implement :echomsg! #1802

justinmk · 2015-01-12T09:40:50Z

Followup of my comments here: #1524 (comment) , #1524 (comment) , #1524 (comment)

Closes #2063
Closes #5103
ref #2308 #1755 #1029
ref #7291
ref #7290
ref #7326

This PR implements "bang" variants of :echomsg ~~and :echoerr~~ (edit: not going to implement :echoerr!; the semantics of :echoerr differ greatly from :echomsg, and have implications for try/catch--it wouldn't make sense for :echoerr to be "passive"). Documented as follows:

+                                                     *:echom!* *:echomsg!* 
+           |:echomsg|  followed by [!] creates a
+           "transient" message, which will never cause a screen
+           scroll or a |press-enter| prompt, and is _always_
+           added to the |:messages| history (even if |:silent| or
+           |:map-<silent>| are used). This is useful for leaving
+           passive diagnostic messages without disturbing
+           usability.

Besides solving a general usability issue (avoiding the "Press ENTER" prompt for non-critical messages), this enables us to be much more consistent and less hesitant about informing the user of possibly-useful context-sensitive discoveries while bootstrapping official plugins, providers (python, clipboard), and general sanity checks (VIMRUNTIME).

Before writing tests, I would like feedback on the general approach. I usually advocate against extending vimscript, but after much thought I could not find a reasonable medium-term solution. Some ideas that I rejected:

extend histadd() to allow adding to the :messages history.
- This is icky becaus then we have to also implement histdel() or introduce a glaring incongruity.
expose :messages list as a read-write v:messages List.
- Same problem as above: giving plugins the ability to delete from :messages is a bad idea.
create an API-only function for adding to history
- This is fine, but I suspect that users will really appreciate being able to easily avoid the "Press ENTER" prompt.
add a new flag to 'shortmess'
- This is still an option that I think is worth exploring, in addition to providing :echomsg! and :echoerr!. But it does not solve the use cases mentioned above (providers, sanity checks).

:echomsg! spec

"+N :messages" spec

test overflow (very large number of notifications)
displays count after event loop
count resets after next user action
showcmd blanks out notification area
appears after all messages posted (until next user command)
appears only if 2 or more non-duplicate messages since last batch
appears in correct location

tarruda · 2015-01-12T11:03:51Z

Before writing tests, I would like feedback on the general approach. I usually advocate against extending vimscript, but after much thought I could not find a reasonable medium-term solution. Some ideas that I rejected:

Why not simply use the screen verification facility for testing this? Here's an example

aktau · 2015-01-13T22:20:42Z

src/nvim/memory.c

-  for (;; )
-    if (delete_first_msg() == FAIL)
-      break;
+  while(msg_delete_first()) {;}


As far as I've seen this is the first instance of the {;} pattern, I though we were using just {}.

ghost · 2015-01-29T22:28:02Z

@justinmk I see that you're including some changes to message.h in this PR, perhaps you could pull https://github.com/Pyrohh/neovim/commit/374d8200bfc2690232917228590862371f5dff13 in as well? I can just submit a PR if it's too far reaching.

justinmk · 2015-02-16T23:17:05Z

test/functional/ex_cmds/echomsg_spec.lua

+  end)
+
+
+  local function echomsg_bang_tests(cmd_under_test)


Tests for new behavior of :echomsg! (:echomsg with ! bang). This test group is executed twice for shortmess-=T and shortmess+=T.

justinmk · 2015-02-18T05:29:20Z

@tarruda After rebasing, I have an unexpected test failure in this PR:

row 1:
  expected: "^                                                    "
  actual:   "                                                     "

If I change the expected ^ in row 1 to a <space>, then the result is reversed:

row 1:
  expected: "                                                     "
  actual:   "^                                                    "

The test is testing the new :echomsg! behavior, so it must be something with how my changes interact with the TUI changes.

