perf: ignore instead of wax

Cargo.toml

@@ -123,7 +123,6 @@ uv-platform-tags = { git = "https://github.com/astral-sh/uv", tag = "0.8.5" }
 uv-pypi-types = { git = "https://github.com/astral-sh/uv", tag = "0.8.5" }
 uv-requirements-txt = { git = "https://github.com/astral-sh/uv", tag = "0.8.5" }
 
-wax = "0.6.0"
 which = "8.0.0"
 
 # Rattler crates

crates/pixi_command_dispatcher/src/source_build_cache_status/mod.rs

@@ -18,11 +18,6 @@ use crate::{
     },
 };
 
-/// A list of globs that should be ignored when calculating any input hash.
-/// These are typically used for build artifacts that should not be included in
-/// the input hash.
-pub const DEFAULT_BUILD_IGNORE_GLOBS: &[&str] = &["!.pixi/**"];
-
 /// A query to retrieve information from the source build cache. This is
 /// memoized to allow querying information from the cache while it is also
 /// overwritten at the same time by a build.
@@ -322,11 +317,7 @@ impl SourceBuildCacheStatusSpec {
         // Compute the modification time of the files that match the source input globs.
         let glob_time = match GlobModificationTime::from_patterns(
             &source_checkout.path,
-            source_info
-                .globs
-                .iter()
-                .map(String::as_str)
-                .chain(DEFAULT_BUILD_IGNORE_GLOBS.iter().copied()),
+            source_info.globs.iter().map(String::as_str), // .chain(DEFAULT_BUILD_IGNORE_GLOBS.iter().copied()),
         ) {
             Ok(glob_time) => glob_time,
             Err(e) => {

crates/pixi_glob/Cargo.toml

@@ -12,14 +12,17 @@ version = "0.1.0"
 [dependencies]
 dashmap = { workspace = true }
 fs-err = { workspace = true }
+ignore = "0.4"
 itertools = { workspace = true }
 memchr = { workspace = true }
+parking_lot = { workspace = true }
 rattler_digest = { workspace = true }
 thiserror = { workspace = true }
 tokio = { workspace = true, features = ["sync", "rt"] }
-wax = { workspace = true }
+tracing = { workspace = true }
 
 [dev-dependencies]
-insta = { workspace = true }
+insta = { workspace = true, features = ["yaml", "redactions"] }
 rstest = { workspace = true }
+serde = { workspace = true, features = ["derive"] }
 tempfile = { workspace = true }

crates/pixi_glob/src/glob_hash.rs

@@ -7,11 +7,10 @@ use std::{
     path::{Path, PathBuf},
 };
 
-use itertools::Itertools;
 use rattler_digest::{Sha256, Sha256Hash, digest::Digest};
 use thiserror::Error;
 
-use crate::glob_set::{self, GlobSet};
+use crate::{GlobSet, GlobSetError};
 
 /// Contains a hash of the files that match the given glob patterns.
 #[derive(Debug, Clone, Default)]
@@ -25,14 +24,14 @@ pub struct GlobHash {
 #[derive(Error, Debug)]
 #[allow(missing_docs)]
 pub enum GlobHashError {
-    #[error(transparent)]
-    FilterGlobError(#[from] glob_set::GlobSetError),
-
     #[error("during line normalization, failed to access {}", .0.display())]
     NormalizeLineEnds(PathBuf, #[source] io::Error),
 
     #[error("the operation was cancelled")]
     Cancelled,
+
+    #[error(transparent)]
+    GlobSetIgnore(#[from] GlobSetError),
 }
 
 impl GlobHash {
@@ -47,14 +46,12 @@ impl GlobHash {
             return Ok(Self::default());
         }
 
-        let glob_set = GlobSet::create(globs)?;
-        let mut entries = glob_set
-            .filter_directory(root_dir)
-            .collect::<Result<Vec<_>, _>>()?
-            .into_iter()
-            .map(|entry| entry.path().to_path_buf())
-            .collect_vec();
-        entries.sort();
+        let glob_set = GlobSet::create(globs);
+        // Collect matching entries and convert to concrete DirEntry list, propagating errors.
+        let mut entries: Vec<ignore::DirEntry> = glob_set.collect_matching(root_dir)?;
+
+        // Sort deterministically by path
+        entries.sort_by_key(|e| e.path().to_path_buf());
 
