@scriptc/runtime 0.0.0 → 0.0.2

package/src/scr_inspect.c

@@ -0,0 +1,803 @@
+/* util.inspect — the runtime half of the static rendering. The compiler
+ * synthesizes one traversal helper per static type (lower-inspect.ts, the
+ * deepStrictEqual precedent); THIS file owns everything the static type
+ * cannot know: Node's exact scalar formatting (number -0, the
+ * single→double→backtick quoting ladder with C0/C1/DEL/lone-surrogate
+ * escapes, the 10000-char string cap and the per-line ` +` continuation
+ * form), Buffer's <Buffer aa bb> hex form, and the LAYOUT engine — a
+ * frame stack reproducing reduceToSingleString, isBelowBreakLength, and
+ * groupArrayElements from Node v24's lib/internal/util/inspect.js for the
+ * DEFAULT options (compact: 3, breakLength: 80, maxArrayLength: 100,
+ * maxStringLength: 10000; non-default options fence in the frontend).
+ *
+ * Lengths follow Node exactly: the break-length math counts UTF-16 code
+ * units (JS .length); the grid-grouping math counts DISPLAY WIDTH. Node
+ * with ICU measures width via icu::getStringWidth over NFC-normalized
+ * text; this file ports the non-ICU fallback tables (full-width/
+ * zero-width code points, no NFC) — pure-ASCII entries agree exactly,
+ * exotic grouped entries may not (SEMANTICS.md divergence).
+ *
+ * The frame stack is process-global (single-threaded runtime, and the
+ * synthesized traversals never call user code — no getters, no custom
+ * inspect — so renders never nest re-entrantly mid-frame).
+ *
+ * Ownership: ScrStr/ScrBytes/ScrDyn/ScrError arguments are BORROWED;
+ * ScrStr results return +1. Nothing here throws; OOM aborts. */
+#include "scr_runtime.h"
+
+#include <math.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+static void insp_oom(void) {
+  fputs("scriptc: out of memory\n", stderr);
+  abort();
+}
+
+/* ── a tiny append-only byte buffer ──────────────────────────────────── */
+
+typedef struct {
+  char *data;
+  size_t len;
+  size_t cap;
+} InspBuf;
+
+static void ib_grow(InspBuf *b, size_t need) {
+  if (b->len + need <= b->cap) return;
+  size_t cap = b->cap ? b->cap * 2 : 64;
+  while (cap < b->len + need) cap *= 2;
+  char *data = realloc(b->data, cap);
+  if (!data) insp_oom();
+  b->data = data;
+  b->cap = cap;
+}
+
+static void ib_bytes(InspBuf *b, const char *s, size_t len) {
+  ib_grow(b, len);
+  memcpy(b->data + b->len, s, len);
+  b->len += len;
+}
+
+static void ib_cstr(InspBuf *b, const char *s) { ib_bytes(b, s, strlen(s)); }
+static void ib_char(InspBuf *b, char c) { ib_bytes(b, &c, 1); }
+static void ib_str(InspBuf *b, const ScrStr *s) { ib_bytes(b, s->data, s->len); }
+
+static void ib_spaces(InspBuf *b, size_t n) {
+  ib_grow(b, n);
+  memset(b->data + b->len, ' ', n);
+  b->len += n;
+}
+
+static ScrStr *ib_take(InspBuf *b) {
+  ScrStr *out = scr_str_new(b->data ? b->data : "", b->len);
+  free(b->data);
+  return out;
+}
+
+/* ── WTF-8 iteration + JS length/width measures ──────────────────────── */
+
+/* Decode the code point at data[i] (WTF-8: lone surrogates ride 3-byte
+ * ED A0..BF xx sequences). Malformed bytes decode as themselves — the
+ * runtime never produces them, this is belt-and-braces. Advances *i. */
