Ignore Preview Content

[aoswalt]

Expected behavior:

When opening a preview with pedit (such as with running a database command with vim-dadbod), it should not have an effect on snippets.

Actual behavior:

Depending on the output, vim can freeze until the python process is killed, which seems to be the same as #155.

Steps to reproduce

1. Stop early in a multi-step snippet. In the following snippet, a valid sql statement will exist at the first step.

snippet sel "select (columns:*) from table"
select ${2:*} from ${1} $0
endsnippet

2. Execute statement with dadbod's :DB, which opens a preview window with the result.
3. Navigate to preview window.

Some larger outputs cause the python process to hang and never complete.

Thoughts

I imagine 2 separate ways to solve my issue:

1. Have UltiSnips ignore preview text (No idea how feasible this is)
2. Provide a function that aborts a snippet. This should be able to provide a solution #1017 and mine through an autocommand.

---

• Operating System: Mac OS 10.14.6
• Vim Version: NVIM v0.4.3, Build type: Release
• UltiSnips Version: 3.2
• Python inside Vim: 2.7.17 / 3.8.1
• Docker repo/vimrc: My Full VimRC

[aoswalt]

Here is a sample of database output that causes the issue.

https://gist.github.com/aoswalt/73ce82cf3ce84ebc6caf0508f3333c9a

Instead of executing with :DB, :pedit that file and navigate to it.

[SirVer]

See #681 and #586 for similar issues, especially #586 (comment).

When opening a preview with pedit (such as with running a database command with vim-dadbod), it should not have an effect on snippets.

I disagree with this. pedit leaves the current buffer and should hence terminate all currently in-flux snippets.

It apparently does not do this here, leading to laggy UI. My hunch here is that dadbot's DB is suppressing AutoCommands and hence does not trigger the BufLeave autocommand (though it totally leaves the buffer) and that confuses UltiSnips.

There are multiple avenues to a fix here:

• DB should be changed to properly send BufLeave Autocommand.
• UltiSnips could actually be changed to figure out that the buffer is a different than the last time it's cursor was moved though, so this could be added as a fix.

[aoswalt]

With some more testing, I realize that I conflated the circumstances around the issue.

The freezing only happens when navigating to the preview window. DB is only calling pedit, and I can replicate it with only a raw pedit. The navigation seems to be more the culprit than the pedit itself, even though it does not happen with a standard split.