r31 changelog

[gokcehan]

Changelog

• (Breaking) There has been some changes in the server protocol. Make sure to kill the old server process when you update to avoid errors (i.e. lf -remote 'quit!').
• (Breaking) A new server command query is added to expose internal state to users (Add query server command to expose internal state #1384) (by @joelim-work). A new builtin command cmds is added to display the commands. The old builtin command jumps is now removed. The builtin commands maps and cmaps now use the new server command.
• (Breaking) Environment variable exporting for files and options are not performed anymore while previewing and cleaning to avoid various issues with race conditions (Remove environment exports from preview thread #1354) (by @joelim-work). Cleaning program should now instead receive an additional sixth argument for the next file path to be previewed to allow comparisons with the previous file path. User options (i.e. user_{option}) are now exported whenever they are changed (Export user-options on change #1418) (by @raslop).
• (Breaking) Command outputs are now exclusively attached to stderr to allow printing the last directory or selection to stdout (Allow lastdir/selection to be written to stdout #1399 and Add -print-last-dir and -print-selection flags #1402) (by @joelim-work). Two new command line options -print-last-dir and -print-selection are added to print the last directory and selection to stdout. The example script etc/lfcd.sh is updated to use -print-last-dir instead. Other lfcd scripts are also likely to be updated in the future to use the new method (patches are welcome).
• (Breaking) The option ruler is now deprecated in favor of its replacement rulerfmt (Add rulerfmt option for improved ruler customization #1386) (by @joelim-work). The new rulerfmt option is more capable (i.e. displays option values, supports colors and attributes, and supports optional fields) and more consistent with the rest of our options. See the documentation for more information.
• (New) Modifier keys (i.e. control, shift, alt) with special keys (e.g. arrows, enter) are now supported for most combinations (Support modifiers in special key mappings #1248) (by @joelim-work).
• (New) A new option borderfmt is added to configure colors for pane borders (Support custom colors for drawbox characters #1251) (by @joelim-work).
• (New) New lf specific environment variables, LF_CONFIG_HOME on Windows and LF_CONFIG/DATA_HOME on Unix, are now supported to set the configuration directory (LF_CONFIG_HOME environment variable  #1253) (by @alok-mishra).
• (New) Tilde (i.e. ~) expansion is performed during completion to be able to use expanded tilde paths as command arguments (Expand tilde during autocompletion #1246) (by @joelim-work).
• (New) A new option preserve is added to preserve attributes (i.e. mode and timestamps) while copying (preserve mtime when copying files #1026) (by @raslop).
• (New) The file etc/icons.example is updated for nerd-fonts v3.0.0 (Update icons.example for nerd-font v3 release #1271) (by @MunifTanjim).
• (New) A new builtin command clearmaps is added to clear all default keybindings except for read (i.e. :) and cmap keybindings to be able to :quit (Add clearmaps command #1286) (by @joelim-work).
• (New) A new option statfmt is added to configure the status line at the bottom (Add statfmt option #1288) (by @joelim-work).
• (New) A new option truncatepct is added to determine the location of truncation from the beginning in terms of percentage (Flexible filename truncation #1029) (by @raslop).
• (New) A new option dupfilefmt is added to configure the names of duplicate files while copying (Make duplicated file format configurable #1315) (by @Limero).
• (New) Shell scripts etc/lf.nu and etc/lfcd.nu are added to the repository to allow completion and directory change with Nushell (Added autocompletion and a simple lfcd script for nushell. #1341) (by @mauerlaeufer).
• (New) Sixels are now supported for previewing (sixel preview #1211) (by @horriblename). A new option sixel is added to enable this behavior.
• (New) A new configuration keyword setlocal is added to configure directory specific options (add 'setlocal' for directory specific options #1381) (by @gokcehan).
• (New) A new command line command cmd-delete-word-back (default a-backspace and a-backspace2) is added to use word boundaries when deleting a word backwards (Add cmd-delete-word-back command #1409) (by @janpeterd).
• (Fix) Cursor positions in the directory should now be preserved after file operations that changes the directory (e.g. create or delete) (Prevent cursor jumping to bottom after deleting file #1247) (by @joelim-work).
• (Fix) Option reverse should now respect to sort stability requirements (Fix stable sort when reverse option is enabled #1261) (by @joelim-work).
• (Fix) Backspace should not exit filter mode anymore (Disallow exiting from 'filter' mode with backspace #1269) (by @jdujava).
• (Fix) Truncated double width characters should not cause misalignment for the file information (Refactor filename truncation code #1272) (by @joelim-work).
• (Fix) Piping shell commands should not refresh the preview anymore (Avoid refreshing previews for shell pipe commands #1281) (by @joelim-work).
• (Fix) Cursor position should now update properly after a terminal resize (Bound cursor position after terminal resize #1290) (by @joelim-work).
• (Fix) Directories should now be reloaded properly after a delete operation (Fix load issue when deleting files #1292) (by @joelim-work).
• (Fix) Executable file completion should not add entries to the log file anymore (Remove test log statement during autocompletion #1307) (by @joelim-work).
• (Fix) Blank input lines are now allowed in piping shell commands (Allow blank input lines for shell-pipe commands #1308) (by @joelim-work).
• (Fix) Shell commands arguments on Windows should now be quoted properly to fix various issues (Improve windows shell handling #1309) (by @joelim-work).
• (Fix) Reloading in a symlink directory should not follow the symlink anymore (Prevent following symbolic links when reloading #1327) (by @joelim-work).
• (Fix) Command load should not flicker image previews anymore (Prevent image preview flicker for load command #1335) (by @joelim-work).
• (Fix) Asynchronous shell commands should now trigger load automatically when they are finished (Call load after shell-async commands #1345) (by @joelim-work).
• (Fix) Changing the value of preview option should now clear image previews (Clean image preview on set preview! #1350) (by @jdujava).
• (Fix) Cursor position in the status line at the bottom should now consider double width characters properly (Fix cursor location when printing statline #1348) (by @joelim-work).
• (Fix) Filenames should only be quoted for cmd on Windows to avoid quoting issues for powershell (Wrap filenames in quotes only for Windows CMD #1371) (by @joelim-work).
• (Fix) Inaccessible files should now be included in the directory list and display their lstat errors in the status line at the bottom (List files even when lstat fails #1382) (by @jantatje).
• (Fix) Command line command cmd-delete-word should now add the deleted text to the yank buffer (Add cmd-delete-word-back command #1409) (by @janpeterd).