+static uint32_t insp_cp_at(const char *data, size_t len, size_t *i) {
+  const unsigned char *p = (const unsigned char *)data;
+  unsigned char c = p[*i];
+  if (c < 0x80) {
+    *i += 1;
+    return c;
+  }
+  if ((c & 0xe0) == 0xc0 && *i + 1 < len) {
+    uint32_t cp = ((uint32_t)(c & 0x1f) << 6) | (p[*i + 1] & 0x3f);
+    *i += 2;
+    return cp;
+  }
+  if ((c & 0xf0) == 0xe0 && *i + 2 < len) {
+    uint32_t cp = ((uint32_t)(c & 0x0f) << 12) | ((uint32_t)(p[*i + 1] & 0x3f) << 6) |
+                  (p[*i + 2] & 0x3f);
+    *i += 3;
+    return cp;
+  }
+  if ((c & 0xf8) == 0xf0 && *i + 3 < len) {
+    uint32_t cp = ((uint32_t)(c & 0x07) << 18) | ((uint32_t)(p[*i + 1] & 0x3f) << 12) |
+                  ((uint32_t)(p[*i + 2] & 0x3f) << 6) | (p[*i + 3] & 0x3f);
+    *i += 4;
+    return cp;
+  }
+  *i += 1;
+  return c;
+}
+
+/* JS .length: UTF-16 code units (astral code points count 2). */
+static size_t insp_utf16_len(const char *data, size_t len) {
+  size_t units = 0;
+  for (size_t i = 0; i < len;) {
+    uint32_t cp = insp_cp_at(data, len, &i);
+    units += cp >= 0x10000 ? 2 : 1;
+  }
+  return units;
+}
+
+/* Node's non-ICU isFullWidthCodePoint table (inspect.js). */
+static bool insp_full_width(uint32_t code) {
+  return code >= 0x1100 &&
+         (code <= 0x115f || code == 0x2329 || code == 0x232a ||
+          (code >= 0x2e80 && code <= 0x3247 && code != 0x303f) ||
+          (code >= 0x3250 && code <= 0x4dbf) || (code >= 0x4e00 && code <= 0xa4c6) ||
+          (code >= 0xa960 && code <= 0xa97c) || (code >= 0xac00 && code <= 0xd7a3) ||
+          (code >= 0xf900 && code <= 0xfaff) || (code >= 0xfe10 && code <= 0xfe19) ||
+          (code >= 0xfe30 && code <= 0xfe6b) || (code >= 0xff01 && code <= 0xff60) ||
+          (code >= 0xffe0 && code <= 0xffe6) || (code >= 0x1b000 && code <= 0x1b001) ||
+          (code >= 0x1f200 && code <= 0x1f251) || (code >= 0x1f300 && code <= 0x1f64f) ||
+          (code >= 0x20000 && code <= 0x3fffd));
+}
+
+/* Node's non-ICU isZeroWidthCodePoint table (inspect.js). */
+static bool insp_zero_width(uint32_t code) {
+  return code <= 0x1f || (code >= 0x7f && code <= 0x9f) ||
+         (code >= 0x300 && code <= 0x36f) || (code >= 0x200b && code <= 0x200f) ||
+         (code >= 0x20d0 && code <= 0x20ff) || (code >= 0xfe00 && code <= 0xfe0f) ||
+         (code >= 0xfe20 && code <= 0xfe2f) || (code >= 0xe0100 && code <= 0xe01ef);
+}
+
+/* Display width (grouping math only): Node's non-ICU getStringWidth,
+ * without the NFC normalization ICU builds apply (SEMANTICS.md). */
+static size_t insp_width(const char *data, size_t len) {
+  size_t width = 0;
+  for (size_t i = 0; i < len;) {
+    uint32_t cp = insp_cp_at(data, len, &i);
+    if (insp_full_width(cp)) width += 2;
+    else if (!insp_zero_width(cp)) width += 1;
+  }
+  return width;
+}
+
+/* ── number rendering ────────────────────────────────────────────────── */
+
+/* formatNumber: JS ToString EXCEPT -0 renders as '-0'. */
+ScrStr *scr_insp_f64(double x) {
+  if (x == 0.0 && signbit(x)) return scr_str_new("-0", 2);
+  char buf[40];
+  size_t n = scr_f64_to_str(x, buf);
+  return scr_str_new(buf, n);
