Add support for bracketed paste #3871

faho · 2017-02-25T14:01:19Z

This is a terminal feature where pastes will be "bracketed" in
\e[200~ and \e[201~.

It is more of a "security" measure (since particularly copying from a
browser can copy text different from what the user sees, which might
be malicious) than a performance optimization.

Fixes #967.

I'd love to see some testing on this, particularly on weird terminals (especially those on OSs I don't have access to) and older versions of emacs' "ansi-term" (I think macOS ships an ancient pre-GPL3 version?).

TODOs:

[N/A] Changes to fish usage are reflected in user documentation/manpages.
Ideally, support is ubiquitous and documentation is unnecessary. We also want to downplay this as \cv is better.
[N/A] Tests have been added for regressions fixed
No regressions fixed, tests have been neutered as I don't know how to test since it depends on the terminal.
User-visible changes noted in CHANGELOG.md

This is a terminal feature where pastes will be "bracketed" in \e\[200~ and \e\[201~. It is more of a "security" measure (since particularly copying from a browser can copy text different from what the user sees, which might be malicious) than a performance optimization. Fixes fish-shell#967.

krader1961

This works for me using iTerm2 and Terminal.app on macOS Sierra.

krader1961 · 2017-02-26T03:42:25Z

share/functions/__fish_config_interactive.fish

    # Load key bindings
    __fish_reload_key_bindings

+    if not set -q FISH_UNIT_TESTS_RUNNING


LOL! It's nice to see another use for that var so soon after I added it.

krader1961 · 2017-02-26T03:50:45Z

share/functions/__fish_config_interactive.fish

+    if not set -q FISH_UNIT_TESTS_RUNNING
+        # Enable bracketed paste before every prompt (see __fish_shared_bindings for the bindings).
+        # Disable it for unit tests so we don't have to add the sequences to bind.expect
+        function __fish_enable_bracketed_paste --on-event fish_prompt


This is a very fishy way to do this. Very cool. I was thinking we need to also emit the magic sequence to disable bracketed paste mode when exiting the shell, such as in response to a signal, but honestly that is such an unlikely scenario I think we should defer doing so until someone complains.

I'm complaining 😉 It's pretty weird to me that if I start a bash shell, type fish, and exit the Fish shell, pasting is now broken.

Fixed in 4ff002b.

Taking for 2.6.0.

krader1961 · 2017-02-26T03:51:38Z

share/functions/__fish_config_interactive.fish

+            printf "\e[?2004l"
+        end
+
+        __fish_enable_bracketed_paste


I'm surprised this is needed. Which means it should probably have a comment explaining why it is needed.

Of course it's needed - we need to tell the terminal we support bracketed paste, and since this is in __f_c_i, the first fish_prompt has already fired. Will add a comment to that effect.

It's the fact that this isn't loaded until the first fish_prompt event has fired that always trips me up. It's counterintuitive. I, and probably most people, expect autoloading this script to happen first. So I'm always surprised when I see this pattern. Hence the need for a comment.

krader1961 · 2017-02-26T03:57:48Z

share/functions/__fish_shared_key_bindings.fish

+    # Re-running `bind` multiple times per mode is still faster than trying to make the list unique,
+    # even without calling `sort -u` or `uniq`, for the vi-bindings.
+    set -l allmodes default
+    set allmodes $allmodes (bind -a | string match -r -- '-M \w+' | string replace -- '-M ' '')


This is fragile in that it depends on the format of the output of bind -a which isn't guaranteed to not change. For example, we might decide to replace -M with --mode in the output for clarity. I recommend opening a new issue to enhance bind with a way to list the known modes and add a "TODO" comment here referencing that issue.

Yes, that would be a much better way of handling this.

krader1961 · 2017-02-28T03:22:21Z

LGTM. I forgot to mention that I've tested this on macOS using Terminal.app and iTerm2. If no one reports any problems in another four days (one week after sending this out for review) feel free to merge it. We can always revert or amend it if someone reports a problem.

zanchey · 2017-02-28T03:24:16Z

Do you think this needs to go into the documentation at all?

krader1961 · 2017-02-28T03:29:35Z

I don't think we need to document this outside of the change log. The new behavior is unlikely to surprise anyone. Having said that it might be worth noting we support it in the FAQ so people searching the docs for "bracketed paste" can find a mention that we do support it and know where to look for implementation details since the latter is pretty cool.

ridiculousfish · 2017-02-28T03:53:56Z

This looks sweet. Thanks for doing this.

floam · 2017-03-01T18:50:12Z

Sorry, I can't test this at the moment. Does this escape the pasted output? i.e. allow someone to type wget, paste a URL, and not worry about the ?s?

faho · 2017-03-01T19:22:10Z

Does this escape the pasted output?

No. What it does is allow pasting input that contains characters bound to something, e.g. \n and such, without triggering those bindings (which is similar to escaping). It could feasibly be adjusted to do simple escaping (e.g. insert \? instead of ?). The main issue with that is that it's not entirely clear when you would want something escaped and when you wouldn't. E.g. when you paste echo *, you want to have the glob there, but when you paste some markdown with **emphasis** you don't.

