Show HN: History suggest box for your shell(github.com)
github.com
Show HN: History suggest box for your shell
https://github.com/adder46/hstr-rs
28 comments
Same here. With fzf, you don't search - you find. One of my most powerful bash aliases is
alias fzubl='fzf -m | xargs subl'
alias fzubl='fzf -m | xargs subl'
fzf for the win. It’s such a boost of productivity. Everyone who spends significant amount of times in the shell should try it.
I was about to comment that this kinda looks like fzf + ncurses.
This is pretty cool. I'm always mashing the up arrow, so I installed this in anticipation of making my life easier :)
I didn't have `cargo` or any other rust language support installed. However, that seems pretty straight forward: https://www.rust-lang.org/tools/install
You may want to throw that into the README.
Anyhow, I'm running this on WSL2/Ubuntu 20.04 using zsh + ohmyzsh. It installs cleanly, and I just added `alias hh=hstr-rs` to ~/.zshrc as the only additional step. Works nicely!
I didn't have `cargo` or any other rust language support installed. However, that seems pretty straight forward: https://www.rust-lang.org/tools/install
You may want to throw that into the README.
Anyhow, I'm running this on WSL2/Ubuntu 20.04 using zsh + ohmyzsh. It installs cleanly, and I just added `alias hh=hstr-rs` to ~/.zshrc as the only additional step. Works nicely!
If you want to keep your zshrc "clean" you can add any number of individual scripts to `~/.oh-my-zsh/custom` and they will be executed for every new shell.
Oh that's good to know. For someone my age, I have embarrassingly little knowledge of shell customization. I just looked in .zshrc and there was a comment to the effect of "Add your aliases below, e.g. ALIAS blah=foo" and so I listened to it!
There is nothing inherently wrong with putting them into .zshrc!
The reasons I keep them separate is because:
a) I can simply backup the `custom` directory and restore onto a "vanilla" oh-my-zsh installation b) I can use descriptive file names which would otherwise have to be comments
Also, some software likes to append stuff (like adding directories to PATH) to .zshrc upon installation so it gets kind of messy over time.
I guess the comment is from the plain zsh installation and oh-my-zsh just didn't remove it.
The reasons I keep them separate is because:
a) I can simply backup the `custom` directory and restore onto a "vanilla" oh-my-zsh installation b) I can use descriptive file names which would otherwise have to be comments
Also, some software likes to append stuff (like adding directories to PATH) to .zshrc upon installation so it gets kind of messy over time.
I guess the comment is from the plain zsh installation and oh-my-zsh just didn't remove it.
hstr-rs is a shell history suggest box that provides quick access to the commands in your shell history. Just search for the command you want to execute (regex mode is supported), or, if you don't precisely know what you're looking for, you can browse through the commands until you find the one you want to execute. Apart from this, you can delete unwanted entries from your history, too, and more. It should work on all terminals ranging from linux term to xterm. It was initially made for bash but it should work as well with zsh, ksh, and tcsh, too.
I use it on a daily basis.
Check out the README for instructions if you want to try it out.
Suggestions, bug reports, etc. are welcome.
Let me know what you think:
https://github.com/adder46/hstr-rs
I use it on a daily basis.
Check out the README for instructions if you want to try it out.
Suggestions, bug reports, etc. are welcome.
Let me know what you think:
https://github.com/adder46/hstr-rs
Small mistake in the readme
> hstr is shell history suggest box. Like hstr, but with pages.
Should be "hstr-rs is shell..." I presume?
Looks like a nice project :) I'll give it a go. I've often wanted to iterate through options when ctrl-r ing in bash (not saying there isn't a way to do that, but if there is I don't know it)
> hstr is shell history suggest box. Like hstr, but with pages.
Should be "hstr-rs is shell..." I presume?
Looks like a nice project :) I'll give it a go. I've often wanted to iterate through options when ctrl-r ing in bash (not saying there isn't a way to do that, but if there is I don't know it)
This looks really cool, I’m going to try it out — by the way, what app/tool did you use to record the screencast?
In case you didn't know, bash has a basic version of this already embedded. You can search your bash history from the command line with Control+R (not Cmd+R even for Macs).
Yes. ZSH has this too. I think the basic shell features can go a long way.
Totally, but the dropdown the rest of the height of the terminal is the feature I'm appreciating here.
Good idea and fairly easy to implement from .bash_history. I never thought of it myself so I must be content with reverse search and grepping through bash history.
One suggestion would be to make it a homebrew recipe for macOS and the Linux version. It would then see a lot more adoption (even if the alias has to be set manually which is fine as it's personal). I'd install it instantly (and probably forget about it for a month).
One suggestion would be to make it a homebrew recipe for macOS and the Linux version. It would then see a lot more adoption (even if the alias has to be set manually which is fine as it's personal). I'd install it instantly (and probably forget about it for a month).
How is this different than reverse-i-search?
The full page display of selections seems like the most obvious difference to me.
None.
But seems like the trend of the last 10 years is to rewrite every single already existing Unix tool in Rust or node.js. Throwing out code is more fun than reading man pages I guess.
This saddens me. It looks like the average dev in 2020 believes he needs to pollute his working environment with thousands of seemingly useless tools to be productive.
fzf, silver searcher, ripgrep, etc.
I will be downvoted to hell, because these people genuinely believe they need these tools, and will find whatever performance or feature argument to explain why. But from my experience, I am much faster at any command line task than pretty much anyone using these fancy stuff, even though I stick to bash, vim, and coreutils...