[gokcehan]

I think we should make a new release soon. I'm thinking of making a new release once the current active PRs are merged. So consider it a feature freeze for r31. @joelim-work @ilyagr Let me know if you need more time for anything.

I have put together a preview for the changelog (@ mentions and # links are excluded to avoid notifications). Let me know if you have any suggestions (e.g. better wording, missing item, misslink/missattribution, label change for breaking/new/fix).

From now on, I'm thinking of maintaining the changelog previews public in a pinned issue so I can update it occasionally from time to time. Otherwise, I'm having trouble motivating myself to make a new release when there are too many changes piled up. This should also be useful for users to see what's currently cooking for the new release. Let me know if you have other suggestions.

[joelim-work]

I think having a freeze period for new releases makes a lot of sense, especially since this project follows a simple master-only branching strategy. This changelog preview is also very helpful in terms of transparency to both collaborators and advanced users who closely follow the development of the project.

This is just a suggestion, but I wonder whether it is worth maintaining a CHANGELOG.md in the repo. By treating the changelog as code itself, you receive the usual benefits from source control management (e.g. easier collaboration, pull request submissions, historical changes, etc.), and you can just copy/paste the contents when creating a new release.

---

Regarding the state of pull requests:

• I have now merged rulerfmt (Add rulerfmt option for improved ruler customization #1386), there was no activity or any other discussions in that PR.
• I have not yet merged the query server command changes (Add query server command to expose internal state #1384), I am not sure if @ilyagr has finished reviewing the changes. Also I have decided to retain cmaps as a default command, although the changelog should mention that jumps is no longer provided as it was actually present in r30.
• Adding a cmd-delete-word-back command (Add cmd-delete-word-back command #1409) seems like a reasonably small change. I don't think it is ready for merging yet, but if it can be fixed soon I think it is unlikely to break anything and can be added for this release.

[gokcehan]

@joelim-work I thought about CHANGELOG.md but I decided we might still need some place to discuss the current state of things, though we can also simply separate the discussion from the changelog. I also wasn't sure about keeping the changelog in the repo since it has github specific information (i.e. @ mentions and # links) and it might increase the commit frequency, though these are not really a big deal. In general I guess it is a good idea to move it to the repository, especially considering the collaboration aspect of it as I'm not the best person when it comes to explaining things. I will try to convince myself that it is a good idea in the upcoming days. I'm also a little hesitant to change our labeling convention to adopt the one in keepachangelog.

