Fail closed on git status and walk deep forms without recursion.

Status records are NUL-terminated, and a git failure returns git's status. .test.cts and .spec.cts are tests. Help lists every skipped directory.

src/dryer/astnorm.py

@@ -152,33 +152,7 @@ def _text(node) -> str:
     return node_text(_DATA, node)
 
 
-def _normalize(node, head: bool = False):
-    if node.type in _SKIP:
-        return None
-    if node.type in _OPERATORS:
-        return [K("symbol"), node.type]
-    if not node.is_named:
-        return None
-    if node.type in _IDENTIFIERS:
-        if head:
-            return [K("symbol"), _text(node)]
-        return K("symbol")
-    if _is_literal(node):
-        return K("literal")
-    if node.type == "parenthesized_expression":
-        named = node.named_children
-        if len(named) == 1:
-            return _normalize(named[0], head)
-    if node.type in _CALLS:
-        return _normalize_call(node)
-    if node.type in _ATTRIBUTES:
-        return _normalize_attr(node, head)
-    children = []
-    for child in node.children:
-        norm = _normalize(child, False)
-        if norm is not None:
-            children.append(norm)
-    return [K(node.type), *children]
+_PENDING = object()
 
 
 def _is_literal(node) -> bool:
@@ -204,54 +178,88 @@ def _without_type_args(nodes):
     return rest, type_args
 
 
-def _extend_normalized(out, nodes) -> None:
-    for node in nodes:
-        norm = _normalize(node, False)
-        if norm is not None:
-            out.append(norm)
+def _try_leaf(node, mode: str, head: bool):
+    if mode == "callee":
+        if node.type in _IDENTIFIERS:
+            return [K("symbol"), _text(node)]
+        return _PENDING
+    if mode == "path":
+        return _PENDING
+    if node.type in _SKIP:
+        return None
+    if node.type in _OPERATORS:
+        return [K("symbol"), node.type]
+    if not node.is_named:
+        return None
+    if node.type in _IDENTIFIERS:
+        if head:
+            return [K("symbol"), _text(node)]
+        return K("symbol")
+    if _is_literal(node):
+        return K("literal")
+    return _PENDING
 
 
-def _normalize_call(node):
+def _call_jobs(node):
     incoming, args = _named_before_args(node)
     rest, type_args = _without_type_args(incoming)
-    out = [K(node.type)]
+    jobs = []
     if rest:
         *receivers, callee = rest
-        _extend_normalized(out, receivers)
-        out.append(_normalize_callee(callee))
-    _extend_normalized(out, type_args)
+        jobs.extend((child, "norm", False) for child in receivers)
+        jobs.append((callee, "callee", False))
+    jobs.extend((child, "norm", False) for child in type_args)
     if args is not None:
-        _extend_normalized(out, [args])
-    return out
+        jobs.append((args, "norm", False))
+    return jobs
 
 
-def _normalize_callee(node):
-    if node.type in _IDENTIFIERS:
-        return [K("symbol"), _text(node)]
-    if node.type in _ATTRIBUTES:
-        return _normalize_attr(node, True)
-    if node.type in _SCOPED or node.type == "generic_type":
-        return _normalize_path(node)
-    return _normalize(node, True)
+def _attr_jobs(node, head: bool):
+    named = list(node.named_children)
+    if not named:
+        return []
+    *objects, name = named
+    jobs = [(obj, "norm", False) for obj in objects]
+    if name.type in _IDENTIFIERS:
+        return jobs
+    if head:
+        jobs.append((name, "callee", False))
+    else:
+        jobs.append((name, "norm", False))
+    return jobs
 
 
-def _normalize_path(node):
-    """A callee path keeps each name in the path. Type arguments stay structural."""
-
-    parts = [K(node.type)]
+def _path_jobs(node):
+    jobs = []
     for child in node.named_children:
         if child.type in _IDENTIFIERS:
-            parts.append([K("symbol"), _text(child)])
-        elif child.type in _SCOPED or child.type == "generic_type":
-            parts.append(_normalize_path(child))
+            continue
+        if child.type in _SCOPED or child.type == "generic_type":
+            jobs.append((child, "path", False))
         else:
-            norm = _normalize(child, False)
-            if norm is not None:
-                parts.append(norm)
-    return parts
+            jobs.append((child, "norm", False))
+    return jobs
 
 
-def _normalized_name(name, head: bool):
+def _jobs(node, mode: str, head: bool):
+    if mode == "callee":
+        if node.type in _ATTRIBUTES:
+            return _attr_jobs(node, True)
+        if node.type in _SCOPED or node.type == "generic_type":
+            return _path_jobs(node)
+        return [(node, "norm", True)]
+    if mode == "path":
+        return _path_jobs(node)
+    if node.type == "parenthesized_expression" and len(node.named_children) == 1:
+        return [(node.named_children[0], "norm", head)]
+    if node.type in _CALLS:
+        return _call_jobs(node)
+    if node.type in _ATTRIBUTES:
+        return _attr_jobs(node, head)
+    return [(child, "norm", False) for child in node.children]
+
+
+def _name_value(name, head: bool, cache):
     """Keep a member's spelling when it is the thing being called."""
 
