ToDo list

[jarun]

Rolled from #881.

Cooking

• Bksp/Del at empty filter prompt to refresh dir
• fix zombies left behind after running plugins (nnn leaves zombies while running plugins #999)
• consider nanosecond resolution when sorting by time (Include milliseconds when sorting by time #978)
• check external selection in 'c'urrent / 's'el prompt (Populating selection with x2sel plugin #976)
• support picker plugins e.g. fzopen (Weird behavior with fzopen plugin mcchrish/nnn.vim#82)
• make batch rename interactive (Swapping names when batch renaming files deletes one of the files #971)
• cached uid/gid for performance improvement
• support QuickLook on WSL in preview-tui* (Get the selected file for live preview to work with a MS Window previewer #959)
• toggle preview-tui* with the same custom plugin key
• smoother preview toggling in preview-tui* (preview-tui: reformat toggle #966)
• config listen_on should be set in kitty.conf for preview-tui* (preview-tui: error to foreground #970)
• minimal bat style in preview-tui*, honors BAT_STYLE
• plugin unmount-parent: unmount a remote mountpoint from within
• plugin xdgdefault: set the default application for hovered file type
• plugin fzplug: fuzzy find, preview and run other plugins
• plugin cmusq: queue/play music in cmus
• plugin mocplay renamed to mocq
• go to last dir on ~ (HOME) or ` (ROOT) key repeat
• key ^Space for range selection/clear dropped, use m (nnn + netbsd-curses + musl libc #998)
• a Themes wiki page for users to add and refer to
• show selection mark (+) in reverse bold for improved visibility
• reversed block instead of > to mark hovered entry in detail mode
• make option O_CKBOARD retired in favour of reversed block cursor
• retain filter in nav-to-type mode after file open
• script to statically compile with netbsd-curses + musl on Ubuntu
• fixes for netbsd-curses + musl (nnn with netbsd-curses sabotage-linux/netbsd-curses#46)
• fix broken picker mode with NNN_TMPFILE exported and q to quit
• disable xterm title setting in picker mode (picker mode broken #974)
• make option NOX11: disable notis, sel->clipboard sync, xterm title

Up for grabs

• multi-threaded disk usage calculation
• notcurses UI backend (notcurses ui backend #1014) [@KlzXS]
• a light colorscheme with NNN_FCOLORS (Default light colorscheme #937)
• update only current context data when saving session (Storable contexts as the missing link halfway between bookmarks and sessions? #837)
• generate static binary for arm, aarch with releases using CI (arm and aarch releases with statically linked dependencies #623)
• support predefined filters like bookmarks
• open new terminal window in spawned shell directory (Spawned shell (!) with kitty (--cwd=current) #759)

For anything else please discuss in this thread.

Contribution guideline.

[luukvbaal]

Following discussion from #267 I do think showing the the last characters of the file name should be prioritized in the case of file names longer than the name column. I was wondering if we should consider changing the unescape() function to return something like abcmnopqrst.txt or abc...pqrst.txt for abcdefghijklmnopqrst.txt.

[jarun]

Doesn't cover all cases inclusively. For some filenames the difference may be in the middle. There's always f/^F for file details. Just leave it be.

[lawnowner]

@jarun (#499 (comment)) The prompt looks better than before with repl now. This morning, I got a three line error message starting with "Assembler: " after running sleep 1; date and typing stuff while asleep, although all I can do is wishing you good luck with reproducing that as I can't even remember the full error message, let alone the keys pressed. As somebody who has been using nnn in monochrome, and with almost none of the official plugins, I guess the next release will be my favourite, except for the tilde contraction next to the context numbers; it looks almost like nothing whem I'm home.

[jarun]

Get used to the tilde. It was requested by multiple users.

[Kabouik]

I noticed preview-tui and preview-tui-ext queue image previews when scrolling through multiple files. It is probably a non-issue on most machines, but breaks previews on slower ones (i.e., SBCs running on SD cards, machines/VM/LXC without hardware acceleration for kitty image preview, etc.) because load times add up for every file hovered along the way.