tarruda · 2015-02-18T20:20:47Z

@justinmk I tested this branch, and it seems you are experiencing a race condition. This patch should fix it:

diff --git a/test/functional/ex_cmds/echomsg_spec.lua b/test/functional/ex_cmds/echomsg_spec.lua
index 0cea285..cf05cba 100644
--- a/test/functional/ex_cmds/echomsg_spec.lua
+++ b/test/functional/ex_cmds/echomsg_spec.lua
@@ -523,6 +523,7 @@ describe('echomsg', function()
       _h.nvim('set_option', 'cmdheight', 2)
       execute('nnoremap <silent> foo :silent '..cmd_under_test..' "'..fullmsg..'"<cr>')
       _h.feed('foo')
+      _h.wait()
       screen:expect([[
       ^                                                    |
       ~

wait() It is simply a way of saying: "Wait until all keys I've sent are fully processed". This is required because the _h.nvim('set_option', 'cmdheight', 2) may set the screen to the state you expect it to be later, and the first redraw notification will succeed, only to be followed by another that fails(sent while foo is being processed)

justinmk · 2015-02-18T20:55:40Z

@tarruda Thanks for looking into that. I was being paranoid so I didn't want to just add a wait() to make it go away, but now that you explain it I suppose there is no reason to expect any particular sequence of redraws until all keys are processed.

This is to ensure ctrl+c is only pressed after the command has started executing in the viml_system_spec.lua system() interrupt test.

see c546875

fmoralesc · 2015-05-30T18:01:35Z

Any progress in this? #2769 would benefit a lot from this, for example.

justinmk · 2015-05-30T21:13:53Z

@fmoralesc I'm thinking the OS X builds are higher priority, but then this was my next priority. The CI failures are out of control...

fmoralesc · 2015-05-30T21:18:13Z

Indeed! Carry on... I was just reminded of this when reviewing #2769, the messages there should probably use something like silent echomsg!.

justinmk · 2025-05-03T19:18:52Z

Closing in favor of #27855

marvim added the WIP label Jan 12, 2015

aktau reviewed Jan 13, 2015
View reviewed changes

fwalch mentioned this pull request Jan 25, 2015

Neovim and virtualenv... #1887

Closed

justinmk force-pushed the echom_bang branch from d29142c to 87e82ce Compare January 29, 2015 07:22

justinmk added enhancement ux user experience labels Jan 29, 2015

justinmk mentioned this pull request Feb 8, 2015

RFC: test: demonstrate vim_command_output() quirk #1959

Closed

justinmk force-pushed the echom_bang branch from 4a382fb to e19816e Compare February 11, 2015 09:00

justinmk mentioned this pull request Feb 13, 2015

abstract_ui vs error messages #1984

Closed

justinmk changed the title ~~[WIP] implement :echomsg!, :echoerr!~~ [WIP] implement :echomsg! Feb 15, 2015

justinmk force-pushed the echom_bang branch from 47454e6 to b118efc Compare February 16, 2015 00:14

justinmk reviewed Feb 16, 2015
View reviewed changes

justinmk force-pushed the echom_bang branch 2 times, most recently from ad1bfd9 to 38616c8 Compare February 18, 2015 05:08

justinmk mentioned this pull request Feb 20, 2015

Nvim doesn't resize itself in certain situations #2026

Closed

tarruda mentioned this pull request Feb 20, 2015

Flaky test: shell/viml_system_spec. #2001

Closed

tarruda referenced this pull request in tarruda/neovim Feb 21, 2015

test: Add extra screen expectation to prevent race condition

f9d327c

This is to ensure ctrl+c is only pressed after the command has started executing in the viml_system_spec.lua system() interrupt test.