     if name.type in _IDENTIFIERS:
@@ -259,21 +267,105 @@ def _normalized_name(name, head: bool):
             return [K("symbol"), _text(name)]
         return K("symbol")
     if head:
-        return _normalize_callee(name)
-    return _normalize(name, False)
+        return cache[(id(name), "callee", False)]
+    return cache[(id(name), "norm", False)]
 
 
-def _normalize_attr(node, head: bool):
+def _assemble_call(node, cache):
+    incoming, args = _named_before_args(node)
+    rest, type_args = _without_type_args(incoming)
+    out = [K(node.type)]
+    if rest:
+        *receivers, callee = rest
+        for child in receivers:
+            norm = cache[(id(child), "norm", False)]
+            if norm is not None:
+                out.append(norm)
+        out.append(cache[(id(callee), "callee", False)])
+    for child in type_args:
+        norm = cache[(id(child), "norm", False)]
+        if norm is not None:
+            out.append(norm)
+    if args is not None:
+        norm = cache[(id(args), "norm", False)]
+        if norm is not None:
+            out.append(norm)
+    return out
+
+
+def _assemble_attr(node, head: bool, cache):
     named = list(node.named_children)
     if not named:
         return [K(node.type)]
     *objects, name = named
     parts = [K(node.type)]
     for obj in objects:
-        norm = _normalize(obj, False)
+        norm = cache[(id(obj), "norm", False)]
         if norm is not None:
             parts.append(norm)
-    nested = _normalized_name(name, head)
+    nested = _name_value(name, head, cache)
     if nested is not None:
         parts.append(nested)
     return parts
+
+
+def _assemble_path(node, cache):
+    """A callee path keeps each name in the path. Type arguments stay structural."""
+
+    parts = [K(node.type)]
+    for child in node.named_children:
+        if child.type in _IDENTIFIERS:
+            parts.append([K("symbol"), _text(child)])
+        elif child.type in _SCOPED or child.type == "generic_type":
+            parts.append(cache[(id(child), "path", False)])
+        else:
+            norm = cache[(id(child), "norm", False)]
+            if norm is not None:
+                parts.append(norm)
+    return parts
+
+
+def _assemble(node, mode: str, head: bool, cache):
+    if mode == "callee":
+        if node.type in _ATTRIBUTES:
+            return _assemble_attr(node, True, cache)
+        if node.type in _SCOPED or node.type == "generic_type":
+            return _assemble_path(node, cache)
+        return cache[(id(node), "norm", True)]
+    if mode == "path":
+        return _assemble_path(node, cache)
+    if node.type == "parenthesized_expression" and len(node.named_children) == 1:
+        child = node.named_children[0]
+        return cache[(id(child), "norm", head)]
+    if node.type in _CALLS:
+        return _assemble_call(node, cache)
+    if node.type in _ATTRIBUTES:
+        return _assemble_attr(node, head, cache)
+    children = []
+    for child in node.children:
+        norm = cache[(id(child), "norm", False)]
+        if norm is not None:
+            children.append(norm)
+    return [K(node.type), *children]
+
+
+def _normalize(node, head: bool = False):
+    # Post-order. A 1,200-deep expression must not raise RecursionError.
+    cache = {}
+    stack = [(node, "norm", head, False)]
+    while stack:
+        current, mode, is_head, expanded = stack.pop()
+        key = (id(current), mode, is_head)
+        if not expanded:
+            if key in cache:
+                continue
+            leaf = _try_leaf(current, mode, is_head)
+            if leaf is not _PENDING:
+                cache[key] = leaf
+                continue
+            stack.append((current, mode, is_head, True))
+            for child, child_mode, child_head in reversed(_jobs(current, mode, is_head)):
+                stack.append((child, child_mode, child_head, False))
+            continue
+        cache[key] = _assemble(current, mode, is_head, cache)
+    return cache[(id(node), "norm", head)]

src/dryer/cli.py

@@ -2,16 +2,23 @@
 
 from __future__ import annotations
 
+import os
 import subprocess
 import sys
 from dataclasses import dataclass, field
 from pathlib import Path
 
-from dryer.discover import is_test_file, iter_source_files, language_of
+from dryer.discover import SKIP_DIRS, TEST_DIRS, is_test_file, iter_source_files, language_of
 from dryer.report import format_text, render_edn, write_metrics
 from dryer.scan import find_duplicates, scan_files
 
-HELP = """\
+
+def _skipped_directory_text() -> str:
+    ordered = sorted(set(SKIP_DIRS) | set(TEST_DIRS))
+    return ", ".join(ordered[:-1]) + ", and " + ordered[-1]
+
+
+HELP = f"""\
 Usage: dryer [options] [path-or-filter ...]
 
 Find candidate duplicate code in Clojure, Java, Go, TypeScript, Rust, and
@@ -48,8 +55,8 @@ Arguments:
                     path contains this text are compared.
 
 With no paths, source files under the project root are compared. Directories
-named test, tests, spec, specs, vendor, node_modules, and target are skipped,
-as are *_test.go, *.spec.ts, test_*.py, and *_test.py files.
+named {_skipped_directory_text()} are skipped, as are *_test.go, *.test.cts,
+*.spec.cts, *.spec.ts, test_*.py, and *_test.py files.
 