+}
+
+/* ── string quoting (strEscape) ──────────────────────────────────────── */
+
+/* The C0 escape table (inspect.js `meta`, entries 0x00-0x1f). */
+static const char *const INSP_C0[32] = {
+    "\\x00", "\\x01", "\\x02", "\\x03", "\\x04", "\\x05", "\\x06", "\\x07",
+    "\\b",   "\\t",   "\\n",   "\\x0B", "\\f",   "\\r",   "\\x0E", "\\x0F",
+    "\\x10", "\\x11", "\\x12", "\\x13", "\\x14", "\\x15", "\\x16", "\\x17",
+    "\\x18", "\\x19", "\\x1A", "\\x1B", "\\x1C", "\\x1D", "\\x1E", "\\x1F",
+};
+
+static bool insp_contains_byte(const char *data, size_t len, char c) {
+  return memchr(data, c, len) != NULL;
+}
+
+static bool insp_contains_dollar_brace(const char *data, size_t len) {
+  for (size_t i = 0; i + 1 < len; i++) {
+    if (data[i] == '$' && data[i + 1] == '{') return true;
+  }
+  return false;
+}
+
+/* Quote selection: single quotes; double when the text has ' but no ";
+ * backtick when it has both but no ` and no ${. -1 = ", -2 = `, 39 = '. */
+static int insp_pick_quote(const char *data, size_t len) {
+  if (!insp_contains_byte(data, len, '\'')) return 39;
+  if (!insp_contains_byte(data, len, '"')) return -1;
+  if (!insp_contains_byte(data, len, '`') && !insp_contains_dollar_brace(data, len)) return -2;
+  return 39;
+}
+
+static char insp_quote_char(int quote) { return quote == -1 ? '"' : quote == -2 ? '`' : '\''; }
+
+/* strEscape's body between the quotes: escape backslash, C0 (the meta
+ * table), DEL + C1 as \xNN (uppercase), lone surrogates as \udNNN
+ * (lowercase — Number.prototype.toString(16)), and the single quote only
+ * when quoting with '. */
+static void insp_escape_into(InspBuf *b, const char *data, size_t len, int quote) {
+  for (size_t i = 0; i < len;) {
+    size_t start = i;
+    uint32_t cp = insp_cp_at(data, len, &i);
+    if (cp < 0x20) {
+      ib_cstr(b, INSP_C0[cp]);
+    } else if (cp == '\'' && quote == 39) {
+      ib_cstr(b, "\\'");
+    } else if (cp == '\\') {
+      ib_cstr(b, "\\\\");
+    } else if (cp >= 0x7f && cp <= 0x9f) {
+      char esc[8];
+      snprintf(esc, sizeof esc, "\\x%02X", cp);
+      ib_cstr(b, esc);
+    } else if (cp >= 0xd800 && cp <= 0xdfff) {
+      char esc[8];
+      snprintf(esc, sizeof esc, "\\u%x", cp);
+      ib_cstr(b, esc);
+    } else {
+      ib_bytes(b, data + start, i - start);
+    }
+  }
+}
+
+static void insp_quote_into(InspBuf *b, const char *data, size_t len) {
+  int quote = insp_pick_quote(data, len);
+  ib_char(b, insp_quote_char(quote));
+  insp_escape_into(b, data, len, quote);
+  ib_char(b, insp_quote_char(quote));
+}
+
+/* ── the frame stack ─────────────────────────────────────────────────── */
+
+typedef struct {
+  ScrStr **items;
+  size_t len;
+  size_t cap;
+  bool all_num; /* every entry so far flagged numeric (grouping order) */
+} InspFrame;
+
+static InspFrame *g_frames;
+static size_t g_nframes;
+static size_t g_frames_cap;
+static size_t g_indent;       /* ctx.indentationLvl */
+static double g_cur_depth;    /* ctx.currentDepth (last-entered composite) */
+
+static InspFrame *insp_top(void) { return &g_frames[g_nframes - 1]; }
+
+/* Begin a non-empty composite: formatRaw's `recurseTimes += 1;