It may be true for other file types, but is most visible for big images since they take longer to load (on my phone running Debian and previewing with kitty's icat, the issue is not huge with images around 500 KB, but very noticeable with images around 2-4 MB). I suppose even on real computers this could be an issue with large images from modern DSLRs or scanned files.

I tried:

        kitty +kitten icat --silent --place "$1"x"$2"@0x0 --transfer-mode=stream --stdin=no \
-           "$3"
        kitty +kitten icat --silent --place "$1"x"$2"@0x0 --transfer-mode=stream --stdin=no \
+           "$3" &

Which seemed to limit queued previews, but resulted in previews being shown only if scrolling slowly.

Should I open a dedicated issue about it?

---

While I mention preview plugins, how about making default bat previews more minimal with:

-           fifo_pager bat --terminal-width="$cols" --paging=never --decorations=always --color=always \
+           fifo_pager bat --terminal-width="$cols" --paging=never --style=numbers,header,plain --color=always \

It is a very minor difference and perhaps just personal taste, but I see no real advantage for the grid view by default:

[jarun]

@leovilok @luukvbaal can you please take a look?

It s a very minor difference and perhaps just personal taste, but I see no real advantage for the grid view by default

Yes, it actually looks awful to anyone from sane more/less background. Please raise a PR to remove this from all plugins.

[luukvbaal]

The preview plugins work with a FIFO so the previews are inherently queued. Not sure anything can be done about a slow machine that can't keep up with how fast you change the selection.

Never noticed the bat defaults since I have BAT_STYLE exported in my profile. I agree the grid is only clutter so I'm fine with getting rid of it.

[Kabouik]

I really have no clue how a FIFO works, but was hoping there was a way to interrupt prior previews if a new one is being issued, without waiting. Something I vaguely remember @leovilok solved with a simple & when I was having a similar issue with viu (hence why I tried & here, but I guess it's only a partial solution and reaches its limits when preview load time is longer).

[Edit] Similar issue with PDF and audio previews using poppler-utils andd ffmpeg in preview-tui-ext. Also I confirmed rapid image scrolling (using key repeat) causes it too on a fast computer, compared to the rather slow phone running Debian above.

[luukvbaal]

I was actually able to reproduce the queuing to some extent with kitty on my laptop with key-repeat and #1006 should fix the issue.

If at all possible I would suggest using ueberzug instead of kitty's icat if speed is an issue for you since it is ALOT faster.

[jarun]

If at all possible I would suggest using ueberzug instead of kitty's icat if speed is an issue for you since it is ALOT faster.

Completely agree! As someone who multitasks all the time my experience with kitty has been horrible. And I have started compiling with O_FIFO=1 only after @luukvbaal added ueberzug support.

[KlzXS]

TIL github will match "fix #XXX" in a middle of a sentence.

[jarun]

Nice image BTW @KlzXS!

[jarun]

@luukvbaal and @Kabouik can you guys audit nuke + preview-tui* (all plugins in general) and see if those can be optimized?

Things I can think of:

• case in place of too many ifs
• use of shell builtins (e.g. type instead of which)
• it would be great to get rid of echo for printf

This optimization document is a good start if you need any reference.

[luukvbaal]

Yeah some of those things have been bothering me since I started working on preview-tui but I didn't want to change the style you guys adopted. I'll see what can be done for preview-tui however I don't use nuke myself so I'd rather if someone else could pick that up.

[jarun]

however I don't use nuke myself

Doesn't matter. Please have a look if you have the time.

[wustho]

@jarun hello there, sorry the thread is locked. About cleanfilename, with latest commit (merged) it won't overwrite any file...

Instead, it will be prepended with _ if file with the same name already exists... Eg. file1 --> _file1 if file1 already exists...

Already stated in doc:

#              And if there are two almost similar filenames
#              like: 'asd]f' and 'asd f' which both will be renamed to 'asd_f',
#              to avoid overwriting, the last file will be prepended by _.
#              So they will be: 'asd_f' and '_asd_f'
#

[jarun]

@wustho awesome! Thanks for the confirmation.