 Clojure compares every top-level form except ns. The other languages compare
 functions and methods. A form is only compared with forms in the same language.
@@ -159,25 +166,55 @@ def parse_args(argv: list[str] | None = None) -> Options:
     return options
 
 
-def _git_status(root: Path):
-    args = ["git", "status", "--porcelain"]
-    return subprocess.run(args, cwd=root, check=False, capture_output=True, text=True)
+class GitStatusError(Exception):
+    """git status could not be read. `code` is git's own status."""
+
+    def __init__(self, code: int, message: str):
+        super().__init__(message)
+        self.code = code
+        self.message = message
+
+
+def _git(root: Path, *args: str) -> bytes:
+    result = subprocess.run(
+        ["git", "-C", os.fspath(root), *args],
+        stdout=subprocess.PIPE,
+        stderr=subprocess.PIPE,
+        check=False,
+    )
+    if result.returncode != 0:
+        message = os.fsdecode(result.stderr).strip() or "git status failed"
+        raise GitStatusError(result.returncode, message)
+    return result.stdout
 
 
 def _changed_files(root: Path) -> list[Path]:
-    result = _git_status(root)
-    if result.returncode != 0:
-        print(result.stderr.strip() or "git status failed", file=sys.stderr)
-        return []
+    """Added and modified files under `root`, as paths from the repo root.
+
+    `-z` keeps spaces and non-ASCII names intact. A missing path, including a
+    deletion, is left out. A rename is the new file only.
+    """
+
+    toplevel = Path(os.fsdecode(_git(root, "rev-parse", "--show-toplevel")).strip())
+    status = _git(
+        root,
+        "status",
+        "--porcelain",
+        "-z",
+        "--no-renames",
+        "--untracked-files=all",
+        "--",
+        ".",
+    )
+    root_resolved = root.resolve()
     found: list[Path] = []
-    for line in result.stdout.splitlines():
-        if len(line) < 4:
+    for entry in status.split(b"\0"):
+        if len(entry) <= 3:
             continue
-        path_text = line[3:].strip()
-        if " -> " in path_text:
-            path_text = path_text.split(" -> ", 1)[1]
-        path_text = path_text.strip('"')
-        found.append((root / path_text).resolve())
+        path = (toplevel / os.fsdecode(entry[3:])).resolve()
+        if not path.is_file() or not path.is_relative_to(root_resolved):
+            continue
+        found.append(path)
     return found
 
 
@@ -195,14 +232,22 @@ def _positionals(root: Path, args: list[str]) -> tuple[list[Path], list[str]]:
     return existing, filters
 
 
-def _tracked_source(path: Path) -> bool:
+def _tracked_source(path: Path, root: Path | None = None) -> bool:
+    parts = path.parts
+    if root is not None:
+        try:
+            parts = path.resolve().relative_to(root.resolve()).parts
+        except ValueError:
+            return False
+    if not SKIP_DIRS.isdisjoint(parts):
+        return False
     if language_of(path) is None:
         return False
     return not is_test_file(path)
 
 
 def _changed_source(root: Path) -> list[Path]:
-    return [path for path in _changed_files(root) if _tracked_source(path)]
+    return [path for path in _changed_files(root) if _tracked_source(path, root)]
 
 
 def _explicit_files(existing: list[Path]) -> list[Path]:
@@ -249,7 +294,11 @@ def run(argv: list[str] | None = None) -> int:
         return options.exit_code
 
     root = options.project_root.resolve()
-    files = select_files(options)
+    try:
+        files = select_files(options)
+    except GitStatusError as exc:
+        print(exc.message, file=sys.stderr)
+        return exc.code
     if not files:
         print("No source files to analyze.")
         return 0

src/dryer/clojure_read.py

@@ -64,6 +64,20 @@ class Splice:
         self.items = items
 
 
+_NEED = object()
+
+
+def _close_coll(kind: str, line: int, items: list, fn_star: bool):
+    if fn_star:
+        return Coll("list", line, items=[Sym("fn*", line), *items])
+    if kind == "map":
+        if len(items) % 2:
+            raise ReadError(line, "map literal has an odd number of forms")
+        pairs = [(items[index], items[index + 1]) for index in range(0, len(items), 2)]
+        return Coll("map", line, pairs=pairs)
+    return Coll(kind, line, items=list(items))
+
+
 class Reader:
     def __init__(self, text: str):
         self.text = text
@@ -146,79 +160,69 @@ class Reader:
             return Kw("", line)
         return Kw(self.read_token(), line)
 