+ * ctx.currentDepth = recurseTimes` plus the uniform +2 the children
+ * format under. `recurse` is the CHILDREN's recursion depth. */
+void scr_insp_begin(double recurse) {
+  if (g_nframes == g_frames_cap) {
+    g_frames_cap = g_frames_cap ? g_frames_cap * 2 : 8;
+    g_frames = realloc(g_frames, g_frames_cap * sizeof(InspFrame));
+    if (!g_frames) insp_oom();
+  }
+  InspFrame *f = &g_frames[g_nframes++];
+  f->items = NULL;
+  f->len = 0;
+  f->cap = 0;
+  f->all_num = true;
+  g_cur_depth = recurse;
+  g_indent += 2;
+}
+
+/* Append one already-rendered entry (borrowed; retained here). is_num
+ * mirrors `typeof value[i] === 'number'` — it decides grid grouping's
+ * padStart/padEnd order. */
+void scr_insp_entry(ScrStr *s, bool is_num) {
+  InspFrame *f = insp_top();
+  if (f->len == f->cap) {
+    f->cap = f->cap ? f->cap * 2 : 8;
+    f->items = realloc(f->items, f->cap * sizeof(ScrStr *));
+    if (!f->items) insp_oom();
+  }
+  f->items[f->len++] = scr_str_retain(s);
+  if (!is_num) f->all_num = false;
+}
+
+/* remainingText: the "... N more items" tail entry. */
+ScrStr *scr_insp_more_items(double remaining) {
+  char buf[64];
+  int n = snprintf(buf, sizeof buf, "... %.0f more item%s", remaining, remaining > 1 ? "s" : "");
+  return scr_str_new(buf, (size_t)n);
+}
+
+/* groupArrayElements: the grid layout for arrays with more than six
+ * entries. Rewrites the frame's items in place when grouping applies. */
+static void insp_group(InspFrame *f, bool trailing_more) {
+  size_t output_len = f->len - (trailing_more ? 1 : 0);
+  double total_length = 0;
+  double max_length = 0;
+  size_t *data_len = malloc((output_len ? output_len : 1) * sizeof(size_t));
+  if (!data_len) insp_oom();
+  for (size_t i = 0; i < output_len; i++) {
+    size_t w = insp_width(f->items[i]->data, f->items[i]->len);
+    data_len[i] = w;
+    total_length += (double)w + 2;
+    if ((double)w > max_length) max_length = (double)w;
+  }
+  double actual_max = max_length + 2;
+  if (actual_max * 3 + (double)g_indent < 80 &&
+      (total_length / actual_max > 5 || max_length <= 6)) {
+    double approx_char_heights = 2.5;
+    double average_bias = sqrt(actual_max - total_length / (double)f->len);
+    double biased_max = fmax(actual_max - 3 - average_bias, 1);
+    double columns_d = fmin(
+        fmin(round(sqrt(approx_char_heights * biased_max * (double)output_len) / biased_max),
+             floor((80 - (double)g_indent) / actual_max)),
+        fmin(3 * 4 /* ctx.compact * 4 */, 15));
+    if (columns_d <= 1) {
+      free(data_len);
+      return;
+    }
+    size_t columns = (size_t)columns_d;
+    size_t *max_line_length = malloc(columns * sizeof(size_t));
+    if (!max_line_length) insp_oom();
+    for (size_t i = 0; i < columns; i++) {
+      size_t line_max = 0;
+      for (size_t j = i; j < output_len; j += columns) {
+        if (data_len[j] > line_max) line_max = data_len[j];
+      }
+      max_line_length[i] = line_max + 2;
+    }
+    bool pad_start = f->all_num;
+    ScrStr **tmp = NULL;
+    size_t tmp_len = 0, tmp_cap = 0;
+    for (size_t i = 0; i < output_len; i += columns) {
+      size_t max = i + columns < output_len ? i + columns : output_len;
