What are the trade-offs between `glob::Pattern::new` and compiled `glob::Pattern` for repeated matching?

glob::Pattern::new compiles a glob pattern string into a reusable Pattern object each time it's called, while reusing a compiled Pattern avoids repeated parsing and compilation overhead when matching multiple paths. The Pattern::new function parses the glob syntax (*, ?, [], **) and builds an internal representation suitable for matching—a non-trivial operation involving syntax validation, bracket expression parsing, and wildcard handling. When matching many files against the same pattern, compiling once and reusing the Pattern yields significant performance benefits. However, for one-off matches or patterns that change frequently, the convenience of Pattern::new inline may outweigh the optimization. The Pattern struct also provides the compile method for more explicit pattern creation, though new is the typical entry point.

Basic Pattern Creation

use glob::Pattern;
 
fn main() {
    // Create a pattern from a string
    let pattern = Pattern::new("src/**/*.rs").unwrap();
    
    // Use the pattern to match paths
    assert!(pattern.matches("src/main.rs"));
    assert!(pattern.matches("src/lib/mod.rs"));
    assert!(!pattern.matches("test/main.rs"));
    
    println!("Pattern compiled successfully");
}

Pattern::new parses and compiles the glob string into a reusable matcher.

Pattern Matching Methods

use glob::Pattern;
 
fn main() {
    let pattern = Pattern::new("*.rs").unwrap();
    
    // matches() - simple boolean check
    println!("matches 'main.rs': {}", pattern.matches("main.rs"));
    println!("matches 'lib.txt': {}", pattern.matches("lib.txt"));
    
    // matches_path() - matches against full Path
    let path = std::path::Path::new("src/main.rs");
    println!("matches_path: {}", pattern.matches_path(path));
    
    // matches_with() - control path separator handling
    use glob::MatchOptions;
    let options = MatchOptions {
        case_sensitive: false,
        require_literal_separator: false,
        require_literal_leading_dot: false,
    };
    println!("case-insensitive: {}", pattern.matches_with("MAIN.RS", &options));
}

Pattern provides multiple matching methods with different behaviors.

The Cost of Repeated Compilation

use glob::Pattern;
use std::time::Instant;
 
fn main() {
    let paths: Vec<&str> = (0..10_000)
        .map(|i| format!("src/file{}.rs", i))
        .collect();
    
    // Approach 1: Compile pattern each time (slow)
    let start = Instant::now();
    for path in &paths {
        let pattern = Pattern::new("src/*.rs").unwrap();  // Compiles EVERY iteration!
        if pattern.matches(path) {
            // process
        }
    }
    let repeated_compile = start.elapsed();
    
    // Approach 2: Compile once, reuse (fast)
    let start = Instant::now();
    let pattern = Pattern::new("src/*.rs").unwrap();  // Compile ONCE
    for path in &paths {
        if pattern.matches(path) {
            // process
        }
    }
    let single_compile = start.elapsed();
    
    println!("Repeated compile: {:?}", repeated_compile);
    println!("Single compile: {:?}", single_compile);
    println!("Speedup: {:.0}x", repeated_compile.as_nanos() as f64 / single_compile.as_nanos() as f64);
}

Compiling on every iteration is wasteful; compile once and reuse.

Pattern Caching Pattern

use glob::Pattern;
use std::collections::HashMap;
 
struct GlobMatcher {
    patterns: HashMap<String, Pattern>,
}
 
impl GlobMatcher {
    fn new() -> Self {
        Self {
            patterns: HashMap::new(),
        }
    }
    
    fn matches(&mut self, pattern_str: &str, path: &str) -> bool {
        // Get or create the compiled pattern
        let pattern = self.patterns
            .entry(pattern_str.to_string())
            .or_insert_with(|| Pattern::new(pattern_str).unwrap());
        
        pattern.matches(path)
    }
    
    fn matches_path(&self, pattern_str: &str, path: &std::path::Path) -> bool {
        self.patterns
            .get(pattern_str)
            .map(|p| p.matches_path(path))
            .unwrap_or(false)
    }
}
 
fn main() {
    let mut matcher = GlobMatcher::new();
    
    // First use compiles the pattern
    println!("Match 1: {}", matcher.matches("*.rs", "main.rs"));
    
    // Subsequent uses reuse the compiled pattern
    println!("Match 2: {}", matcher.matches("*.rs", "lib.rs"));
    println!("Match 3: {}", matcher.matches("*.rs", "test.txt"));
    
    // Different pattern gets compiled separately
    println!("Match 4: {}", matcher.matches("*.txt", "readme.txt"));
}

Caching compiled patterns avoids redundant parsing.