[Kabouik]

If at all possible I would suggest using ueberzug instead of kitty's icat if speed is an issue for you since it is ALOT faster.

Completely agree! As someone who multitasks all the time my experience with kitty has been horrible. And I have started compiling with O_FIFO=1 only after @luukvbaal added ueberzug support.

Ha, thanks for the advice! That's frustrating after I spent time configuring kitty over the last two days exactly because I wanted to use icat and finally have real image previews in terminal. I assumed icat being native to kitty, it would perform better than alternatives. I'll try the combination kitty + ueberzug on the slow machine instead.

@jarun what exactly was horrible for you with kitty? Overall performance regardless of icat?

[jarun]

what exactly was horrible for you with kitty?

Image rendering performance.

[Kabouik]

I failed at prioritizing ueberzug over kitty's icat in the plugin (I shifted its clause to the top but broke previews, won't bother you with that here), but to get a better idea of ueberzug's performance, I compared image preview between a fff fork using ueberzug and fff upstream using w3mimgdisplay.

I got much better rendering speed with w3mimgdisplay. Through my reading I learnt its advantages over ueberzug are it works on Wayland too, can print images on the tty, and doesn't need a daemon like ueberzug (apparently).

I am sure it has drawbacks too, ueberzug is vey popular. But would it make sense to add support for w3mimgdisplay in preview plugins? nnn has a broad audience that certainly includes slow systems (like SBCs), machines using Wayland, or others with just a tty, where both ueberzug and kitty would be unusable. Since w3mimgdisplay appears less restrictive and faster (at least I can distinguish it from ueberzug on a slow system, maybe not on a fast one), I guess it's not too redundant. viu could work on all those machines but there's a gap in functionality.

[Kabouik]

w3mimgdisplay doesn't support tmux so problem solved!

[luukvbaal]

You're right, prioritzing ueberzug over icat in kitty requires more changes than shifting the order in image_preview(). See e.g. lines such as: [ "$TERMINAL" != "kitty" ] && exists ueberzug && ueberzug_remove in preview_fifo().

ueberzug previews are pretty much instantaneous on the machines I use nnn on but I can believe w3mimgdisplay could be faster as its written in C as opposed to python.

The problem I have with w3mimgdisplay is that I've never gotten it to work in any of the terminals I've used. This is also one of the main reasons ueberzug exists in the first place if you go by the README(it's definitely not less restirictive than w3mimgdisplay atleast for X11 environments).

Having said that, if you can manage to add support for w3mimgdisplay to preview-tui I say go ahead. Again, I've had nothing with problems with w3mimgdisplay and haven't touched it since ueberzug was created so I see no reason add it myself.

w3mimgdisplay doesn't support tmux so problem solved!

Yeah, considering that, w3mimgdisplay doesn't really fit in preview-tui I guess.

[luukvbaal]

@Kabouik This is the diff you would need for ueberzug previews in kitty.

[Kabouik]

Nice, thanks a lot @luukvbaal! It's not instantaneous on my slow device but I can feel an improvement indeed.

The queuing is there though (which is expected, it was prevented just for icat as far as I understand). However I have mixed feelings about the queuing in this case. On the one hand it's ugly and can go wrong if scrolling for several seconds continuously [Edit: really wrong on long scrolls]. On the other hand ueberzug appears to create some kind of cache for the images it already previewed, which means this cache is fed by the queued files, so later preview of the scrolled files is almost instantaneous even on this machine.

[luukvbaal]

I do believe ueberzug does some kind of caching yeah but I would also suggest following
# Consider setting NNN_PREVIEWDIR to $XDG_CACHE_HOME/nnn/previews if you want to keep previews on disk between reboots on a slow device if you havent already. If disk space is less of an issue that is.

This will place converted images in $XDG_CACHE_HOME instead of the default $TMPDIR which is lost after a reboot.

[Kabouik]

Will do. Disk space is an issue since this is a phone but I know where to clear the cache when I reach the limit.