But seems like the trend of the last 10 years is to rewrite every single already existing Unix tool in Rust or node.js. Throwing out code is more fun than reading man pages I guess.
This saddens me. It looks like the average dev in 2020 believes he needs to pollute his working environment with thousands of seemingly useless tools to be productive.
fzf, silver searcher, ripgrep, etc.
I will be downvoted to hell, because these people genuinely believe they need these tools, and will find whatever performance or feature argument to explain why. But from my experience, I am much faster at any command line task than pretty much anyone using these fancy stuff, even though I stick to bash, vim, and coreutils...
The standard tools have a lot of cruft. `grep` being case sensitive by default is I think a good example of a default that used to make sense, but is probably better the other way around today. There are countless other such examples.
Your comment reads as a no-true-scottsman "look how skilled I am" signaling device. You're not necessarily wrong, but you're also clearly fighting a pointless battle, and sound a bit out-of-touch.
Yes, the tools already exist and work. No, they do not need to be re-written, but the new versions are often substantially more usable and newcomers enjoy that. Such is the nature of life. Things change. Accept that and you'll be happier.
Your attitude might win some points with other fight-the-tide-till-my-last-dying-breath types, but it reads to me as "It is the children who are wrong!"
I worked with a guy who had the same attitude about always using kubectl over the gui because it was "faster". He was demonstrably slower at quite a few common tasks, like discovering errors in logs across many pods, than others who used the gui.
Your comment reads as a no-true-scottsman "look how skilled I am" signaling device. You're not necessarily wrong, but you're also clearly fighting a pointless battle, and sound a bit out-of-touch.
Yes, the tools already exist and work. No, they do not need to be re-written, but the new versions are often substantially more usable and newcomers enjoy that. Such is the nature of life. Things change. Accept that and you'll be happier.
Your attitude might win some points with other fight-the-tide-till-my-last-dying-breath types, but it reads to me as "It is the children who are wrong!"
I worked with a guy who had the same attitude about always using kubectl over the gui because it was "faster". He was demonstrably slower at quite a few common tasks, like discovering errors in logs across many pods, than others who used the gui.
Yes it looks much nicer imo. One thing that annoys me with ctrl-r is that if you make a typo and backspace it, you will not be back to where you were. The "cursor" has moved back in history, potentially past the command you were interested in. It's also nice you can see many different matches at the same time.
I kind of think the author should have foreseen this question and provided an answer.
I kind of think the author should have foreseen this question and provided an answer.
You should see my git aliases that makes heavy use of fzf. It definitely speeds me up. `ga` gives an fzf-powered selection only on files in my git status to `git add`. I can cherry pick commits without ever typing a branch or commit hash. Fuzzy-find on reverse search with fzf is way better than the built-in ctrl+r. ripgrep is way better than grep. I could go on.
I appreciate your call to be familiar with the core unix experience, but the unix userland is definitely enriched by these newfangled tools.
I appreciate your call to be familiar with the core unix experience, but the unix userland is definitely enriched by these newfangled tools.
So which man page from "bash, vim, and coreutils" should I look into to get the equivalent of fzf?
I can easily give examples of using `fzf` or `ripgrep` to perform the same action in many less keystrokes than the built-in utilities would require, e.g., traversing a deep directory hierarchy using `fzf` to fuzzy search for the right directory to `cd` to, or performing a text search of only `git` included files over many repos.
Given these tasks are accomplishable in many less keystrokes using `rg` and `fzf` than the built-in utilities, how would you make the argument that you're able to accomplish these command line tasks faster using only "bash, vim, and coreutils"?
Given these tasks are accomplishable in many less keystrokes using `rg` and `fzf` than the built-in utilities, how would you make the argument that you're able to accomplish these command line tasks faster using only "bash, vim, and coreutils"?
You’ve just picked an arbitrary point in the development of these tools that you’re happy with and are suggesting that any progress beyond that point is wasted. I’m sure there were folks around when bash development started saying that they were super productive in Bourne shell so why start a rewrite from scratch. Is vim the terminal point of editor development, forever? Why wasn’t it when nvi replaced Joy’s original vi? Why aren’t we still using ex?
All of the tools you mentioned started as rewrites of existing tools and only became better over time as they added features. That process hasn’t stopped.
All of the tools you mentioned started as rewrites of existing tools and only became better over time as they added features. That process hasn’t stopped.
Well put. I got very good at typing:
And it automatically ignores the directories like .git, build, node_modules, etc., whatever is in .gitignore, without having to add a bunch of "-prune" clauses.
find . -type f -exec egrep -il "whatever" {} \;
rg whatever
And it automatically ignores the directories like .git, build, node_modules, etc., whatever is in .gitignore, without having to add a bunch of "-prune" clauses.
Silver searcher and ripgrep can ignore git-ignored files very easily. Don’t know how to do that with grep.
Fzf gives you a list of hits that you can go through to select the one you want. Ctrl-r in bash just shows you the current hit.
Those are real advantages.
Fzf gives you a list of hits that you can go through to select the one you want. Ctrl-r in bash just shows you the current hit.
Those are real advantages.
How does this differ from peco [1]?
[1]: https://github.com/peco/peco
[1]: https://github.com/peco/peco
ctrl-r & fuck have been more than enough for me most of the time
I personally use fzf [0] for a similar functionality. It brings fuzzy-search to a lot of aspects of the shell, like reverse search but also autocompletion of `kill` for example.
[0]: https://github.com/junegunn/fzf