Glob Syntax and Complexity

use glob::Pattern;
 
fn main() {
    // Simple wildcards
    let simple = Pattern::new("*.rs").unwrap();
    
    // Character classes
    let char_class = Pattern::new("file[0-9].txt").unwrap();
    
    // Negation in character classes
    let negation = Pattern::new("file[!0-9].txt").unwrap();
    
    // Single character match
    let single = Pattern::new("file?.txt").unwrap();
    
    // Recursive glob
    let recursive = Pattern::new("src/**/*.rs").unwrap();
    
    // Each pattern has different parsing complexity
    // Character classes require more parsing than simple wildcards
    
    println!("All patterns compiled successfully");
}

Complex patterns with character classes require more parsing work.

MatchOptions for Control

use glob::{Pattern, MatchOptions};
 
fn main() {
    let pattern = Pattern::new("*.RS").unwrap();
    
    // Default: case-sensitive on Unix, case-insensitive on Windows
    println!("Default (case-sensitive on Unix): {}", pattern.matches("file.rs"));
    
    // Case-insensitive matching
    let case_insensitive = MatchOptions {
        case_sensitive: false,
        require_literal_separator: false,
        require_literal_leading_dot: false,
    };
    println!("Case-insensitive: {}", pattern.matches_with("file.rs", &case_insensitive));
    
    // Require literal separator (don't match '/' with '*')
    let literal_sep = MatchOptions {
        case_sensitive: true,
        require_literal_separator: true,
        require_literal_leading_dot: false,
    };
    
    let all_files = Pattern::new("*").unwrap();
    println!("Matches 'src/file.rs' with literal sep: {}", 
        all_files.matches_with("src/file.rs", &literal_sep));
    // With literal separator, '*' won't match '/'
}

MatchOptions provides fine-grained control over matching behavior.

Error Handling in Pattern Compilation

use glob::Pattern;
 
fn main() {
    // Valid pattern
    match Pattern::new("*.rs") {
        Ok(pattern) => println!("Compiled: {:?}", pattern.as_str()),
        Err(e) => println!("Error: {}", e),
    }
    
    // Invalid pattern (unmatched bracket)
    match Pattern::new("file[0-9.txt") {
        Ok(pattern) => println!("Compiled: {:?}", pattern.as_str()),
        Err(e) => println!("Error: {}", e),
    }
    
    // Invalid pattern (unclosed range)
    match Pattern::new("file[0-") {
        Ok(pattern) => println!("Compiled: {:?}", pattern.as_str()),
        Err(e) => println!("Error: {}", e),
    }
}

Pattern::new returns Result to handle invalid glob syntax.

Using with glob::glob Function

use glob::glob;
 
fn main() {
    // glob() compiles pattern internally each call
    for entry in glob("src/**/*.rs").unwrap() {
        match entry {
            Ok(path) => println!("Found: {:?}", path),
            Err(e) => println!("Error: {}", e),
        }
    }
    
    // For repeated calls with same pattern,
    // compile and use glob::Paths directly
}

glob::glob compiles the pattern internally; consider caching for repeated use.

Pattern from Configuration

use glob::Pattern;
use std::collections::HashMap;
 
struct FileFilter {
    // Cache of compiled patterns from config
    include_patterns: Vec<Pattern>,
    exclude_patterns: Vec<Pattern>,
}
 
impl FileFilter {
    fn from_config(include: &[&str], exclude: &[&str]) -> Result<Self, glob::PatternError> {
        let include_patterns = include
            .iter()
            .map(|p| Pattern::new(p))
            .collect::<Result<Vec<_>, _>>()?;
        
        let exclude_patterns = exclude
            .iter()
            .map(|p| Pattern::new(p))
            .collect::<Result<Vec<_>, _>>()?;
        
        Ok(Self {
            include_patterns,
            exclude_patterns,
        })
    }
    
    fn should_process(&self, path: &str) -> bool {
        // Check if matches any include pattern
        let included = self.include_patterns.iter().any(|p| p.matches(path));
        
        // Check if matches any exclude pattern
        let excluded = self.exclude_patterns.iter().any(|p| p.matches(path));
        
        included && !excluded
    }
}
 
fn main() {
    let filter = FileFilter::from_config(
        &["src/**/*.rs", "tests/**/*.rs"],
        &["**/target/**", "**/.git/**"],
    ).unwrap();
    
    println!("Process src/main.rs: {}", filter.should_process("src/main.rs"));
    println!("Process target/main.rs: {}", filter.should_process("target/main.rs"));
    println!("Process tests/test.rs: {}", filter.should_process("tests/test.rs"));
}

Compile patterns from configuration once at startup, then reuse.

