treezip 1.0.0 → 1.1.1

package/README.md

@@ -108,7 +108,34 @@ my-project/
 
 ### Default ignores
 
-`node_modules`, `.git`, `.DS_Store`, `__pycache__`, `.pnpm`
+treezip hardcodes an aggressive noise filter tuned for LLM context — build
+artifacts, caches, and OS junk never reach the output:
+
+- **VCS**: `.git` (dir and worktree pointer file), `.hg`, `.svn`, `.jj`, …
+- **OS**: `.DS_Store`, `Thumbs.db`, `desktop.ini`, `._*` AppleDouble files, …
+- **JS/TS**: `node_modules`, `.next`, `.nuxt`, `.turbo`, `.vite`, `.svelte-kit`,
+  `.astro`, `coverage`, `.wrangler`, `.vercel`, `*.tsbuildinfo`, …
+- **Python**: `__pycache__`, `.venv`/`venv`, `.pytest_cache`, `.mypy_cache`,
+  `.ruff_cache`, `*.egg-info`, `*.pyc`, …
+- **Swift/Xcode**: `DerivedData`, `.build`, `.swiftpm`, `xcuserdata`,
+  `Index.noindex` and friends, …
+- **Others**: `.gradle`, `zig-out`, `_build` (Elixir), `.dart_tool`,
+  `CMakeFiles`, `cmake-build-*`, `.terraform`, `.idea`, `.vs`, …
+
+Three smarter rules on top of the basename list:
+
+1. **Content markers** — a directory containing `pyvenv.cfg` is a venv, one
+   containing `Index.noindex` is an Xcode DerivedData root; both are skipped
+   *whatever they're named* (catches custom `-derivedDataPath` dirs).
+2. **Sibling confirmation** — generic names are only skipped when the
+   ecosystem is proven: `target/` next to `Cargo.toml`/`pom.xml`, `build/`
+   next to `build.gradle`, `bin`/`obj` next to a `.csproj`. A docs folder
+   named `build/` stays visible.
+3. **Scoped rules** — `.claude/worktrees` is hidden but `.claude/skills`
+   stays; `.yarn/cache` is hidden but `.yarn/patches` stays.
+
+Meaningful files are never hidden: lockfiles (`package-lock.json`,
+`Package.resolved`, `uv.lock`, …), `.env`, `.envrc`, `.vscode`, `.github`.
 
 ## Output header
 

package/dist/ignore.d.ts

@@ -0,0 +1,10 @@
+/** Directory basenames that are always noise. */
+export declare const NOISE_DIRS: Set<string>;
+/** File basenames that are always noise. */
+export declare const NOISE_FILES: Set<string>;
+/** True if a directory entry should be skipped, judged by name + context. */
+export declare function isNoiseDir(name: string, parentName: string, siblings: string[]): boolean;
+/** True if a file entry should be skipped. */
+export declare function isNoiseFile(name: string, parentName: string): boolean;
+/** True if a directory's own entry list marks the whole dir as noise. */
+export declare function hasNoiseMarker(entries: string[]): boolean;

