chore: development v0.2.31 - comprehensive testing complete [auto-commit]

Author: githubrobbi

Cargo.lock

@@ -32,7 +32,7 @@ members = [
 # Workspace Package Metadata (inherited by all crates)
 # ─────────────────────────────────────────────────────────────────────────────
 [workspace.package]
-version = "0.2.29"
+version = "0.2.31"
 edition = "2024"
 rust-version = "1.85"
 license = "MPL-2.0 OR LicenseRef-UFFS-Commercial"

Cargo.toml

@@ -21,7 +21,7 @@ Traditional file search tools (including `os.walk`, `FindFirstFile`, etc.) work
 
 **UFFS reads the MFT directly** - once - and queries it in memory using Polars DataFrames. This is like reading the entire phonebook once instead of looking up each name individually.
 
-### Benchmark Results (v0.2.29)
+### Benchmark Results (v0.2.31)
 
 | Drive Type | Records | Time | Throughput |
 |------------|---------|------|------------|
@@ -33,7 +33,7 @@ Traditional file search tools (including `os.walk`, `FindFirstFile`, etc.) work
 
 | Comparison | Records | Time | Notes |
 |------------|---------|------|-------|
-| **UFFS v0.2.29** | **18.7 Million** | **~142 seconds** | All disks, fast mode |
+| **UFFS v0.2.31** | **18.7 Million** | **~142 seconds** | All disks, fast mode |
 | UFFS v0.1.30 | 18.7 Million | ~315 seconds | Baseline |
 | Everything | 19 Million | 178 seconds | All disks |
 | WizFile | 6.5 Million | 299 seconds | Single HDD |

README.md

@@ -541,7 +541,7 @@ async fn load_and_filter_data(
 
         // Build path resolver from FULL data BEFORE filtering
         let t_resolver = std::time::Instant::now();
-        let mut path_resolver = if needs_paths {
+        let path_resolver = if needs_paths {
             Some(
                 uffs_core::FastPathResolver::build(&full_df, drive_letter)
                     .context("Failed to build path resolver")?,
@@ -557,12 +557,11 @@ async fn load_and_filter_data(
         let filter_ms = t_filter.elapsed().as_millis();
         let filtered_count = filtered.height();
 
-        // Add paths using the pre-built resolver (auto-selects parallel for large
-        // DataFrames)
+        // Add paths using the pre-built resolver with directory suffix (C++ parity)
         let t_paths = std::time::Instant::now();
-        if let Some(ref mut resolver) = path_resolver {
+        if let Some(resolver) = &path_resolver {
             filtered = resolver
-                .add_path_column_auto(&filtered)
+                .add_path_column_with_dir_suffix(&filtered)
                 .context("Failed to add path column")?;
             // Add path_only column (directory portion of path)
             filtered = uffs_core::add_path_only_column(&filtered)
@@ -979,10 +978,9 @@ async fn search_multi_drive_filtered(
 
             let matches = filtered.height();
 
-            // Add paths using the pre-built resolver (auto-selects parallel for large
-            // DataFrames)
-            let with_paths = if let Some(ref mut resolver) = path_resolver {
-                match resolver.add_path_column_auto(&filtered) {
+            // Add paths using the pre-built resolver with directory suffix (C++ parity)
+            let with_paths = if let Some(resolver) = &path_resolver {
+                match resolver.add_path_column_with_dir_suffix(&filtered) {
                     Ok(df) => {
                         // Add path_only column (directory portion of path)
                         match uffs_core::add_path_only_column(&df) {
@@ -1272,9 +1270,9 @@ async fn search_multi_drive_streaming<W: Write + Send + 'static>(
 
             let matches = filtered.height();
 
-            // Add paths using the pre-built resolver
-            let with_paths = if let Some(ref mut resolver) = path_resolver {
-                match resolver.add_path_column_auto(&filtered) {
+            // Add paths using the pre-built resolver with directory suffix (C++ parity)
+            let with_paths = if let Some(resolver) = &path_resolver {
+                match resolver.add_path_column_with_dir_suffix(&filtered) {
                     Ok(df) => {
                         // Add path_only column (directory portion of path)
                         match uffs_core::add_path_only_column(&df) {