Probably out of scope since it's not the official plugin, but it doesn't hurt to report it: with the diff to prioritize ueberzug over icat in kitty, the preview does not properly follow layout changes. See here with patched preview-tui-ext (ueberzug previews) on the left, and vanilla preview-tui (icat previews) on the right:

Peek.2021-05-15.00-23.mp4

(Similar issue when using tmux panes: the preview follows layout changes but is cropped in certain positions.)

[luukvbaal]

Yeah preview-tui is not aware of the preview pane origin as is the case for tmux. I already accounted for this for horizontal splits in the initial diff but forgot about vertical splits.

Try the following ueberzug_layer()

ueberzug_layer() {
    if [ "$TERMINAL" = "kitty" ]; then
        if [ "$SPLIT" = "v" ]; then
            x=$(($(tput cols) / 2 + 1))
            y=0
        else
            x=0
            y=$(($(tput lines) / 2 + 1))
        fi
    else
        x=0
        y=0
    fi
    printf '{"action": "add", "identifier": "nnn_ueberzug", "x": %d, "y": %d, "width": "%d", "height": "%d", "scaler": "fit_contain", "path": "%s"}\n' "$x" "$y" "$1" "$2" "$3" > "$FIFO_UEBERZUG"
}

Updated the gist as well.

[Kabouik]

This didn't solve the issue from my first tests, but that was just a quick test while on the move, I'll test more thoroughly with more terminals and scenarios.

Unrelated: It was mentioned in #977 that scripts and vanilla functions outputs are garbled after several actions because they are displayed on top of each other. In cmusq I just added \n at the end of all prompts and this turned out to work decently for the use case. Would that be usable for nnn itself and possibly other scripts?

Example of garbled output with cpg here:

(By the way, noob question, how do we copy/paste from nnn and accept overwrite for the whole selection?)

I tried adding \n after overwrite? at L584 of nnn.c but it didn't work, I suppose the issue might be with cpg instead.

It may seem only an aesthetic issue but sometimes, depending on the length of sentences, it can be unreadable and it may be impossible to tell if the output expects an answer, what are the accepted answers, unless experienced with daily nnn use already.

[jarun]

Would that be usable for nnn itself and possibly other scripts?

You can do that for all the scripts. Just ensure you see the problem in the script and then you do the change.

how do we copy/paste from nnn and accept overwrite for the whole selection?

This is not an nnn issue. nnn uses the interactive option of cp. See cp options. If you have something you'll have to recompile the code with that.

I tried adding \n after overwrite? at L584 of nnn.c but it didn't work

That prompt is from advcpmv. You can try to add an fflush() after every progress bar ending and see if that works. Please raise a PR if it works for you.

[mizlan]

May I request some better documentation in the form of docstrings and comments for variables and functions for the C source code? Since it isn't too clear what variable names mean at a glance, it's quite difficult to gain an understanding of the code from reading control flow alone.

[jarun]

@Kabouik the broken overwrit eprompt is fixed at jarun/advcpmv@6adf544.

@mizlan unfortunately no such documentation exists. Please feel free to ask any questions you have. Also, I invited to the nnn devs group. Please join that. We generally discuss there.

[mizlan]

So sorry, but I am unaware of the dev group and can't seem to find it. What platform is it on? Gitter?

[Kabouik]

Awesome, thanks @jarun. I didn't even have the time to try what you suggested, can't keep up with your pace. :>

…

On 2021-05-18 18:33 Piña Colada ***@***.***> wrote: @Kabouik the broken overwrit eprompt is fixed at jarun/advcpmv@6adf544 0545d23b94. @mizlan unfortunately no such documentation exists. Please feel free to ask any questions you have. Also, I invited to the nnn devs group. Please join that. We generally discuss there. -- You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub: #935 (comment)

[jarun]

but I am unaware of the dev group and can't seem to find it.

Please check your mail. I sent you an invitation 1/2 days back.

[mizlan]

The invitation for was collaboration (push access) right? I do not see any sort of discussion group associated with it. Sorry for the inconvenience.

[jarun]

Not that one. I've resent the invitation just now. Please check.