-    def read_collection(self, end: str, kind: str) -> Coll:
-        line = self.line
-        self.get()
-        items: list = []
+    def read_form(self):
+        """One form. None and Splice are real results.
+
+        An explicit stack: a 1,200-deep form must not raise RecursionError.
+        """
+
+        stack: list[tuple] = []
+        produced = _NEED
         while True:
-            self.skip_ws()
+            if produced is _NEED:
+                produced = self._next(stack)
+                continue
+            if not stack:
+                return produced
+            produced = self._deliver(stack, produced)
+
+    def _next(self, stack: list[tuple]):
+        self.skip_ws()
+        if stack and stack[-1][0] == "coll":
+            _tag, end, kind, line, items, fn_star = stack[-1]
             if self.eof():
                 raise ReadError(line, f"unterminated {kind}")
             if self.peek() == end:
                 self.get()
-                break
-            form = self.read_form()
-            if form is None:
-                continue
-            if isinstance(form, Splice):
-                items.extend(form.items)
-                continue
-            items.append(form)
-        if kind == "map":
-            if len(items) % 2:
-                raise ReadError(line, "map literal has an odd number of forms")
-            pairs = [(items[i], items[i + 1]) for i in range(0, len(items), 2)]
-            return Coll("map", line, pairs=pairs)
-        return Coll(kind, line, items=items)
+                stack.pop()
+                return _close_coll(kind, line, items, fn_star)
+        if self.eof():
+            raise ReadError(self.line, "unexpected end of file")
+        return self._open(stack)
 
-    def _wrap(self, name: str, line: int) -> Coll:
-        inner = self.read_form()
-        items = [Sym(name, line)]
-        if inner is not None and not isinstance(inner, Splice):
-            items.append(inner)
-        return Coll("list", line, items=items)
-
-    def read_dispatch(self):
-        line = self.line
-        self.get()
-        ch = self.peek()
-        if ch is None:
-            raise ReadError(line, "incomplete dispatch")
-        if ch == "_":
-            self.get()
-            self.read_form()
+    def _deliver(self, stack: list[tuple], produced):
+        frame = stack[-1]
+        tag = frame[0]
+        if tag == "coll":
+            items = frame[4]
+            if produced is None:
+                return _NEED
+            if isinstance(produced, Splice):
+                items.extend(produced.items)
+            else:
+                items.append(produced)
+            return _NEED
+        stack.pop()
+        if tag == "wrap":
+            _tag, name, line = frame
+            items = [Sym(name, line)]
+            if produced is not None and not isinstance(produced, Splice):
+                items.append(produced)
+            return Coll("list", line, items=items)
+        if tag == "discard":
             return None
-        if ch == "{":
-            return self.read_collection("}", "set")
-        if ch == "(":
-            inner = self.read_collection(")", "list")
-            return Coll("list", line, items=[Sym("fn*", line), *inner.items])
-        if ch == '"':
-            self.read_string()
-            return Lit(line)
-        if ch == "?":
-            self.get()
-            splicing = False
-            if self.peek() == "@":
-                self.get()
-                splicing = True
-            return self.read_cond(splicing)
-        if ch == ":":
-            self.get()
-            if self.peek() == ":":
-                self.get()
-            elif self.peek() not in (None, "{") and self.peek() not in _WHITESPACE:
-                self.read_token()
-            return self.read_form()
-        self.read_token()
-        self.read_form()
-        return Lit(line)
+        if tag == "meta":
+            return _NEED
+        if tag == "lit":
+            return Lit(frame[1])
+        if tag == "cond":
+            return self._finish_cond(frame[1], produced)
+        raise AssertionError(tag)
 
-    def read_cond(self, splicing: bool):
-        form = self.read_form()
-        if not isinstance(form, Coll) or form.kind != "list":
+    def _finish_cond(self, splicing: bool, produced):
+        if not isinstance(produced, Coll) or produced.kind != "list":
             raise ReadError(self.line, "reader conditional body must be a list")
-        chosen = _chosen_branch(form.items)
+        chosen = _chosen_branch(produced.items)
         if chosen is None:
             return None
         if not splicing:
@@ -228,18 +232,21 @@ class Reader:
             raise ReadError(self.line, "splicing reader conditional needs a collection")
         return spliced
 
-    def read_form(self):
-        self.skip_ws()
-        if self.eof():
-            raise ReadError(self.line, "unexpected end of file")
+    def _open(self, stack: list[tuple]):
         ch = self.peek()
         line = self.line
         if ch == "(":
-            return self.read_collection(")", "list")
+            self.get()
+            stack.append(("coll", ")", "list", line, [], False))
+            return _NEED
         if ch == "[":
-            return self.read_collection("]", "vector")
+            self.get()
+            stack.append(("coll", "]", "vector", line, [], False))
+            return _NEED
         if ch == "{":
-            return self.read_collection("}", "map")
+            self.get()
+            stack.append(("coll", "}", "map", line, [], False))
+            return _NEED
         if ch == '"':
             return self.read_string()
         if ch == "\\":
@@ -248,31 +255,75 @@ class Reader:
             return self.read_keyword()
         if ch == "'":
             self.get()
-            return self._wrap("quote", line)
+            stack.append(("wrap", "quote", line))
+            return _NEED
         if ch == "@":
             self.get()
-            return self._wrap("deref", line)
+            stack.append(("wrap", "deref", line))
+            return _NEED
         if ch == "`":
             self.get()
-            return self._wrap("syntax-quote", line)
+            stack.append(("wrap", "syntax-quote", line))
+            return _NEED
         if ch == "~":
             self.get()
             name = "unquote"
             if self.peek() == "@":
                 self.get()
                 name = "unquote-splicing"
-            return self._wrap(name, line)
+            stack.append(("wrap", name, line))
+            return _NEED
         if ch == "^":
             self.get()
-            self.read_form()
-            return self.read_form()
+            stack.append(("meta",))
+            return _NEED
         if ch == "#":
-            return self.read_dispatch()
+            return self._dispatch(stack, line)
         token = self.read_token()
         if _NUMBER.match(token):
             return Lit(line)
         return Sym(token, line)
 