+      InspBuf line = {0};
+      size_t j = i;
+      for (; j + 1 < max; j++) {
+        size_t width = data_len[j];
+        size_t target = max_line_length[j - i];
+        if (pad_start) {
+          if (target > width + 2) ib_spaces(&line, target - width - 2);
+          ib_str(&line, f->items[j]);
+          ib_cstr(&line, ", ");
+        } else {
+          ib_str(&line, f->items[j]);
+          ib_cstr(&line, ", ");
+          if (target > width + 2) ib_spaces(&line, target - width - 2);
+        }
+      }
+      if (pad_start) {
+        size_t width = data_len[j];
+        size_t target = max_line_length[j - i] - 2;
+        if (target > width) ib_spaces(&line, target - width);
+        ib_str(&line, f->items[j]);
+      } else {
+        ib_str(&line, f->items[j]);
+      }
+      if (tmp_len == tmp_cap) {
+        tmp_cap = tmp_cap ? tmp_cap * 2 : 8;
+        tmp = realloc(tmp, tmp_cap * sizeof(ScrStr *));
+        if (!tmp) insp_oom();
+      }
+      tmp[tmp_len++] = ib_take(&line);
+    }
+    if (trailing_more) {
+      if (tmp_len == tmp_cap) {
+        tmp_cap += 1;
+        tmp = realloc(tmp, tmp_cap * sizeof(ScrStr *));
+        if (!tmp) insp_oom();
+      }
+      tmp[tmp_len++] = scr_str_retain(f->items[f->len - 1]);
+    }
+    for (size_t i = 0; i < f->len; i++) scr_str_release(f->items[i]);
+    free(f->items);
+    f->items = tmp;
+    f->len = tmp_len;
+    f->cap = tmp_cap;
+    free(max_line_length);
+  }
+  free(data_len);
+}
+
+/* isBelowBreakLength: UTF-16 lengths against breakLength 80. */
+static bool insp_below_break_length(InspFrame *f, size_t start, const ScrStr *base) {
+  size_t total = f->len + start;
+  if (total + f->len > 80) return false;
+  for (size_t i = 0; i < f->len; i++) {
+    total += insp_utf16_len(f->items[i]->data, f->items[i]->len);
+    if (total > 80) return false;
+  }
+  return base->len == 0 || !insp_contains_byte(base->data, base->len, '\n');
+}
+
+/* reduceToSingleString for the default options (compact: 3). Pops the
+ * frame. base/b0/b1 are borrowed; +1 result. `recurse` is the frame's
+ * recurseTimes (the children's depth); array_extras marks kArrayExtrasType
+ * (grid grouping applies); trailing_more marks a "... N more items" tail
+ * entry (excluded from the grid). */
+ScrStr *scr_insp_end(ScrStr *base, ScrStr *b0, ScrStr *b1, double recurse, bool array_extras,
+                     bool trailing_more) {
+  InspFrame *f = insp_top();
+  g_indent -= 2;
+  size_t entries = f->len;
+  if (array_extras && entries > 6) insp_group(f, trailing_more);
+  InspBuf out = {0};
+  if (g_cur_depth - recurse < 3 && entries == f->len) {
+    size_t start = f->len + g_indent + insp_utf16_len(b0->data, b0->len) +
+                   insp_utf16_len(base->data, base->len) + 10;
+    if (insp_below_break_length(f, start, base)) {
+      bool multiline = false;
+      for (size_t i = 0; i < f->len && !multiline; i++) {
+        multiline = insp_contains_byte(f->items[i]->data, f->items[i]->len, '\n');
+      }
+      if (!multiline) {
+        if (base->len) {
+          ib_str(&out, base);
+          ib_char(&out, ' ');
+        }