justinmk mentioned this pull request Feb 28, 2015

[RDY] Builtin terminal emulation #2076

Merged

ghost mentioned this pull request Mar 14, 2015

[RFC] misc2.c: Move emsg* functions to message.c #2152

Closed

justinmk mentioned this pull request Mar 15, 2015

[RFC] Migrate legacy test 19 and 89 #2074

Closed

tarruda mentioned this pull request Mar 22, 2015

Functional test screen_basic_spec failing #2130

Closed

justinmk added 2 commits April 24, 2015 01:58

WIP: test cmdheight=4

a549ec7

remove cursor_off

d543cdd

see c546875

justinmk force-pushed the echom_bang branch from 363d090 to f5eabc4 Compare April 24, 2015 06:26

justinmk mentioned this pull request Apr 30, 2015

vim_err_write might ask "Press ENTER" in the middle of an (async) message #2547

Closed

bfredl mentioned this pull request Jul 26, 2015

[RFC] don't wait_return in the middle of a vim_err_write message #3106

Merged

justinmk mentioned this pull request Sep 2, 2015

[RFC] main: Add a warning if runtimepath doesn't point to a Neovim runtime #3159

Closed

fwalch mentioned this pull request Sep 3, 2015

Add message if runtime not found. #3283

Closed

justinmk added this to the 0.2 milestone Nov 12, 2015

justinmk mentioned this pull request Aug 8, 2016

:messages pane #5189

Closed

7 tasks

rew: fix tests

ee34ba7

justinmk force-pushed the echom_bang branch from f5eabc4 to ee34ba7 Compare October 19, 2016 10:35

justinmk mentioned this pull request Oct 20, 2016

Add ability to hide warnings #5103

Closed

justinmk modified the milestones: 0.3, 0.2 Jan 9, 2017

This was referenced Sep 19, 2017

Unmodified buffers with set hidden causes multiple error messages on quit #7291

Open

Switch buffers before emitting modified buffer message #7290

Open

neovim deleted a comment from wsdjeg Sep 19, 2017

justinmk mentioned this pull request Sep 28, 2017

win: python: "OSError: [WinError 6] The handle is invalid" #7326

Closed

bfredl modified the milestones: 0.6, backlog Oct 30, 2021

dundargoc removed the WIP label Feb 7, 2023

dundargoc changed the title ~~[WIP] implement :echomsg!~~ implement :echomsg! Apr 2, 2023

neovim deleted a comment from tarruda Jan 5, 2025

neovim deleted a comment from Shougo Jan 5, 2025

justinmk closed this May 3, 2025

justinmk deleted the echom_bang branch May 3, 2025 19:18

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

implement :echomsg! #1802

implement :echomsg! #1802

Uh oh!

justinmk commented Jan 12, 2015 •

edited

Loading

Uh oh!

tarruda commented Jan 12, 2015

Uh oh!

aktau Jan 13, 2015

Uh oh!

justinmk Jan 13, 2015

Uh oh!

ghost commented Jan 29, 2015

Uh oh!

justinmk Feb 16, 2015

Uh oh!

justinmk commented Feb 18, 2015

Uh oh!

tarruda commented Feb 18, 2015

Uh oh!

justinmk commented Feb 18, 2015

Uh oh!

fmoralesc commented May 30, 2015

Uh oh!

justinmk commented May 30, 2015

Uh oh!

fmoralesc commented May 30, 2015

Uh oh!

justinmk commented May 3, 2025

Uh oh!

Uh oh!

Uh oh!

implement :echomsg! #1802

implement :echomsg! #1802

Uh oh!

Conversation

justinmk commented Jan 12, 2015 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tarruda commented Jan 12, 2015

Uh oh!

aktau Jan 13, 2015

Choose a reason for hiding this comment

Uh oh!

justinmk Jan 13, 2015

Choose a reason for hiding this comment

Uh oh!

ghost commented Jan 29, 2015

Uh oh!

justinmk Feb 16, 2015

Choose a reason for hiding this comment

Uh oh!

justinmk commented Feb 18, 2015

Uh oh!

tarruda commented Feb 18, 2015

Uh oh!

justinmk commented Feb 18, 2015

Uh oh!

fmoralesc commented May 30, 2015

Uh oh!

justinmk commented May 30, 2015

Uh oh!

fmoralesc commented May 30, 2015

Uh oh!

justinmk commented May 3, 2025

Uh oh!

Uh oh!

justinmk commented Jan 12, 2015 •

edited

Loading