+    def _dispatch(self, stack: list[tuple], line: int):
+        self.get()
+        ch = self.peek()
+        if ch is None:
+            raise ReadError(line, "incomplete dispatch")
+        if ch == "_":
+            self.get()
+            stack.append(("discard",))
+            return _NEED
+        if ch == "{":
+            brace = self.line
+            self.get()
+            stack.append(("coll", "}", "set", brace, [], False))
+            return _NEED
+        if ch == "(":
+            self.get()
+            stack.append(("coll", ")", "list", line, [], True))
+            return _NEED
+        if ch == '"':
+            self.read_string()
+            return Lit(line)
+        if ch == "?":
+            self.get()
+            splicing = False
+            if self.peek() == "@":
+                self.get()
+                splicing = True
+            stack.append(("cond", splicing))
+            return _NEED
+        if ch == ":":
+            self.get()
+            if self.peek() == ":":
+                self.get()
+            elif self.peek() not in (None, "{") and self.peek() not in _WHITESPACE:
+                self.read_token()
+            return _NEED
+        self.read_token()
+        stack.append(("lit", line))
+        return _NEED
+
 
 def _branch_name(feature):
     if isinstance(feature, Kw):
@@ -341,33 +392,18 @@ def _children(form):
 
 def max_line(form) -> int:
     best = getattr(form, "line", 1) or 1
-    for child in _children(form):
-        if child is not None:
-            best = max(best, max_line(child))
+    stack = [form]
+    while stack:
+        current = stack.pop()
+        if current is None:
+            continue
+        best = max(best, getattr(current, "line", 1) or 1)
+        if isinstance(current, Coll):
+            stack.extend(_children(current))
     return best
 
 
-def _normalize_list(form):
-    if not form.items:
-        return [K("list"), K("literal")]
-    head_form, *args = form.items
-    return [K("list"), normalize(head_form, True), *[normalize(arg, False) for arg in args]]
-
-
-def normalize(form, head: bool = False):
-    """dry4clj's `normalize-form`. Collection heads are normalized in full."""
-
-    if isinstance(form, Coll):
-        if form.kind == "list":
-            return _normalize_list(form)
-        if form.kind == "vector":
-            return [K("vector"), *[normalize(item, False) for item in form.items]]
-        if form.kind == "set":
-            return [K("set"), *[normalize(item, False) for item in form.items]]
-        if form.kind == "map":
-            pairs = [[normalize(key, False), normalize(value, False)] for key, value in form.pairs]
-            return [K("map"), *pairs]
-        return [K("literal")]
+def _normalize_atom(form, head: bool):
     if isinstance(form, Sym):
         if head:
             return [K("symbol"), form.name]
@@ -377,6 +413,69 @@ def normalize(form, head: bool = False):
     return K("literal")
 
 
+def _norm_children(form) -> list[tuple[object, bool]]:
+    if form.kind == "list":
+        if not form.items:
+            return []
+        head_form, *args = form.items
+        return [(head_form, True), *[(arg, False) for arg in args]]
+    if form.kind in {"vector", "set"}:
+        return [(item, False) for item in form.items]
+    if form.kind == "map":
+        children = []
+        for key, value in form.pairs:
+            children.append((key, False))
+            children.append((value, False))
+        return children
+    return []
+
+
+def _assemble_form(form, cache):
+    if form.kind == "list":
+        if not form.items:
+            return [K("list"), K("literal")]
+        head_form, *args = form.items
+        return [
+            K("list"),
+            cache[(id(head_form), True)],
+            *[cache[(id(arg), False)] for arg in args],
+        ]
+    if form.kind == "vector":
+        return [K("vector"), *[cache[(id(item), False)] for item in form.items]]
+    if form.kind == "set":
+        return [K("set"), *[cache[(id(item), False)] for item in form.items]]
+    if form.kind == "map":
+        pairs = [
+            [cache[(id(key), False)], cache[(id(value), False)]] for key, value in form.pairs
+        ]
+        return [K("map"), *pairs]
+    return [K("literal")]
+
+
+def normalize(form, head: bool = False):
+    """dry4clj's `normalize-form`. Collection heads are normalized in full."""
+
+    if not isinstance(form, Coll):
+        return _normalize_atom(form, head)
+    cache = {}
+    stack: list[tuple] = [(form, head, False)]
+    while stack:
+        current, is_head, expanded = stack.pop()
+        key = (id(current), is_head)
+        if not expanded:
+            if key in cache:
+                continue
+            if not isinstance(current, Coll):
+                cache[key] = _normalize_atom(current, is_head)
+                continue
+            stack.append((current, is_head, True))
+            for child, child_head in reversed(_norm_children(current)):
+                stack.append((child, child_head, False))
+            continue
+        cache[key] = _assemble_form(current, cache)
+    return cache[(id(form), head)]
+
+
 def is_candidate_form(form) -> bool:
     """A non-empty top-level list whose head is not `ns`."""
 