Anyway, what we took from #967 was "support bracketed-paste", which isn't quite what the discussion focussed on. So we can debate whether this "fixes" it, or whether that issue should be will-not-implement or implemented.

faho · 2017-03-05T23:23:04Z

Merged as db63be7.

elliott-beach · 2017-08-08T00:20:24Z

share/functions/__fish_shared_key_bindings.fish

+    # This sequence ends paste-mode and returns to the previous mode we have saved before.
+    bind -M paste \e\[201~ 'set fish_bind_mode $__fish_last_bind_mode; commandline -f force-repaint'
+    # In paste-mode, everything self-inserts except for the sequence to get out of it
+    bind -M paste "" self-insert


Is this paste mode documented anywhere? I was looking to make a plugin similar to https://www.npmjs.com/package/hyperterm-paste with fish, and need this feature to do it.

@e-beach: That right there is basically the extent of our documentation on it.

For those transformations, implementing it here seems difficult. You'll want to use something like fish_clipboard_paste instead. Look at that function and modify it to your liking, then use ctrl-v instead of ctrl-shift-v to paste.

Right on. Thanks!

jamesgecko · 2017-08-08T17:21:55Z

Is there a way to disable bracketed paste mode? When using fish in tmux I've started getting [200~clipboard content[201~ whenever I try to paste.

faho · 2017-08-08T18:02:25Z

Currently, there is not. Which tmux version are you using in which terminal? I checked it just now in tmux 2.5 (and I had tested it before merging) and I just do not see that.

jamesgecko · 2017-08-08T18:34:20Z

I'm using tmux 2.5 with fish 2.6.0 in iTerm 3.0.15. The problem popped up when I ran brew upgrade for the first time in several months; I'm pretty sure that I upgraded both fish and tmux at the same time.

krader1961 · 2017-08-08T20:06:46Z

@jamesgecko, I'm using a similar setup and cannot reproduce your problem. Bracketed paste works for me when running fish inside tmux inside iTerm. I'm guessing you've set the tmux escape-time var to zero or close to it. In general setting that to less than 10 ms is going to cause problems.

jamesgecko · 2017-08-08T21:07:22Z

I did have escape-time set to zero. I also had keybindings for using alt+[ and alt+] to tab between tmux windows. 🤦‍♂️

bind-key -n M-[ previous-window
bind-key -n M-] next-window

Thanks!

krader1961 · 2017-09-06T03:27:19Z

@jamesgecko, fish doesn't have a bind-key command so presumably you've got a wrapper that translates what you wrote to bind \e\[ previous-window. And therein lies the problem. Bracketed paste works by having the terminal send \e\[200~ and \e\[201~. Fish implements bindings for those sequences. See type __fish_shared_key_bindings. Your binding, because it's a prefix, takes precedence which interferes with fish's handling of bracketed paste. Don't do that 😄

faho added the enhancement label Feb 25, 2017

faho added this to the 2.6.0 milestone Feb 25, 2017

krader1961 reviewed Feb 26, 2017

View reviewed changes

Add more comments

a362ddf

faho closed this Mar 5, 2017

faho mentioned this pull request Mar 17, 2017

Add "--list-modes" option to bind #3902

Closed

3 tasks

geoffmacintosh mentioned this pull request Mar 23, 2017

youtube-dl command does not work in fish shell #3907

Closed

laughedelic mentioned this pull request Aug 2, 2017

Copy paste laughedelic/pisces#6

Closed

elliott-beach reviewed Aug 8, 2017

View reviewed changes

sarum9in mentioned this pull request Dec 10, 2017

Bracketed paste mode causes fish to hang on WSL if started as PID=2 #4582

Closed

github-actions bot locked as resolved and limited conversation to collaborators Apr 17, 2020

faho deleted the bracketed-paste branch January 17, 2024 09:35

Uh oh!

Add support for bracketed paste #3871

Add support for bracketed paste #3871

Uh oh!

Conversation

faho commented Feb 25, 2017

TODOs:

Uh oh!

krader1961 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lilyball May 24, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

krader1961 commented Feb 28, 2017

Uh oh!

zanchey commented Feb 28, 2017

Uh oh!

krader1961 commented Feb 28, 2017

Uh oh!

ridiculousfish commented Feb 28, 2017

Uh oh!

floam commented Mar 1, 2017

Uh oh!

faho commented Mar 1, 2017

Uh oh!

faho commented Mar 5, 2017

Uh oh!

elliott-beach Aug 8, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jamesgecko commented Aug 8, 2017

Uh oh!

faho commented Aug 8, 2017

Uh oh!

jamesgecko commented Aug 8, 2017

Uh oh!

krader1961 commented Aug 8, 2017

Uh oh!

jamesgecko commented Aug 8, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

krader1961 commented Sep 6, 2017

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

lilyball May 24, 2017 •

edited

Loading

elliott-beach Aug 8, 2017 •

edited

Loading

jamesgecko commented Aug 8, 2017 •

edited

Loading