You get all this and more with a direct perl one liner. Without the interactivity. I'd argue that if it is a lot of files, interactivity would be a pain. Also, since the original is preserved as a .bak file, one can be fearless about trying
#change xxx to yyy in all html
bash> perl -pi.bak -e 's/xxx/yyy/g' *.html
#change xxx10 (say) to yyy10 in all html
bash> perl -pi.bak -e 's/xxx(\d+)/yyy$1/g' *.html
# Change x4 to yyyy, where the number of y's equals to the number after x.
bash> perl -pi.bak -e 's/x(\d+)/"y" x $1/ge' *.
The last example shows the /e operator, which evaluates an expression and uses the result as substitute, instead of a simple string.
And finally, to exclude files, one can use a subshell. For example, suppose you want to change all html, but exclude undesirable.html..
Since we're all giving replacements to this, nobody's mentioned my preferred one, so: use git add's patch mode. On a clean worktree, do the search/replace in bulk. Then use `git add --patch` to selectively add the good replacements and skip the bad ones. Finally, `git checkout -- .` to throw away all the bad ones. The nice part about this is that it's not much to remember. If you can make a global search and replace, and you can use `git add --patch` (which is useful loads of times), you can do a selective search and replace by combining them.
As far as this actual tool: the demo GIF is way too fast-paced to show what's going on. A better demo would maybe search "ring", have 10 or so results instead of pages and pages, and show how you can unselect "spring" matches which were unintentionally caught.
I dunno, seems reasonable to me that we might have nice things without requiring everyone to use emacs. (And for those who do use emacs, I guess you're ahead of the curve?)
On the other hand, it seems reasonable that we should be able to have nice things without giving up our editors. I know I’ve been spoiled by Kakoune’s cursors, but this feels like a tool that should work by spawning $EDITOR in the middle of its execution (or perhaps just having two phases and a control file). I don’t know if that’s actually possible with the current capabilities of $EDITORs (which are not Emacs). I just feel, in the darkest hour of the night which I spend reflecting on UIs, like it should be.
You could also use vim in a loop. Say you want to replace "hello" in all files in the current dir with "world" and confirm every replace, then you would do:
for f in $(grep -l 'hello' *); do vim -c ':%s/hello/world/gc | wq' "$f"; done
Or if you want to use some more vim magic, this simpler command will do the same:
vim -c "argdo %s/hello/world/gce | update" -c "qall" *
"argdo" will do the replace command for each file, the additional e modifier in gce will ignore files that do not contain the search string and the second command "qall" is to quit vim after the work is done.
There was another comment about the difficulty in installing scooter and in the issues section, there are some requests to add more installation options.
In Emacs, there is [helm-ag-edit](https://github.com/emacsorphanage/helm-ag) (but uses ripgrep if present). It's almost identical to your workflow, but all done inside the same app.
1. helm-ag <pattern> # the search results are updated as you type
2. helm-ag-edit # edit the search result as regular text. Use multi-cursors, macros, whatever.
3. helm-ag-edit-save # commits the changes to the affected files
All those commands have keybindings, so it's pretty fast. I'll often open up Emacs just to do that and then go back to my JetBrains IDE.
"For a Linux user, you can already build such a system yourself quite trivially by getting an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem"
It's great and clearly the community appreciates it! I'll put Show HN in the title since that's the convention for sharing one's projects on HN (https://news.ycombinator.com/showhn.html).
Btw, do you want to include some text giving the backstory of how you came to work on this, and explaining what's different about it? that's also the convention. If you post it in a reply to this comment, I'll move your text to the top of the thread.