crates/uffs-cli/src/commands.rs

@@ -385,6 +385,57 @@ impl FastPathResolver {
         }
     }
 
+    /// Add a "path" column with trailing slashes for directories (C++ parity).
+    ///
+    /// This method adds a trailing backslash to directory paths to match
+    /// the C++ `UltraFastFileSearch` output format.
+    ///
+    /// # Errors
+    ///
+    /// Returns an error if the frs or `is_directory` columns are missing.
+    pub fn add_path_column_with_dir_suffix(&self, df: &DataFrame) -> Result<DataFrame> {
+        let frs_col = df.column("frs")?.u64()?;
+        let is_dir_col = df.column("is_directory")?.bool()?;
+        let stream_name_col = df.column("stream_name").ok().and_then(|col| col.str().ok());
+
+        // Collect values for parallel iteration
+        let frs_values: Vec<Option<u64>> = frs_col.into_iter().collect();
+        let is_dir_values: Vec<Option<bool>> = is_dir_col.into_iter().collect();
+        let stream_names: Vec<Option<&str>> = stream_name_col.map_or_else(
+            || vec![None; frs_values.len()],
+            |col| col.into_iter().collect(),
+        );
+
+        // Resolve paths in parallel with directory suffix
+        let paths: Vec<String> = frs_values
+            .par_iter()
+            .zip(is_dir_values.par_iter())
+            .zip(stream_names.par_iter())
+            .map(|((frs, is_dir), stream_name)| {
+                let mut path =
+                    frs.map_or_else(|| "<null>".to_owned(), |frs_val| self.resolve(frs_val));
+                // Add trailing backslash for directories
+                if is_dir.unwrap_or(false) && !path.ends_with('\\') {
+                    path.push('\\');
+                }
+                // Append stream name for ADS (e.g., "file.txt:Zone.Identifier")
+                if let Some(stream) = stream_name {
+                    if !stream.is_empty() {
+                        path.push(':');
+                        path.push_str(stream);
+                    }
+                }
+                path
+            })
+            .collect();
+
+        let path_series = Column::new("path".into(), paths);
+        let mut result = df.clone();
+        result.with_column(path_series)?;
+
+        Ok(result)
+    }
+
     /// Get statistics about the resolver.
     #[must_use]
     pub fn stats(&self) -> FastPathResolverStats {

crates/uffs-core/src/path_resolver.rs

@@ -812,6 +812,8 @@ pub struct ParsedColumns {
     pub name_count: Vec<u16>,
     /// Number of data streams per record.
     pub stream_count: Vec<u16>,
+    /// Stream name (empty for default stream, non-empty for ADS).
+    pub stream_name: Vec<String>,
 
     // Attribute flags (all boolean columns for C++ parity)
     /// Read-only flag.
@@ -876,6 +878,7 @@ impl ParsedColumns {
             is_directory: Vec::with_capacity(capacity),
             name_count: Vec::with_capacity(capacity),
             stream_count: Vec::with_capacity(capacity),
+            stream_name: Vec::with_capacity(capacity),
             is_readonly: Vec::with_capacity(capacity),
             is_hidden: Vec::with_capacity(capacity),
             is_system: Vec::with_capacity(capacity),
@@ -925,6 +928,7 @@ impl ParsedColumns {
         self.is_directory.push(record.is_directory);
         self.name_count.push(record.name_count());
         self.stream_count.push(record.stream_count());
+        self.stream_name.push(String::new()); // Default stream (no ADS)
         self.is_readonly.push(record.std_info.is_readonly);
         self.is_hidden.push(record.std_info.is_hidden);
         self.is_system.push(record.std_info.is_system);
@@ -946,56 +950,82 @@ impl ParsedColumns {
         self.flags.push(record.std_info.to_raw_flags());
     }
 
