feat: add --ignore-ext and --only-ext flags for extension filtering

[ericzakariasson]

Note

Medium Risk
Changes scan inclusion/exclusion behavior across multiple scanners and affects scan caching keys; incorrect normalization or filter interaction could silently hide files from results.

Overview
Adds extension-based filtering to the CLI and config via new --ignore-ext and --only-ext flags (comma-delimited) and persisted config fields (ignored_extensions, only_extensions).

Applies these filters during scanning (downloads/large/old) through a new Config::should_skip_extension helper and updates scan-cache fingerprinting so cached results are invalidated when extension filters change, while normalizing configured/CLI extensions for consistent matching.

^{Written by Cursor Bugbot for commit 150f89e. This will update automatically on new commits. Configure here.}

Made with Cursor

[ericzakariasson]

@cursor push f9c9592

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

Bugbot Autofix prepared fixes for both issues found in the latest run.

• ✅ Fixed: Missing whitespace trim causes silent filter failures
  • I updated normalize_extension to trim surrounding whitespace before stripping leading dots and lowercasing, and added a test covering spaced extension inputs.
• ✅ Fixed: Extension filtering inconsistently applied across file scanners
  • I added config.should_skip_extension(path) filtering in the duplicates scanner so --ignore-ext and --only-ext are consistently enforced there.

Or push these changes by commenting:

@cursor push d9db976547

Preview (d9db976547)

diff --git a/src/config.rs b/src/config.rs
--- a/src/config.rs
+++ b/src/config.rs
@@ -348,7 +348,7 @@
 }
 
 fn normalize_extension(ext: &str) -> String {
-    ext.trim_start_matches('.').to_lowercase()
+    ext.trim().trim_start_matches('.').to_lowercase()
 }
 
 #[cfg(test)]
@@ -373,6 +373,12 @@
     }
 
     #[test]
+    fn test_normalize_extension_trims_whitespace() {
+        assert_eq!(normalize_extension(" iso"), "iso");
+        assert_eq!(normalize_extension(" .DMG "), "dmg");
+    }
+
+    #[test]
     fn test_is_excluded_plain_pattern() {
         let mut config = Config::default();
         config.excluded_paths = vec![".local/share/cursor-agent".to_string()];

diff --git a/src/scanner/duplicates.rs b/src/scanner/duplicates.rs
--- a/src/scanner/duplicates.rs
+++ b/src/scanner/duplicates.rs
@@ -102,6 +102,11 @@
                 continue;
             }
 
+            // Skip based on extension filters
+            if config.should_skip_extension(path) {
+                continue;
+            }
+
             // Skip hidden files
             if let Some(name) = path.file_name() {
                 if name.to_string_lossy().starts_with('.') {

_{You can send follow-ups to this agent here.}

[cursor]

Missing whitespace trim causes silent filter failures

Medium Severity · Logic Bug

normalize_extension only strips leading dots but doesn't trim whitespace. With clap's value_delimiter = ',', input like --ignore-ext "dmg, iso, pkg" produces ["dmg", " iso", " pkg"]. The leading space in " iso" survives normalization, so it never matches path.extension() which returns "iso" (no space). Extensions after the first one silently fail to filter. Adding .trim() before the dot-strip would fix this.

 

[cursor]

Extension filtering inconsistently applied across file scanners

Low Severity · Logic Bug

should_skip_extension is applied to downloads, large files, and old files scanners, but not to the duplicates scanner — which also iterates individual files. A user running duster scan --all --only-ext dmg would still see duplicate results for non-.dmg files, contradicting the documented behavior of "only include files with these extensions."

Additional Locations (2)

• src/scanner/downloads.rs#L62-L66
• src/scanner/old_files.rs#L117-L121