src/dryer/discover.py

@@ -69,6 +69,8 @@ def is_test_file(path: str | Path) -> bool:
             ".spec.tsx",
             ".test.mts",
             ".spec.mts",
+            ".test.cts",
+            ".spec.cts",
         )
     ):
         return True

src/dryer/extract.py

@@ -73,8 +73,9 @@ def _rust_in_tests(data: bytes, node) -> bool:
 
 def _java_nodes(data: bytes, root) -> list:
     found = []
-
-    def visit(node) -> None:
+    stack = [root]
+    while stack:
+        node = stack.pop()
         if node.type == "method_declaration":
             nested = _inside(
                 node,
@@ -82,11 +83,8 @@ def _java_nodes(data: bytes, root) -> list:
             )
             if _has(node, "block") and not nested:
                 found.append(node)
-            return
-        for child in node.children:
-            visit(child)
-
-    visit(root)
+            continue
+        stack.extend(reversed(node.children))
     return found
 
 

src/dryer/shape.py

@@ -18,38 +18,91 @@ class K:
         return ":" + self.name
 
 
-def pr(node) -> str:
-    """Print a normalized node the way Clojure `pr-str` prints it."""
-
+def _pr_atom(node) -> str:
     if isinstance(node, K):
         return ":" + node.name
     if isinstance(node, str):
         escaped = node.replace("\\", "\\\\").replace('"', '\\"')
         return f'"{escaped}"'
-    if isinstance(node, list):
-        return "[" + " ".join(pr(child) for child in node) + "]"
     raise TypeError(f"cannot print {node!r}")
 
 
+def pr(node) -> str:
+    """Print a normalized node the way Clojure `pr-str` prints it."""
+
+    if not isinstance(node, list):
+        return _pr_atom(node)
+    # Post-order. A 1,200-deep tree must not raise RecursionError.
+    done: dict[int, str] = {}
+    stack: list[tuple[list, bool]] = [(node, False)]
+    while stack:
+        current, expanded = stack.pop()
+        key = id(current)
+        if expanded:
+            pieces = []
+            for child in current:
+                if isinstance(child, list):
+                    pieces.append(done[id(child)])
+                else:
+                    pieces.append(_pr_atom(child))
+            done[key] = "[" + " ".join(pieces) + "]"
+            continue
+        if key in done:
+            continue
+        stack.append((current, True))
+        for child in reversed(current):
+            if isinstance(child, list):
+                stack.append((child, False))
+    return done[id(node)]
+
+
 def node_count(node) -> int:
     """One plus the count of every nested node. Atoms count as one."""
 
-    if isinstance(node, list):
-        return 1 + sum(node_count(child) for child in node)
-    return 1
+    if not isinstance(node, list):
+        return 1
+    total = 0
+    stack: list = [node]
+    while stack:
+        current = stack.pop()
+        total += 1
+        if isinstance(current, list):
+            stack.extend(current)
+    return total
 
 
 def fingerprints(node) -> frozenset[str]:
-    found: set[str] = set()
-
-    def walk(form) -> None:
-        found.add(pr(form))
-        if isinstance(form, list):
-            for child in form:
-                walk(child)
-
-    walk(node)
-    return frozenset(found)
+    if not isinstance(node, list):
+        return frozenset({pr(node)})
+    done: dict[int, str] = {}
+    stack: list[tuple[list, bool]] = [(node, False)]
+    while stack:
+        current, expanded = stack.pop()
+        key = id(current)
+        if expanded:
+            pieces = []
+            for child in current:
+                if isinstance(child, list):
+                    pieces.append(done[id(child)])
+                else:
+                    pieces.append(_pr_atom(child))
+            done[key] = "[" + " ".join(pieces) + "]"
+            continue
+        if key in done:
+            continue
+        stack.append((current, True))
+        for child in reversed(current):
+            if isinstance(child, list):
+                stack.append((child, False))
+    atoms: set[str] = set()
+    pending: list = [node]
+    while pending:
+        current = pending.pop()
+        if isinstance(current, list):
+            pending.extend(current)
+        else:
+            atoms.add(_pr_atom(current))
+    return frozenset(set(done.values()) | atoms)
 
 
 def jaccard(left: frozenset[str], right: frozenset[str]) -> float:

tests/test_cli.py

@@ -1,8 +1,18 @@
+import subprocess
 from pathlib import Path
 
 import pytest
 
-from dryer.cli import _changed_files, _count, _tracked_source, main, parse_args, run, select_files
+from dryer.cli import (
+    GitStatusError,
+    _changed_files,
+    _count,
+    _tracked_source,
+    main,
+    parse_args,
+    run,
+    select_files,
+)
 from dryer.discover import is_test_file, iter_source_files, language_of
 