I'm using this quickly put-together shell script called replace
#!/usr/bin/env bash
# Function to escape special characters for sed
escape_sed_string() {
printf '%s\n' "$1" | gsed -e 's/[]\/$*.^[]/\\&/g'
}
help() {
gum style --foreground cyan --italic "\
Usage (everything optional, you will be prompted):\n\
$0\n\
--ext .js --ext .ts\n\
--from \"source string\"\n\
--to \"replacement string\"\n\
--dir somePath"
}
# Parse command line arguments
while [[ "$#" -gt 0 ]]; do
case $1 in
-h)
help
exit 0
;;
--help)
help
exit 0
;;
--ext) EXTENSIONS+=("$2"); shift ;;
--from) REPLACE_FROM="$2"; shift ;;
--to) REPLACE_TO="$2"; shift ;;
--dir) DIRECTORY="$2"; shift ;;
*) gum style --foreground red --bold "Unknown parameter: $1"; exit 1 ;;
esac
shift
done
# Check for missing parameters and prompt using gum
if [ -z "${EXTENSIONS+set}" ]; then
EXTENSIONS=($(gum choose \
--no-limit \
--selected .ts,.mts,.tsx,.vue,.js,.cjs,.mjs \
.ts .mts .tsx .vue .js .cjs .mjs .txt .md .html .json))
fi
# Exit if no extension is selected
if [ ${#EXTENSIONS[@]} -eq 0 ]; then
gum style --foreground red --bold " Error: No extensions selected. Exiting."
exit 1
fi
if [ -z "${REPLACE_FROM+set}" ]; then
REPLACE_FROM=$(gum input --placeholder "Search string:")
if [ -z "${REPLACE_FROM}" ]; then
echo "No replace from string, exiting"
exit 1
fi
fi
if [ -z "${REPLACE_TO+set}" ]; then
REPLACE_TO=$(gum input --placeholder "Replace string:")
fi
if [ -z "${DIRECTORY+set}" ]; then
DIRECTORY="."
fi
# Escape strings for sed
ESCAPED_FROM=$(escape_sed_string "$REPLACE_FROM")
ESCAPED_TO=$(escape_sed_string "$REPLACE_TO")
# Run the replacement
for ext in "${EXTENSIONS[@]}"; do
gum style --foreground blue " Replacing ${ext} files..."
find "$DIRECTORY" -type f -name "*$ext" ! -path "*/node_modules/*" -exec gsed -i "s/$ESCAPED_FROM/$ESCAPED_TO/g" {} \;
done
gum style --foreground green --bold " Replacement complete."
You get all this and more with a direct perl one liner. Without the interactivity. I'd argue that if it is a lot of files, interactivity would be a pain. Also, since the original is preserved as a .bak file, one can be fearless about trying
The last example shows the /e operator, which evaluates an expression and uses the result as substitute, instead of a simple string.And finally, to exclude files, one can use a subshell. For example, suppose you want to change all html, but exclude undesirable.html..
Since we're all giving replacements to this, nobody's mentioned my preferred one, so: use git add's patch mode. On a clean worktree, do the search/replace in bulk. Then use `git add --patch` to selectively add the good replacements and skip the bad ones. Finally, `git checkout -- .` to throw away all the bad ones. The nice part about this is that it's not much to remember. If you can make a global search and replace, and you can use `git add --patch` (which is useful loads of times), you can do a selective search and replace by combining them.
As far as this actual tool: the demo GIF is way too fast-paced to show what's going on. A better demo would maybe search "ring", have 10 or so results instead of pages and pages, and show how you can unselect "spring" matches which were unintentionally caught.
We're losing the art of bash ``` find -type f -iname '*.go' | xargs -r -n1 sed -i 's,foo,foobar,g' ```
Cool! is it possible to support structural search like ast-grep[1]? ast-grep has some interactive mode but it is nothing near Scooter.
1: https://ast-grep.github.io/
Neat ... but I'd probably just do this by opening a "grep -l" list in nano for interactive replacement directly instead. Easy peasy.
Very cool! I currently use `sad` for this, if you're already an fzf user you should check it out.
https://github.com/ms-jpq/sad
Cool.
I assumed it uses ripgrep (or the underlying walkdir) because that's the established high-performance tool for this. But apparently not.
It uses https://docs.rs/ignore/latest/ignore/ to walk dirs while respecting ignore files
(And `ignore` uses `walkdir` internally)
For single threaded use cases. For multi-threaded, it has its own parallel directory traversal. :-)
Feels like we just keep making tools that already exist in Emacs.
I dunno, seems reasonable to me that we might have nice things without requiring everyone to use emacs. (And for those who do use emacs, I guess you're ahead of the curve?)