+        ib_str(&out, b0);
+        ib_char(&out, ' ');
+        for (size_t i = 0; i < f->len; i++) {
+          if (i) ib_cstr(&out, ", ");
+          ib_str(&out, f->items[i]);
+        }
+        ib_char(&out, ' ');
+        ib_str(&out, b1);
+        goto done;
+      }
+    }
+  }
+  if (base->len) {
+    ib_str(&out, base);
+    ib_char(&out, ' ');
+  }
+  ib_str(&out, b0);
+  for (size_t i = 0; i < f->len; i++) {
+    ib_cstr(&out, i ? ",\n" : "\n");
+    ib_spaces(&out, g_indent + 2);
+    ib_str(&out, f->items[i]);
+  }
+  ib_char(&out, '\n');
+  ib_spaces(&out, g_indent);
+  ib_str(&out, b1);
+done:;
+  for (size_t i = 0; i < f->len; i++) scr_str_release(f->items[i]);
+  free(f->items);
+  g_nframes--;
+  return ib_take(&out);
+}
+
+/* ── string values (formatPrimitive's string half) ───────────────────── */
+
+/* Byte offset of UTF-16 unit `units` — when the boundary splits an astral
+ * code point (JS slice splitting a surrogate pair), the left half keeps
+ * the HIGH surrogate: *lone_high is set and the offset excludes the
+ * 4-byte sequence. */
+static size_t insp_utf16_offset(const char *data, size_t len, size_t units, bool *lone_high,
+                                uint32_t *high_cp) {
+  size_t count = 0;
+  *lone_high = false;
+  for (size_t i = 0; i < len;) {
+    if (count == units) return i;
+    size_t start = i;
+    uint32_t cp = insp_cp_at(data, len, &i);
+    size_t w = cp >= 0x10000 ? 2 : 1;
+    if (count + w > units) {
+      /* boundary splits this astral pair: keep its high surrogate */
+      *lone_high = true;
+      *high_cp = 0xd800 + ((cp - 0x10000) >> 10);
+      return start;
+    }
+    count += w;
+  }
+  return len;
+}
+
+/* formatPrimitive over a string value under the default options:
+ * maxStringLength 10000, then the per-line ` +` continuation when the
+ * (truncated) text is longer than 16 chars AND breakLength(80) -
+ * indentationLvl - 4, split after each newline; the trailer counts the
+ * cut characters. Uses the CURRENT indentation. */
+ScrStr *scr_insp_str(ScrStr *s) {
+  InspBuf out = {0};
+  const char *data = s->data;
+  size_t byte_len = s->len;
+  size_t units = insp_utf16_len(data, byte_len);
+  char trailer[64] = "";
+  char surrogate_tail[4];
+  size_t surrogate_tail_len = 0;
+  if (units > 10000) {
+    size_t remaining = units - 10000;
+    bool lone_high;
+    uint32_t high_cp = 0;
+    byte_len = insp_utf16_offset(data, byte_len, 10000, &lone_high, &high_cp);
+    if (lone_high) {
+      /* re-encode the kept high surrogate as WTF-8 */
+      surrogate_tail[0] = (char)(0xe0 | (high_cp >> 12));
+      surrogate_tail[1] = (char)(0x80 | ((high_cp >> 6) & 0x3f));
+      surrogate_tail[2] = (char)(0x80 | (high_cp & 0x3f));
+      surrogate_tail_len = 3;
+    }
+    snprintf(trailer, sizeof trailer, "... %zu more character%s", remaining,
+             remaining > 1 ? "s" : "");
+    units = 10000;
+  }
+  /* the (possibly truncated) value as one buffer */
+  InspBuf val = {0};
+  ib_bytes(&val, data, byte_len);
+  if (surrogate_tail_len) ib_bytes(&val, surrogate_tail, surrogate_tail_len);
+  size_t vlen = val.len;
+  const char *v = val.data ? val.data : "";