 
@@ -28,6 +38,8 @@ def test_help_does_not_scan(capsys):
     assert run(["--help"]) == 0
     out = capsys.readouterr().out
     assert "Usage: dryer" in out
+    assert "__pycache__" in out
+    assert ".test.cts" in out
     assert "--threshold" in out
 
 
@@ -153,6 +165,8 @@ def test_names_that_are_tests():
         "ui/a.spec.tsx",
         "ui/a.test.mts",
         "ui/a.spec.mts",
+        "ui/a.test.cts",
+        "ui/a.spec.cts",
         "tests/conftest.py",
         "pkg/foo_test.py",
         "pkg/test_foo.py",
@@ -163,54 +177,71 @@ def test_names_that_are_tests():
     assert not is_test_file("src/app.py")
 
 
-class _Git:
-    def __init__(self, returncode, stdout, stderr=""):
-        self.returncode = returncode
-        self.stdout = stdout
-        self.stderr = stderr
+def _init_repo(path: Path) -> None:
+    subprocess.run(["git", "init"], cwd=path, check=True, capture_output=True)
+    subprocess.run(["git", "config", "user.email", "[email protected]"], cwd=path, check=True)
+    subprocess.run(["git", "config", "user.name", "Dev"], cwd=path, check=True)
 
 
-def test_changed_files_reads_porcelain(tmp_path, monkeypatch):
-    stdout = ' M src/a.py\nR  old.py -> src/b.py\n?? "src/c d.py"\n\nX\n'
-    monkeypatch.setattr(
-        "dryer.cli.subprocess.run",
-        lambda *args, **kwargs: _Git(0, stdout),
-    )
-    assert _changed_files(tmp_path) == [
-        (tmp_path / "src/a.py").resolve(),
-        (tmp_path / "src/b.py").resolve(),
-        (tmp_path / "src/c d.py").resolve(),
-    ]
+def _commit(path: Path, message: str) -> None:
+    subprocess.run(["git", "add", "-A"], cwd=path, check=True, capture_output=True)
+    subprocess.run(["git", "commit", "-m", message], cwd=path, check=True, capture_output=True)
 
 
-def test_changed_files_reports_git_failure(tmp_path, monkeypatch, capsys):
-    monkeypatch.setattr(
-        "dryer.cli.subprocess.run",
-        lambda *args, **kwargs: _Git(1, "", "fatal: not a git repository\n"),
-    )
-    assert _changed_files(tmp_path) == []
-    assert capsys.readouterr().err == "fatal: not a git repository\n"
-
-    monkeypatch.setattr(
-        "dryer.cli.subprocess.run",
-        lambda *args, **kwargs: _Git(1, "", "  \n"),
-    )
-    assert _changed_files(tmp_path) == []
-    assert capsys.readouterr().err == "git status failed\n"
+def test_changed_files_reads_real_status(tmp_path):
+    _init_repo(tmp_path)
+    write_source(tmp_path, "src/a.py", "x = 1\n")
+    write_source(tmp_path, "src/old.py", "x = 1\n")
+    write_source(tmp_path, "gone.py", "x = 1\n")
+    _commit(tmp_path, "base")
+    (tmp_path / "gone.py").unlink()
+    subprocess.run(["git", "mv", "src/old.py", "src/b.py"], cwd=tmp_path, check=True, capture_output=True)
+    write_source(tmp_path, "src/c d.py", "x = 1\n")
+    write_source(tmp_path, "src/café.py", "x = 1\n")
+    write_source(tmp_path, "fresh/nested/new.py", "x = 1\n")
+    found = set(_changed_files(tmp_path))
+    assert (tmp_path / "src/b.py").resolve() in found
+    assert (tmp_path / "src/c d.py").resolve() in found
+    assert (tmp_path / "src/café.py").resolve() in found
+    assert (tmp_path / "fresh/nested/new.py").resolve() in found
+    assert (tmp_path / "gone.py").resolve() not in found
+    assert (tmp_path / "src/old.py").resolve() not in found
 
 
-def test_changed_limits_the_report(tmp_path, monkeypatch, capsys):
+def test_changed_files_reports_git_failure(tmp_path, capsys):
+    with pytest.raises(GitStatusError) as caught:
+        _changed_files(tmp_path)
+    assert caught.value.code == 128
+    assert "not a git repository" in caught.value.message
+    assert capsys.readouterr().err == ""
+
+    code = run(["--changed", "--root", str(tmp_path)])
+    captured = capsys.readouterr()
+    assert code == 128
+    assert "not a git repository" in captured.err
+    assert "No source files" not in captured.out
+
+
+def test_an_empty_git_error_uses_the_fallback_text(monkeypatch, tmp_path):
+    def fail(args, **kwargs):
+        return subprocess.CompletedProcess(args, 128, b"", b"   \n")
+
+    monkeypatch.setattr("dryer.cli.subprocess.run", fail)
+    with pytest.raises(GitStatusError, match="git status failed") as caught:
+        _changed_files(tmp_path)
+    assert caught.value.code == 128
+
+
+def test_changed_limits_the_report(tmp_path, capsys):
     body = "def alpha(xs):\n    ys = filter(xs, odd)\n    zs = map(ys, inc)\n    return list(zs)\n"
     other = "def beta(items):\n    kept = filter(items, even)\n    out = map(kept, dec)\n    return list(out)\n"
+    _init_repo(tmp_path)
+    write_source(tmp_path, "src/c.py", body)
+    _commit(tmp_path, "base")
     write_source(tmp_path, "src/a.py", body)
     write_source(tmp_path, "src/b.py", other)
-    write_source(tmp_path, "src/c.py", body)
     write_source(tmp_path, "tests/test_a.py", other)
-    stdout = " M src/a.py\n M src/b.py\n M tests/test_a.py\n ?? README.md\n"
-    monkeypatch.setattr(
-        "dryer.cli.subprocess.run",
-        lambda *args, **kwargs: _Git(0, stdout),
-    )
+    write_source(tmp_path, "README.md", "notes\n")
     assert run(["--root", str(tmp_path), "--changed", "--min-lines", "3", "--min-nodes", "1"]) == 0
     out = capsys.readouterr().out
     assert "src/a.py" in out
@@ -275,35 +306,28 @@ def test_source_root_limits_the_walk(tmp_path):
     assert files == ["src/a.py"]
 