package/dist/ignore.js

@@ -0,0 +1,205 @@
+// Hardcoded noise filtering — build artifacts, caches, and OS junk that
+// should never appear in an LLM-context tree. Sources: github/gitignore
+// templates, vendor docs, and the default-ignore lists shipped by
+// repomix/gitingest (the closest prior art).
+//
+// Design rules:
+//   1. Only distinctive basenames are ignored unconditionally. Generic
+//      names (dist, build, out, bin, obj, vendor, Pods) are kept visible
+//      unless corroborated by a sibling marker (e.g. target + Cargo.toml).
+//   2. Content markers catch arbitrarily-named build roots: a dir that
+//      directly contains pyvenv.cfg is a venv; one containing
+//      Index.noindex / ModuleCache.noindex is an Xcode DerivedData root
+//      (custom -derivedDataPath dirs can be named anything).
+//   3. Lockfiles (package-lock.json, Package.resolved, uv.lock, ...) and
+//      meaningful dotfiles (.env, .envrc, .vscode) are never hidden.
+/** Directory basenames that are always noise. */
+export const NOISE_DIRS = new Set([
+    // VCS internals
+    ".git",
+    ".hg",
+    ".svn",
+    ".bzr",
+    ".jj",
+    ".sapling",
+    // OS junk
+    ".Spotlight-V100",
+    ".Trashes",
+    ".fseventsd",
+    ".TemporaryItems",
+    ".DocumentRevisions-V100",
+    "$RECYCLE.BIN",
+    "__MACOSX",
+    // JavaScript / TypeScript
+    "node_modules",
+    ".pnpm",
+    ".pnpm-store",
+    ".next",
+    ".nuxt",
+    ".output",
+    ".turbo",
+    ".parcel-cache",
+    ".cache",
+    ".vite",
+    ".svelte-kit",
+    ".astro",
+    ".angular",
+    ".docusaurus",
+    ".expo",
+    ".expo-shared",
+    ".wrangler",
+    ".vercel",
+    ".netlify",
+    ".serverless",
+    ".rpt2_cache",
+    ".sass-cache",
+    "coverage",
+    ".nyc_output",
+    "storybook-static",
+    // Python
+    "__pycache__",
+    ".venv",
+    "venv",
+    ".tox",
+    ".nox",
+    ".pytest_cache",
+    ".mypy_cache",
+    ".ruff_cache",
+    ".hypothesis",
+    ".ipynb_checkpoints",
+    ".eggs",
+    "site-packages",
+    "htmlcov",
+    "__pypackages__",
+    ".pytype",
+    ".pyre",
+    "cython_debug",
+    // Swift / Xcode
+    "DerivedData",
+    ".build",
+    ".swiftpm",
+    "xcuserdata",
+    "Carthage",
+    "SourcePackages",
+    "Index.noindex",
+    "ModuleCache.noindex",
+    "Intermediates.noindex",
+    "CompilationCache.noindex",
+    "SDKStatCaches.noindex",
+    // JVM
+    ".gradle",
+    // Zig
+    ".zig-cache",
+    "zig-cache",
+    "zig-out",
+    // Elixir
+    "_build",
+    ".elixir_ls",
+    // Dart / Flutter
+    ".dart_tool",
+    // C / CMake
+    "CMakeFiles",
+    "_deps",
+    // .NET / IDE caches (.vscode/.idea configs are often committed — .vs never is)
+    ".vs",
+    ".idea",
+    // Ruby
+    ".bundle",
+    ".yardoc",
+    // Infra
+    ".terraform",
+    ".terragrunt-cache",
+    ".vagrant",
+    ".direnv",
+]);
+/** File basenames that are always noise. */
+export const NOISE_FILES = new Set([
+    ".DS_Store",
+    "Thumbs.db",
+    "desktop.ini",
+    ".directory",
+    ".git", // worktree/submodule pointer file
+    ".eslintcache",
+    ".stylelintcache",
+    ".node_repl_history",
+    ".yarn-integrity",
+    ".coverage",
+    "coverage.xml",
+    ".dmypy.json",
+    "dmypy.json",
+    "nosetests.xml",
+    "Package.pins",
+]);
+const NOISE_FILE_PREFIXES = [
+    "._", // AppleDouble resource forks
+    "npm-debug.log",
+    "yarn-debug.log",
+    "yarn-error.log",
+    "lerna-debug.log",
+];
+const NOISE_FILE_SUFFIXES = [".pyc", ".pyo", ".tsbuildinfo"];
+const NOISE_DIR_PREFIXES = ["cmake-build-"];
+const NOISE_DIR_SUFFIXES = [".egg-info"];
+/**
+ * Generic dir names that are only noise when a sibling file proves the
+ * ecosystem (predicate runs against each sibling basename).
+ */
+const SIBLING_CONFIRMED_DIRS = {
+    target: (s) => s === "Cargo.toml" || s === "pom.xml",
+    build: (s) => s === "build.gradle" ||
+        s === "build.gradle.kts" ||
+        s === "settings.gradle" ||
+        s === "settings.gradle.kts",
+    deps: (s) => s === "mix.exs",
+    bin: (s) => /\.(cs|fs|vb)proj$|\.sln$/.test(s),
+    obj: (s) => /\.(cs|fs|vb)proj$|\.sln$/.test(s),
+};
+/** Child names that are noise only inside a specific parent dir. */
+const SCOPED_NOISE = {
+    ".claude": new Set(["worktrees"]),
+    ".yarn": new Set(["cache", "unplugged", "install-state.gz", "build-state.yml"]),
+};
+/**
+ * If a directory directly contains any of these entries, the entire
+ * directory is noise regardless of its name.
+ */
+const CONTENT_MARKERS = new Set([
+    "pyvenv.cfg", // Python venv (catches env/, myenv/, any name)
+    "Index.noindex", // Xcode DerivedData root (default or custom -derivedDataPath)
+    "ModuleCache.noindex",
+    "SDKStatCaches.noindex",
+    "CompilationCache.noindex",
+]);
+/** True if a directory entry should be skipped, judged by name + context. */
+export function isNoiseDir(name, parentName, siblings) {
+    if (NOISE_DIRS.has(name))
+        return true;
+    if (NOISE_DIR_PREFIXES.some((p) => name.startsWith(p)))
+        return true;
+    if (NOISE_DIR_SUFFIXES.some((s) => name.endsWith(s)))
+        return true;
+    const scoped = SCOPED_NOISE[parentName];
+    if (scoped?.has(name))
+        return true;
+    const confirm = SIBLING_CONFIRMED_DIRS[name];
+    if (confirm && siblings.some(confirm))
+        return true;
+    return false;
+}
+/** True if a file entry should be skipped. */
+export function isNoiseFile(name, parentName) {
+    if (NOISE_FILES.has(name))
+        return true;
+    if (NOISE_FILE_PREFIXES.some((p) => name.startsWith(p)))
+        return true;
+    if (NOISE_FILE_SUFFIXES.some((s) => name.endsWith(s)))
+        return true;
+    const scoped = SCOPED_NOISE[parentName];
+    if (scoped?.has(name))
+        return true;
+    return false;
+}
+/** True if a directory's own entry list marks the whole dir as noise. */
+export function hasNoiseMarker(entries) {
+    return entries.some((e) => CONTENT_MARKERS.has(e));
+}