-    /// Pushes a record with hard link expansion.
+    /// Pushes a record with full expansion (names × streams).
     ///
-    /// This matches C++ behavior: one row per hard link (name).
-    /// If a file has 3 hard links, this creates 3 rows with different
-    /// name/parent_frs but the same FRS and other attributes.
+    /// This matches C++ behavior: one row per (hard link × stream) combination.
+    /// If a file has 2 hard links and 3 streams, this creates 6 rows.
     ///
     /// This is the default behavior for user-facing output, as users
-    /// expect to see each hard link as a separate entry (matching Explorer).
+    /// expect to see each hard link and ADS as separate entries.
     #[inline]
     pub fn push_record_expanded(&mut self, record: &ParsedRecord) {
-        if record.names.is_empty() {
-            // No names - use the primary name (fallback)
-            self.push_record(record);
-            return;
-        }
-
-        // Create one row per hard link
-        for name_info in &record.names {
-            self.frs.push(record.frs);
-            self.parent_frs.push(name_info.parent_frs);
-            self.name.push(name_info.name.clone());
-            self.size.push(record.size);
-            self.allocated_size.push(record.allocated_size);
-            self.created.push(record.std_info.created);
-            self.modified.push(record.std_info.modified);
-            self.accessed.push(record.std_info.accessed);
-            self.mft_changed.push(record.std_info.mft_changed);
-            self.is_directory.push(record.is_directory);
-            // For expanded records, name_count is always 1 (this row represents one link)
-            self.name_count.push(1);
-            self.stream_count.push(record.stream_count());
-            self.is_readonly.push(record.std_info.is_readonly);
-            self.is_hidden.push(record.std_info.is_hidden);
-            self.is_system.push(record.std_info.is_system);
-            self.is_archive.push(record.std_info.is_archive);
-            self.is_compressed.push(record.std_info.is_compressed);
-            self.is_encrypted.push(record.std_info.is_encrypted);
-            self.is_sparse.push(record.std_info.is_sparse);
-            self.is_reparse.push(record.std_info.is_reparse);
-            self.is_offline.push(record.std_info.is_offline);
-            self.is_not_indexed
-                .push(record.std_info.is_not_content_indexed);
-            self.is_temporary.push(record.std_info.is_temporary);
-            self.is_integrity_stream
-                .push(record.std_info.is_integrity_stream);
-            self.is_no_scrub_data.push(record.std_info.is_no_scrub_data);
-            self.is_pinned.push(record.std_info.is_pinned);
-            self.is_unpinned.push(record.std_info.is_unpinned);
-            self.is_virtual.push(record.std_info.is_virtual);
-            self.flags.push(record.std_info.to_raw_flags());
+        // Get names to iterate over (use primary name if names is empty)
+        let names: Vec<_> = if record.names.is_empty() {
+            vec![NameInfo {
+                name: record.name.clone(),
+                parent_frs: record.parent_frs,
+                namespace: 3, // Win32+DOS
+            }]
+        } else {
+            record.names.clone()
+        };
+
+        // Get streams to iterate over (use empty stream if streams is empty)
+        let streams: Vec<_> = if record.streams.is_empty() {
+            vec![StreamInfo {
+                name: String::new(),
+                size: record.size,
+                allocated_size: record.allocated_size,
+                is_sparse: false,
+                is_compressed: false,
+            }]
+        } else {
+            record.streams.clone()
+        };
+
+        // Create one row per (name × stream) combination
+        for name_info in &names {
+            for stream_info in &streams {
+                self.frs.push(record.frs);
+                self.parent_frs.push(name_info.parent_frs);
+                self.name.push(name_info.name.clone());
+                // Use stream-specific size for ADS, file size for default stream
+                let (size, alloc) = if stream_info.name.is_empty() {
+                    (record.size, record.allocated_size)
+                } else {
+                    (stream_info.size, stream_info.allocated_size)
+                };
+                self.size.push(size);
+                self.allocated_size.push(alloc);
+                self.created.push(record.std_info.created);
+                self.modified.push(record.std_info.modified);
+                self.accessed.push(record.std_info.accessed);
+                self.mft_changed.push(record.std_info.mft_changed);
+                self.is_directory.push(record.is_directory);
+                // For expanded records, counts are 1 (this row = one link + one stream)
+                self.name_count.push(1);
+                self.stream_count.push(1);
+                self.stream_name.push(stream_info.name.clone());
+                self.is_readonly.push(record.std_info.is_readonly);
+                self.is_hidden.push(record.std_info.is_hidden);
+                self.is_system.push(record.std_info.is_system);
+                self.is_archive.push(record.std_info.is_archive);
+                self.is_compressed.push(record.std_info.is_compressed);
+                self.is_encrypted.push(record.std_info.is_encrypted);
+                self.is_sparse.push(record.std_info.is_sparse);
+                self.is_reparse.push(record.std_info.is_reparse);
+                self.is_offline.push(record.std_info.is_offline);
+                self.is_not_indexed
+                    .push(record.std_info.is_not_content_indexed);
+                self.is_temporary.push(record.std_info.is_temporary);
+                self.is_integrity_stream
+                    .push(record.std_info.is_integrity_stream);
+                self.is_no_scrub_data.push(record.std_info.is_no_scrub_data);
+                self.is_pinned.push(record.std_info.is_pinned);
+                self.is_unpinned.push(record.std_info.is_unpinned);
+                self.is_virtual.push(record.std_info.is_virtual);
+                self.flags.push(record.std_info.to_raw_flags());
+            }
         }
     }
 
@@ -1015,6 +1045,7 @@ impl ParsedColumns {
         self.is_directory.extend(other.is_directory);
         self.name_count.extend(other.name_count);
         self.stream_count.extend(other.stream_count);
+        self.stream_name.extend(other.stream_name);
         self.is_readonly.extend(other.is_readonly);
         self.is_hidden.extend(other.is_hidden);
         self.is_system.extend(other.is_system);
@@ -1048,6 +1079,7 @@ impl ParsedColumns {
         self.is_directory.reserve(additional);
         self.name_count.reserve(additional);
         self.stream_count.reserve(additional);
+        self.stream_name.reserve(additional);
         self.is_readonly.reserve(additional);
         self.is_hidden.reserve(additional);
         self.is_system.reserve(additional);