Performance improve

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "codem8"
-version = "0.1.0"
+version = "0.2.0"
 edition = "2021"
 license = "MIT"
 description = "A deterministic source code analysis CLI for duplicate code reports."
@@ -9,5 +9,7 @@ keywords = ["cli", "duplicate-detection", "source-code", "analysis"]
 categories = ["command-line-utilities", "development-tools"]
 
 [dependencies]
+ignore = "0.4"
+rayon = "1"
 regex = "1"
 xxhash-rust = { version = "0.8", features = ["xxh3"] }

README.md

@@ -40,13 +40,13 @@ cargo run -- --report-duplicate
 
 ## Usage
 
-Analyze TypeScript files from the current directory:
+Analyze supported source files from the current directory:
 
 ```bash
 codem8 --report-duplicate
 ```
 
-Analyze multiple extensions:
+Restrict analysis to specific extensions:
 
 ```bash
 codem8 --report-duplicate -file-extension=ts,tsx,js,jsx
@@ -65,16 +65,17 @@ base branch:
 codem8 --report-duplicate -git-branch
 ```
 
-Include duplicate block metrics:
+Include duplicate block metrics and timing information:
 
 ```bash
 codem8 --report-duplicate -verbose
 ```
 
 ## Duplicate Report
 
-By default, CodeM8 analyzes `.ts` files. Recursive discovery skips common
-irrelevant directories such as `.git`, `node_modules`, `target`, `dist`,
+By default, CodeM8 analyzes all registered source file extensions. Recursive
+discovery respects Git ignore rules, works outside Git repositories, and skips
+common irrelevant directories such as `.git`, `node_modules`, `target`, `dist`,
 `build`, `coverage`, `.next`, `.nuxt`, `.svelte-kit`, `.idea`, and `.vscode`.
 Symbolic links are not followed.
 
@@ -99,55 +100,22 @@ Reports are sorted deterministically by descending weight, then by line count,
 character count, first location, and normalized block text.
 
 By default, each duplicate block prints the duplicated code before its
-locations. Use `-verbose` to also show weight, line count, and occurrence
-count. Character counts are used internally for scoring and sorting, but are
-not printed.
-
-## Language Heuristics
-
-CodeM8 includes a hard-coded registry of block-only line patterns for common
-languages and markup formats:
-
-- TypeScript / JavaScript
-- Rust
-- C / C++ / Objective-C
-- C#
-- Java / Kotlin / Scala
-- Go
-- Python
-- Ruby
-- PHP
-- Swift
-- Shell
-- PowerShell
-- HTML / XML
-- CSS / SCSS / Sass / Less
-- SQL
-- YAML / JSON / TOML
-
-Block-only lines, such as braces or closing tags, cannot start a duplicate by
-themselves. They can still be included inside a larger duplicated block when
-surrounding comparison lines match.
+locations. Use `-verbose` to also show weight, line count, occurrence count, and
+timings for discovery, file processing, and duplicate detection. Character
+counts are used internally for scoring and sorting, but are not printed.
 
 ## Development
 
 Run the full local verification set:
 
 ```bash
+cargo test
 cargo fmt --all -- --check
-cargo clippy --workspace --all-targets --all-features -- -D warnings -W clippy::too_many_lines -W clippy::too_many_arguments -W clippy::type_complexity -W clippy::excessive_nesting -W clippy::cognitive_complexity
-rtk cargo build --locked --all-targets
-cargo test --all-targets
+cargo clippy --workspace --all-targets --all-features -- -D warnings -W clippy::too_many_lines -W clippy::too_many_arguments -W clippy::type_complexity -W clippy::excessive_nesting -W clippy::cognitive_complexity -W clippy::pedantic -W clippy::nursery -W clippy::cargo
+cargo build --locked --all-targets
 ```
 
 The repository includes GitHub Actions workflows for Rust CI and a CodeRabbit
 review gate. CI verifies formatting, build success, and tests on pushes and pull
 requests. The CodeRabbit gate runs when CodeRabbit submits or edits a pull
 request review and fails if CodeRabbit requests changes on the current PR head.
-
-## Dependency Policy
-
-CodeM8 avoids external packages for functionality that is simple to implement
-and maintain directly. The first implementation uses one runtime dependency,
-`xxhash-rust`, for the required XXH3 128-bit hash implementation. The crate is
-widely used and permissively licensed under MIT or Apache-2.0.