Performance Comparison by Pattern Complexity

use glob::Pattern;
use std::time::Instant;
 
fn main() {
    let iterations = 100_000;
    
    // Simple pattern
    let simple_pattern = "*.rs";
    
    // Complex pattern with character classes
    let complex_pattern = "src/**/file[0-9][a-z].rs";
    
    // Test simple pattern compilation
    let start = Instant::now();
    for _ in 0..iterations {
        let _ = Pattern::new(simple_pattern).unwrap();
    }
    let simple_compile_time = start.elapsed();
    
    // Test complex pattern compilation
    let start = Instant::now();
    for _ in 0..iterations {
        let _ = Pattern::new(complex_pattern).unwrap();
    }
    let complex_compile_time = start.elapsed();
    
    println!("Simple pattern compile: {:?}", simple_compile_time);
    println!("Complex pattern compile: {:?}", complex_compile_time);
    println!("Complex is {:.1}x slower to compile",
        complex_compile_time.as_nanos() as f64 / simple_compile_time.as_nanos() as f64);
    
    // Compare matching time
    let simple = Pattern::new(simple_pattern).unwrap();
    let complex = Pattern::new(complex_pattern).unwrap();
    
    let start = Instant::now();
    for _ in 0..iterations {
        let _ = simple.matches("src/main.rs");
    }
    let simple_match_time = start.elapsed();
    
    let start = Instant::now();
    for _ in 0..iterations {
        let _ = complex.matches("src/dir/file1a.rs");
    }
    let complex_match_time = start.elapsed();
    
    println!("\nSimple pattern match: {:?}", simple_match_time);
    println!("Complex pattern match: {:?}", complex_match_time);
}

Complex patterns cost more to compile and match than simple ones.

Pattern Validation at Startup

use glob::Pattern;
 
fn validate_patterns(patterns: &[&str]) -> Result<Vec<Pattern>, String> {
    patterns
        .iter()
        .map(|p| {
            Pattern::new(p).map_err(|e| {
                format!("Invalid pattern '{}': {}", p, e)
            })
        })
        .collect()
}
 
fn main() {
    // Validate all patterns at startup
    let patterns = ["src/**/*.rs", "tests/**/*.rs", "invalid[pattern"];
    
    match validate_patterns(&patterns) {
        Ok(compiled) => {
            println!("All patterns valid, {} compiled", compiled.len());
        }
        Err(e) => {
            println!("Pattern validation failed: {}", e);
        }
    }
    
    // Alternative: compile on demand with error handling
    let on_demand = |pattern_str: &str| -> Result<bool, glob::PatternError> {
        let pattern = Pattern::new(pattern_str)?;
        Ok(pattern.matches("src/main.rs"))
    };
}

Validate and compile patterns at startup to catch errors early.

Comparing with std::path Matching

use glob::Pattern;
use std::path::Path;
 
fn main() {
    // glob::Pattern - matches against glob syntax
    let glob_pattern = Pattern::new("src/**/*.rs").unwrap();
    
    // std::path - matches against path components
    let path = Path::new("src/main.rs");
    
    // glob pattern matching
    println!("Glob matches: {}", glob_pattern.matches_path(path));
    
    // Manual path matching (more limited)
    println!("Extension: {:?}", path.extension());
    println!("Parent: {:?}", path.parent());
    
    // For complex patterns, glob is more expressive
    // For simple checks, Path methods may be faster
}

Pattern provides expressive matching; Path methods are simpler for basic checks.

Pattern Reuse in Multithreaded Context

use glob::Pattern;
use std::sync::Arc;
use std::thread;
 
fn main() {
    // Pattern is Send + Sync, can be shared across threads
    let pattern = Arc::new(Pattern::new("src/**/*.rs").unwrap());
    
    let handles: Vec<_> = (0..4)
        .map(|_| {
            let pattern = Arc::clone(&pattern);
            thread::spawn(move || {
                // Each thread uses the same compiled pattern
                let test_paths = vec
!["src/main.rs", "tests/test.rs", "src/lib/mod.rs"];
                for path in test_paths {
                    if pattern.matches(path) {
                        println!("Thread matched: {}", path);
                    }
                }
            })
        })
        .collect();
    
    for handle in handles {
        handle.join().unwrap();
    }
}

Pattern can be shared across threads without recompilation.

