciate your fixing my (too) aggressive change!

Anytime, I'm pretty happy the tests caught it :)

(As a data point, I only include "stable" and "nightly" tags in CI; I try not to break the previous (stable-1) versions by intentionally dropping support, but I only guarantee coverage for those two.)

Ah, that's reasonable, maybe I'll do it like this on future plugins :)
IIRC there were a few times where keeping backwards compatibility so far back meant additional work.. it's just a shame it's hard to tell if anyone actually used such an old version and was (positively) affected by that, there are (usually) no issues about software working correctly :P