On the other hand, it seems reasonable that we should be able to have nice things without giving up our editors. I know I’ve been spoiled by Kakoune’s cursors, but this feels like a tool that should work by spawning $EDITOR in the middle of its execution (or perhaps just having two phases and a control file). I don’t know if that’s actually possible with the current capabilities of $EDITORs (which are not Emacs). I just feel, in the darkest hour of the night which I spend reflecting on UIs, like it should be.
Nonsense, lots of people are doing text editors.
nice! Find and replace across a codebase is one of the few times I open an IDE.
Being able to interactively ignore instances for replacement is great!
Am I alone in initially thinking this was specifically for the fish shell because of this tool's name?
Perhaps. I as a fish user thought “oh, like `string replace`”
Excellent, thank you. I do this with sed & awk & sometimes an IDE, and scooter looks better in every way.
I'm adding scooter to my cargo install favorites:
https://github.com/sixarm/cargo-install-favorites
Very nice, it might be a good alternative when I can't use vscode remote connections.
A Homebrew install option will help this take off on Macs. https://github.com/thomasschafer/scooter/issues/6
That was my first problem with trying to install this. But agree that it should be on Homebrew.
You could also use vim in a loop. Say you want to replace "hello" in all files in the current dir with "world" and confirm every replace, then you would do:
Or if you want to use some more vim magic, this simpler command will do the same: "argdo" will do the replace command for each file, the additional e modifier in gce will ignore files that do not contain the search string and the second command "qall" is to quit vim after the work is done.There was another comment about the difficulty in installing scooter and in the issues section, there are some requests to add more installation options.
https://github.com/thomasschafer/scooter/issues/6
Not everyone has the Rust toolchain installed on their machine. The `cargo install` installation directive needs to be discouraged.
Similarly, on bash/ksh: set -o vi Ctrl-[ v (or ESC) set -o emacs Ctrl-x e
A useful feature of bash and zsh is the "edit command". The standard shortcut is "ctrl-x ctrl-e".
It opens the current command line in $EDITOR, which often defaults to vim.
That is very useful. What does it have to do with this?
If you want to search and replace a command line, there's tools to do it in your favourite editor.
Ah, so you didn't click through and actually see what this tool is, you just read the title.
I did click through, but misinterpreted what it was doing. Apologies, I'm "multitasking".
Looks handy!
Couldn't find it in nixpkgs.
I opened a PR for it: https://github.com/NixOS/nixpkgs/pull/356310
In the meantime you can also add packages that aren't yet in nixpkgs using pkgs.callPackage.
Also see the excellent https://github.com/your-tools/ruplacer.
For more advanced needs, I have a custom thing called greprep that let's you make changes using your favorite editor. Workflow is like this:
In Emacs, there is [helm-ag-edit](https://github.com/emacsorphanage/helm-ag) (but uses ripgrep if present). It's almost identical to your workflow, but all done inside the same app.
1. helm-ag <pattern> # the search results are updated as you type 2. helm-ag-edit # edit the search result as regular text. Use multi-cursors, macros, whatever. 3. helm-ag-edit-save # commits the changes to the affected files
All those commands have keybindings, so it's pretty fast. I'll often open up Emacs just to do that and then go back to my JetBrains IDE.
[flagged]
Can you please not post shallow dismissals of other people's work? This is in the site guidelines: https://news.ycombinator.com/newsguidelines.html.
It's important, when people share something they've made on HN, that they don't run into this sort of bilious internet putdown.
Edit - these are other examples of the same thing (i.e. the thing we don't want in HN threads, and which we'd appreciate if you'd not do any more of):
https://news.ycombinator.com/item?id=41810426
https://news.ycombinator.com/item?id=41224056
"For a Linux user, you can already build such a system yourself quite trivially by getting an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem"
https://news.ycombinator.com/item?id=8863
Not affiliated, I just built a little tool to make my life easier and thought I'd share
It's great and clearly the community appreciates it! I'll put Show HN in the title since that's the convention for sharing one's projects on HN (https://news.ycombinator.com/showhn.html).
Btw, do you want to include some text giving the backstory of how you came to work on this, and explaining what's different about it? that's also the convention. If you post it in a reply to this comment, I'll move your text to the top of the thread.
I think it's cool. Thanks for sharing
I'm using this quickly put-together shell script called replace
What is gum?