Pattern as Part of a Struct

use glob::Pattern;
 
#[derive(Debug)]
struct Rule {
    pattern: Pattern,
    action: String,
}
 
impl Rule {
    fn new(pattern_str: &str, action: &str) -> Result<Self, glob::PatternError> {
        Ok(Self {
            pattern: Pattern::new(pattern_str)?,
            action: action.to_string(),
        })
    }
    
    fn matches(&self, path: &str) -> Option<&str> {
        if self.pattern.matches(path) {
            Some(&self.action)
        } else {
            None
        }
    }
}
 
fn main() {
    let rules = [
        Rule::new("*.rs", "compile").unwrap(),
        Rule::new("*.md", "copy").unwrap(),
        Rule::new("*.toml", "process").unwrap(),
    ];
    
    let files = ["main.rs", "readme.md", "config.toml", "test.txt"];
    
    for file in files {
        for rule in &rules {
            if let Some(action) = rule.matches(file) {
                println!("{}: {} (rule: {})", file, action, rule.pattern.as_str());
            }
        }
    }
}

Storing Pattern in structs enables efficient rule-based matching.

Synthesis

Compilation cost comparison:

| Pattern Type | Compilation Cost | Match Cost | |--------------|------------------|------------| | Simple (*) | Low | Low | | Character class ([a-z]) | Medium | Medium | | Recursive (**) | Low | Higher | | Complex (src/**/[A-Z]*.rs) | Higher | Higher |

When to compile once:

| Scenario | Recommendation | |----------|----------------| | Matching many paths | Compile once, reuse | | Pattern from config | Compile at startup | | Long-running service | Cache patterns | | Multiple files, same pattern | Pattern struct | | Concurrent matching | Arc<Pattern> |

When inline compilation is acceptable:

| Scenario | Reasoning | |----------|-----------| | One-off match | Overhead negligible | | Dynamic patterns | Must compile anyway | | CLI with one match | Simplicity over optimization | | Prototyping | Optimize later |

Key insight: The glob::Pattern::new function performs real work—parsing the glob syntax, validating bracket expressions, and building an internal representation that can efficiently match paths. This work is done once per new call, regardless of how many subsequent matches occur. For a script that matches a single path, compiling inline is fine. But for a file watcher that matches thousands of files against the same patterns, or a web server that routes requests by path patterns, the amortized cost of repeated compilation adds up. The Pattern struct is designed to be created once and used many times: it's Send + Sync for sharing across threads, Clone for cheap copying, and its matching methods are &self borrows that work on the compiled representation. The trade-off isn't just performance—it's also about error handling. If you compile patterns at startup, validation errors surface immediately with clear context. If you compile on-demand, a typo in a configuration pattern only reveals itself when that pattern is first needed, possibly in production during an edge case. Compile early, validate early, and let the compiled Pattern do the heavy lifting of matching without re-parsing.

Loading page…

What are the trade-offs between glob::Pattern::new and compiled glob::Pattern for repeated matching?

Basic Pattern Creation

Pattern Matching Methods

The Cost of Repeated Compilation

Pattern Caching Pattern

Glob Syntax and Complexity

MatchOptions for Control

Error Handling in Pattern Compilation

Using with glob::glob Function

Pattern from Configuration

Performance Comparison by Pattern Complexity

Pattern Validation at Startup

Comparing with std::path Matching

Pattern Reuse in Multithreaded Context

Pattern as Part of a Struct

Synthesis

What are the trade-offs between glob::Pattern::new and compiled glob::Pattern for repeated matching?

Basic Pattern Creation

Pattern Matching Methods

The Cost of Repeated Compilation

Pattern Caching Pattern

Glob Syntax and Complexity

MatchOptions for Control

Error Handling in Pattern Compilation

Using with glob::glob Function

Pattern from Configuration

Performance Comparison by Pattern Complexity

Pattern Validation at Startup

Comparing with std::path Matching

Pattern Reuse in Multithreaded Context

Pattern as Part of a Struct

Synthesis

What are the trade-offs between `glob::Pattern::new` and compiled `glob::Pattern` for repeated matching?

What are the trade-offs between `glob::Pattern::new` and compiled `glob::Pattern` for repeated matching?