 
-def test_git_status_asks_for_text_without_checking(tmp_path, monkeypatch):
-    seen = {}
-
-    def fake_run(args, **kwargs):
-        seen["args"] = args
-        seen["kwargs"] = kwargs
-        return _Git(0, " M src/a.py\n")
-
-    monkeypatch.setattr("dryer.cli.subprocess.run", fake_run)
-    assert _changed_files(tmp_path) == [(tmp_path / "src/a.py").resolve()]
-    assert seen["args"] == ["git", "status", "--porcelain"]
-    assert seen["kwargs"]["check"] is False
-    assert seen["kwargs"]["capture_output"] is True
-    assert seen["kwargs"]["text"] is True
-    assert seen["kwargs"]["cwd"] == tmp_path
+def test_changed_from_a_subdirectory_stays_inside_it(tmp_path):
+    repo = tmp_path / "repo"
+    sub = repo / "sub"
+    repo.mkdir()
+    _init_repo(repo)
+    write_source(repo, "src/above.py", "x = 1\n")
+    write_source(sub, "src/below.py", "x = 1\n")
+    chosen = select_files(parse_args(["--root", str(sub), "--changed"]))
+    assert chosen == [(sub / "src/below.py").resolve()]
 
 
-def test_a_status_line_of_four_characters_is_a_path(tmp_path, monkeypatch):
-    monkeypatch.setattr(
-        "dryer.cli.subprocess.run",
-        lambda *args, **kwargs: _Git(0, " M a\n"),
-    )
-    assert _changed_files(tmp_path) == [(tmp_path / "a").resolve()]
+def test_a_short_untracked_name_is_a_file(tmp_path):
+    _init_repo(tmp_path)
+    write_source(tmp_path, "a.py", "x = 1\n")
+    assert _changed_files(tmp_path) == [(tmp_path / "a.py").resolve()]
 
 
 def test_changed_source_skips_unknown_and_test_files(tmp_path):
     assert _tracked_source(tmp_path / "README.md") is False
     assert _tracked_source(tmp_path / "tests" / "test_a.py") is False
     assert _tracked_source(tmp_path / "src" / "a.py") is True
+    assert _tracked_source(tmp_path / "target" / "app.py") is False
 
 
 def test_test_files_and_skipped_directories_are_left_out(tmp_path):
@@ -311,8 +335,11 @@ def test_test_files_and_skipped_directories_are_left_out(tmp_path):
     write_source(tmp_path, "src/foo_test.py", "x = 1\n")
     write_source(tmp_path, "target/app.py", "x = 1\n")
     write_source(tmp_path, "tests/app.py", "x = 1\n")
+    write_source(tmp_path, "src/app.test.cts", "export {}\n")
+    write_source(tmp_path, "src/app.spec.cts", "export {}\n")
+    write_source(tmp_path, "src/widget.cts", "export {}\n")
     relative = [path.relative_to(tmp_path).as_posix() for path in iter_source_files([tmp_path])]
-    assert relative == ["src/app.py"]
+    assert sorted(relative) == ["src/app.py", "src/widget.cts"]
 
 
 def test_a_second_report_replaces_the_snapshot(tmp_path, capsys):

tests/test_languages.py

@@ -599,3 +599,14 @@ def test_keep_uses_the_line_and_node_minimums():
     assert _keep(_entry("a.py", 1, 4, {"q"}, nodes=20), 4, 20) is True
     assert _keep(_entry("a.py", 1, 3, {"q"}, nodes=20), 4, 20) is False
     assert _keep(_entry("a.py", 1, 4, {"q"}, nodes=19), 4, 20) is False
+
+
+def test_a_deep_expression_is_read():
+    expr = " + ".join(["1"] * 1200)
+    entries, warning = entries_in_source("python", f"def place():\n    return {expr}\n", "app.py", "app.py")
+    assert warning is None
+    assert entries
+    nested = "(inc " * 1200 + "x" + ")" * 1200
+    entries, warning = entries_in_source("clojure", f"(defn place [] {nested})\n", "app.clj", "app.clj")
+    assert warning is None
+    assert entries