feat: add strip_trailing_whitespace, on by default (#3390)

I can't think of any reason you would want this disabled by default -
trailing whitespace means nothing, breaks dedupe, and wastes a few bytes

closes #3387 

## Checks
- [ ] I am happy for maintainers to push small adjustments to this PR,
to speed up the review cycle
- [ ] I have checked that there are no existing pull requests for the
same thing

3 files changed, 97 insertions, 0 deletions

diff --git a/crates/atuin-client/config.toml b/crates/atuin-client/config.toml
index 6e67a4e1..0d0672bf 100644
--- a/crates/atuin-client/config.toml
+++ b/crates/atuin-client/config.toml
@@ -113,6 +113,11 @@
 ## default history list format - can also be specified with the --format arg
 # history_format = "{time}\t{command}\t{duration}"
 
+## Defaults to true. If enabled, strip trailing spaces and tabs from commands
+## before saving them to history.
+## Escaped trailing spaces (for example `printf foo\\ `) are preserved.
+# strip_trailing_whitespace = true
+
 ## prevent commands matching any of these regexes from being written to history.
 ## Note that these regular expressions are unanchored, i.e. if they don't start
 ## with ^ or end with $, they'll match anywhere in the command.
diff --git a/crates/atuin-client/src/settings.rs b/crates/atuin-client/src/settings.rs
index 22b892d1..e2624136 100644
--- a/crates/atuin-client/src/settings.rs
+++ b/crates/atuin-client/src/settings.rs
@@ -1092,6 +1092,7 @@ pub struct Settings {
     pub word_chars: String,
     pub scroll_context_lines: usize,
     pub history_format: String,
+    pub strip_trailing_whitespace: bool,
     pub prefers_reduced_motion: bool,
     pub store_failed: bool,
     pub no_mouse: bool,
@@ -1497,6 +1498,7 @@ impl Settings {
             .set_default("workspaces", false)?
             .set_default("ctrl_n_shortcuts", false)?
             .set_default("secrets_filter", true)?
+            .set_default("strip_trailing_whitespace", true)?
             .set_default("network_connect_timeout", 5)?
             .set_default("network_timeout", 30)?
             .set_default("local_timeout", 2.0)?
diff --git a/crates/atuin/src/command/client/history.rs b/crates/atuin/src/command/client/history.rs
index fe9a7e32..4d81d144 100644
--- a/crates/atuin/src/command/client/history.rs
+++ b/crates/atuin/src/command/client/history.rs
@@ -365,6 +365,30 @@ fn parse_fmt(format: &str) -> ParsedFmt<'_> {
 }
 
 impl Cmd {
+    fn normalize_command_for_storage<'a>(command: &'a str, settings: &Settings) -> &'a str {
+        if !settings.strip_trailing_whitespace {
+            return command;
+        }
+
+        let trimmed = command.trim_end_matches([' ', '\t']);
+        if trimmed.len() == command.len() {
+            return command;
+        }
+
+        let trailing_backslashes = trimmed
+            .as_bytes()
+            .iter()
+            .rev()
+            .take_while(|&&byte| byte == b'\\')
+            .count();
+
+        if trailing_backslashes % 2 == 1 {
+            command
+        } else {
+            trimmed
+        }
+    }
+
     fn apply_start_metadata(history: &mut History, author: Option<&str>, intent: Option<&str>) {
         if let Some(author) = author.map(str::trim).filter(|author| !author.is_empty()) {
             author.clone_into(&mut history.author);
@@ -388,6 +412,7 @@ impl Cmd {
         // It's better for atuin to silently fail here and attempt to
         // store whatever is ran, than to throw an error to the terminal
         let cwd = utils::get_current_dir();
+        let command = Self::normalize_command_for_storage(command, settings);
 
         let mut h: History = History::capture()
             .timestamp(OffsetDateTime::now_utc())
@@ -424,6 +449,7 @@ impl Cmd {
         // It's better for atuin to silently fail here and attempt to
         // store whatever is ran, than to throw an error to the terminal
         let cwd = utils::get_current_dir();
+        let command = Self::normalize_command_for_storage(command, settings);
 
         let mut h: History = History::capture()
             .timestamp(OffsetDateTime::now_utc())
@@ -831,6 +857,70 @@ mod tests {
     use super::*;
 
     #[test]
+    fn normalize_command_strips_trailing_spaces_and_tabs() {
+        let settings = Settings::utc();
+
+        assert!(settings.strip_trailing_whitespace);
+        assert_eq!(
+            Cmd::normalize_command_for_storage("ls   \t", &settings),
+            "ls"
+        );
+    }
+
+    #[test]
+    fn normalize_command_preserves_escaped_trailing_space() {
+        let settings = Settings::utc();
+
+        assert_eq!(
+            Cmd::normalize_command_for_storage("printf foo\\ ", &settings),
+            "printf foo\\ "
+        );
+        assert_eq!(
+            Cmd::normalize_command_for_storage("printf foo\\\\ ", &settings),
+            "printf foo\\\\"
+        );
+    }
+
+    #[tokio::test]
+    async fn handle_start_saves_trimmed_command() {
+        let db = Sqlite::new("sqlite::memory:", 2.0).await.unwrap();
+        let settings = Settings::utc();
+
+        Cmd::handle_start(&db, &settings, "ls   \t", None, None)
+            .await
+            .unwrap();
+
+        let history = db
+            .before(OffsetDateTime::now_utc() + time::Duration::SECOND, 1)
+            .await
+            .unwrap()
+            .pop()
+            .unwrap();
+        assert_eq!(history.command, "ls");
+    }
+
+    #[tokio::test]
+    async fn handle_start_can_keep_trailing_whitespace() {
+        let db = Sqlite::new("sqlite::memory:", 2.0).await.unwrap();
+        let settings = Settings {
+            strip_trailing_whitespace: false,
+            ..Settings::utc()
+        };
+
+        Cmd::handle_start(&db, &settings, "ls   \t", None, None)
+            .await
+            .unwrap();
+
+        let history = db
+            .before(OffsetDateTime::now_utc() + time::Duration::SECOND, 1)
+            .await
+            .unwrap()
+            .pop()
+            .unwrap();
+        assert_eq!(history.command, "ls   \t");
+    }
+
+    #[test]
     fn test_format_string_no_panic() {
         // Don't panic but provide helpful output (issue #2776)
         let malformed_json = r#"{"command":"{command}","key":"value"}"#;