         #[cfg(test)]
         let mut matching_files = Vec::new();
@@ -63,17 +60,19 @@ impl GlobHash {
         for entry in entries {
             // Construct a normalized file path to ensure consistent hashing across
             // platforms. And add it to the hash.
-            let relative_path = entry.strip_prefix(root_dir).unwrap_or(&entry);
+            let relative_path = entry.path().strip_prefix(root_dir).unwrap_or(entry.path());
             let normalized_file_path = relative_path.to_string_lossy().replace("\\", "/");
             rattler_digest::digest::Update::update(&mut hasher, normalized_file_path.as_bytes());
 
             #[cfg(test)]
             matching_files.push(normalized_file_path);
 
             // Concatenate the contents of the file to the hash.
-            File::open(&entry)
+            File::open(entry.path())
                 .and_then(|mut file| normalize_line_endings(&mut file, &mut hasher))
-                .map_err(move |e| GlobHashError::NormalizeLineEnds(entry, e))?;
+                .map_err(move |e| {
+                    GlobHashError::NormalizeLineEnds(entry.path().to_path_buf(), e)
+                })?;
         }
 
         if let Some(additional_hash) = additional_hash {

crates/pixi_glob/src/glob_mtime.rs

@@ -5,7 +5,7 @@ use std::{
 
 use thiserror::Error;
 
-use crate::glob_set::{self, GlobSet};
+use crate::{GlobSet, GlobSetError};
 
 /// Contains the newest modification time for the files that match the given glob patterns.
 #[derive(Debug, Clone)]
@@ -27,7 +27,7 @@ pub enum GlobModificationTimeError {
     #[error("error calculating modification time for {}", .0.display())]
     CalculateMTime(PathBuf, #[source] std::io::Error),
     #[error(transparent)]
-    GlobSet(#[from] glob_set::GlobSetError),
+    GlobSetIgnore(#[from] GlobSetError),
 }
 
 impl GlobModificationTime {
@@ -36,38 +36,49 @@ impl GlobModificationTime {
         root_dir: &Path,
         globs: impl IntoIterator<Item = &'a str>,
     ) -> Result<Self, GlobModificationTimeError> {
-        // If the root is not a directory or does not exist, return NoMatches.
+        // Delegate to the ignore-based implementation for performance.
+        Self::from_patterns_ignore(root_dir, globs)
+    }
+
+    /// Same as `from_patterns` but uses the `ignore` crate for walking/matching.
+    pub fn from_patterns_ignore<'a>(
+        root_dir: &Path,
+        globs: impl IntoIterator<Item = &'a str>,
+    ) -> Result<Self, GlobModificationTimeError> {
+        // Normalize root to a directory if a file was passed.
         let mut root = root_dir.to_owned();
         if !root.is_dir() {
             root.pop();
         }
 
-        let glob_set = GlobSet::create(globs)?;
-        let entries: Vec<_> = glob_set
-            .filter_directory(root_dir)
-            .collect::<Result<Vec<_>, _>>()?;
+        let glob_set = GlobSet::create(globs);
+        let entries = glob_set.collect_matching(root_dir)?;
 
         let mut latest = None;
         let mut designated_file = PathBuf::new();
 
-        // Find the newest modification time and the designated file
         for entry in entries {
-            let matched_path = entry.path().to_owned();
-            let metadata = entry.metadata().map_err(|e| {
-                GlobModificationTimeError::CalculateMTime(matched_path.clone(), e.into())
-            })?;
-            let modified_entry = metadata
+            let matched_path = entry.path().to_path_buf();
+            let md = match entry.metadata() {
+                Ok(md) => md,
+                Err(e) => {
+                    return Err(GlobModificationTimeError::CalculateMTime(
+                        matched_path,
+                        std::io::Error::new(std::io::ErrorKind::Other, e.to_string()),
+                    ));
+                }
+            };
+            let modified = md
                 .modified()
                 .map_err(|e| GlobModificationTimeError::CalculateMTime(matched_path.clone(), e))?;
 
-            if let Some(ref current_latest) = latest {
-                if *current_latest >= modified_entry {
+            if let Some(cur) = latest {
+                if cur >= modified {
                     continue;
                 }
             }
-
-            latest = Some(modified_entry);
-            designated_file = matched_path.clone();
+            latest = Some(modified);
+            designated_file = matched_path;
         }
 
         match latest {