[joelim-work]

@gokcehan I forgot to reply to this earlier but the use of CHANGELOG.md is just a suggestion, you don't have to use it. For instance I have found that (as a collaborator?) I am able to edit your comments so maybe just creating an issue like this for each release is sufficient.

That being said, if you do want to add in a CHANGELOG.md, I don't think there's any need to keep it up to date all the time, you can update it once in a while and/or just before a new release. There's also not really any need to follow the exact format specified in https://keepachangelog.com/en/1.0.0/

[ilyagr]

I haven't had any time for PR review this weekend. My brother got married. 🎉

I'm mildly worried about the complexity of server-client interaction that was added in #1384, but I am mostly worried about documenting this in places where someone might try, say, to send a message to the server over a channel that the server is no longer listening to. I haven't followed the entire discussion that lead to the design of #1384, but I don't expect to come up with something better in terms of the overall design. There's always the option of merging that PR and making any changes, if necessary, in follow-up PRs.

[joelim-work]

I haven't had any time for PR review this weekend. My brother got married. 🎉

No worries at all, my congratulations! 🎉

I do acknowledge that there is some more complexity in the interaction between the server and the client. Up until now, the communication was effectively one way - the server forwards send commands to the client(s) to execute. With the addition of query, the client can now send data back to the server, which should work if only one server thread is reading from the socket (the thread that is handling the current query request). This is why the thread handling the original conn request has to return - it was not doing anything useful anyway.

To be clear, the original design was actually proposed by @gokcehan in #1310 (comment), and I just merely got it to work, and then further built upon that idea.

In terms of what to do, the discussion starting from #1277 has already lasted a few months, I am not in a rush to merge this and I don't mind giving you more time if you want. This PR can even skip r31 if you feel it to be necessary. Otherwise I also don't mind merging this and dealing with any issues afterwards.

P.S.: I'm not sure if the release notes need to clarify this, but to use the new query command, the server has to be restarted after upgrading.

[gokcehan]

@ilyagr @joelim-work I looked at #1384 again now and it looks good to me. Since the new code path with query is partly separate, the changes are even more limited. So I don't really expect any issues with the previous server commands. If anything, maybe there could be some issues with the new query command, though that should be acceptable. To be honest, I don't remember many contributions to the server-client before, so it probably consists of my quick and dirty solutions and I wouldn't really trust them much at this point. If there is still a possibility to replace #1384 with one the alternative approaches proposed in #1277, then we should skip it for r31. Otherwise, I don't see much point in waiting further, so I'm in favor of including it in r31. Let me know what you decide.

@ilyagr Congratulations to your brother. Speaking of irl, I'm in the process of changing my life setup these days, and I will probably disappear for a while towards the end of this month, hopefully not permanently. This is why I want to push a new release soon, preferably this weekend. Looking at the changelog, it is appearent that @joelim-work is handling most of the development these days, and I only do occasional complaints about breaking changes, so I don't expect any issues when I'm not around.

@joelim-work I forgot to think about the changelog, but since you can also edit the comment maybe we should just keep it as an issue for now. I have now added a breaking change about the server restart.

[joelim-work]

@gokcehan I've decided to merge #1384, I think we might as well not wait any longer. Regarding all the other alternative approaches discussed over the last few months, I consider them to be inferior to the query approach, and I just merely kept them for reference.

I have merged the Upcoming Changes section into the Unreleased Changes section. I think I would prefer if you wait for another week before releasing r31 just in case, but I don't think there will be any issues. I should be around during this time to respond to issues if anything comes up.

[ilyagr]

I'm entirely fine with merging #1384. I still hope to look it over, but I never meant for that to slow everything down so much; I can always make follow-up PRs if I will feel like it. Sorry, I continue to have very limited free time and, more importantly, brainpower for lf. I hope that changes in the near future, but I can't make any promises.

@gokcehan, feel free to release whenever works for you. I'd also prefer that to be later (e.g. in a week as Joe suggested) rather than sooner, but it's more important that it works with your schedule.

[gokcehan]

@joelim-work @ilyagr Waiting for an extra week sounds good.

[tayopi]

In the mean time, is there a workaround for #1371?

[joelim-work]

In the mean time, is there a workaround for #1371?

You can always build from source, which will give you access to all the latest bug fixes and features.