package/dist/index.d.ts

@@ -3,3 +3,4 @@ export type { TreeNode } from "./scanner.js";
 export { compress } from "./compress.js";
 export type { CompressedItem } from "./compress.js";
 export { render } from "./renderer.js";
+export { NOISE_DIRS, NOISE_FILES, isNoiseDir, isNoiseFile, hasNoiseMarker } from "./ignore.js";

package/dist/index.js

@@ -1,3 +1,4 @@
 export { scan, countNodes } from "./scanner.js";
 export { compress } from "./compress.js";
 export { render } from "./renderer.js";
+export { NOISE_DIRS, NOISE_FILES, isNoiseDir, isNoiseFile, hasNoiseMarker } from "./ignore.js";

package/dist/scanner.d.ts

@@ -3,7 +3,7 @@ export interface TreeNode {
     isDir: boolean;
     children: TreeNode[];
 }
-export declare function scan(rootPath: string, ignores?: Set<string>): TreeNode;
+export declare function scan(rootPath: string): TreeNode;
 /** Count total files and directories in a TreeNode (excludes root) */
 export declare function countNodes(node: TreeNode): {
     dirs: number;

package/dist/scanner.js

@@ -1,25 +1,23 @@
-import { readdirSync, statSync, lstatSync } from "node:fs";
-import { join } from "node:path";
-const DEFAULT_IGNORES = new Set([
-    "node_modules",
-    ".git",
-    ".DS_Store",
-    ".Spotlight-V100",
-    ".Trashes",
-    "__pycache__",
-    ".pnpm",
-]);
-export function scan(rootPath, ignores = DEFAULT_IGNORES) {
+import { readdirSync, lstatSync, statSync } from "node:fs";
+import { join, basename } from "node:path";
+import { isNoiseDir, isNoiseFile, hasNoiseMarker } from "./ignore.js";
+export function scan(rootPath) {
     const stat = statSync(rootPath);
     const name = rootPath.split("/").filter(Boolean).pop() ?? rootPath;
     if (!stat.isDirectory()) {
         return { name, isDir: false, children: [] };
     }
-    const entries = readdirSync(rootPath).filter((e) => !ignores.has(e)).sort();
+    // Never bail on the root itself, even if it looks like a build dir —
+    // the user explicitly asked for it.
+    return scanDir(rootPath, name);
+}
+function scanDir(dirPath, name) {
+    const entries = readdirSync(dirPath).sort();
+    const parentName = basename(dirPath);
     const files = [];
     const dirs = [];
     for (const entry of entries) {
-        const fullPath = join(rootPath, entry);
+        const fullPath = join(dirPath, entry);
         let entryStat;
         try {
             entryStat = lstatSync(fullPath);
@@ -31,9 +29,24 @@ export function scan(rootPath, ignores = DEFAULT_IGNORES) {
         if (entryStat.isSymbolicLink())
             continue;
         if (entryStat.isDirectory()) {
-            dirs.push(scan(fullPath, ignores));
+            if (isNoiseDir(entry, parentName, entries))
+                continue;
+            // Content-marker check: a dir whose entries reveal it as a venv or
+            // DerivedData root is noise regardless of what it's named.
+            let childEntries;
+            try {
+                childEntries = readdirSync(fullPath);
+            }
+            catch {
+                continue;
+            }
+            if (hasNoiseMarker(childEntries))
+                continue;
+            dirs.push(scanDir(fullPath, entry));
         }
         else if (entryStat.isFile()) {
+            if (isNoiseFile(entry, parentName))
+                continue;
             files.push({ name: entry, isDir: false, children: [] });
         }
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "treezip",
-  "version": "1.0.0",
+  "version": "1.1.1",
   "description": "Compressed tree output for AI-optimized codebase summaries",
   "type": "module",
   "bin": {