+  if (units > 16 && units > 80 - g_indent - 4 && memchr(v, '\n', vlen) != NULL) {
+    /* split after each newline; quote each chunk; join with ` +\n<indent+2>` */
+    size_t chunk_start = 0;
+    bool first = true;
+    for (size_t i = 0; i <= vlen; i++) {
+      bool at_end = i == vlen;
+      if (at_end ? i > chunk_start : v[i] == '\n') {
+        size_t end = at_end ? i : i + 1;
+        if (!first) {
+          ib_cstr(&out, " +\n");
+          ib_spaces(&out, g_indent + 2);
+        }
+        insp_quote_into(&out, v + chunk_start, end - chunk_start);
+        chunk_start = end;
+        first = false;
+      }
+    }
+  } else {
+    insp_quote_into(&out, v, vlen);
+  }
+  if (trailer[0]) ib_cstr(&out, trailer);
+  free(val.data);
+  return ib_take(&out);
+}
+
+/* ── regex / Buffer / Error leaves ───────────────────────────────────── */
+
+/* RegExp: `/source/flags` (RegExp.prototype.toString — regexes render
+ * fully at any depth; our regexes never carry extra own properties). */
+ScrStr *scr_insp_regex(ScrRegex *re) {
+  InspBuf out = {0};
+  ib_char(&out, '/');
+  ib_bytes(&out, re->source->data, re->source->len);
+  ib_char(&out, '/');
+  ib_bytes(&out, re->flags->data, re->flags->len);
+  return ib_take(&out);
+}
+
+/* Buffer: the custom-inspect hex form `<Buffer aa bb ...>`, capped at
+ * INSPECT_MAX_BYTES (50); renders fully at any depth (custom inspection
+ * runs before the depth check in Node). u8 element kind only (the
+ * frontend fences the rest). */
+ScrStr *scr_insp_buffer(ScrBytes *b) {
+  InspBuf out = {0};
+  ib_cstr(&out, "<Buffer ");
+  size_t max = b->len < 50 ? b->len : 50;
+  for (size_t i = 0; i < max; i++) {
+    char hex[4];
+    snprintf(hex, sizeof hex, "%02x", ((const unsigned char *)b->data)[i]);
+    if (i) ib_char(&out, ' ');
+    ib_bytes(&out, hex, 2);
+  }
+  if (b->len > 50) {
+    char more[48];
+    size_t remaining = b->len - 50;
+    int n = snprintf(more, sizeof more, " ... %zu more byte%s", remaining,
+                     remaining > 1 ? "s" : "");
+    ib_bytes(&out, more, (size_t)n);
+  }
+  ib_char(&out, '>');
+  return ib_take(&out);
+}
+
+/* Errors: the STACKLESS rendering — compiled binaries carry no stack, so
+ * the base is formatError's bracket form `[Name: message]` / `[Name]`
+ * (Node prints these exact forms for an error whose stack is empty;
+ * stack-carrying Node output is a documented divergence). A stamped
+ * `code` slot renders as the one extra own property Node would show:
+ * `[Error: m] { code: 'X' }`, `[Error]` beyond depth. Message newlines
+ * indent like Node's formatError tail. */
+ScrStr *scr_insp_error(ScrError *e, double recurse, double depth) {
+  bool has_code = e->code != NULL; /* the stamped slot, read directly (borrowed) */
+  if (has_code && recurse > depth) {
+    /* `[${getCtxStyle(...)}]` with the trailing space sliced: [Name] */
+    InspBuf out = {0};
+    ib_char(&out, '[');
+    ib_str(&out, e->name);
+    ib_char(&out, ']');
+    return ib_take(&out);
+  }
+  InspBuf base = {0};
+  ib_char(&base, '[');
+  ib_str(&base, e->name);
+  if (e->message->len) {
+    ib_cstr(&base, ": ");
+    /* indent embedded newlines by the CURRENT level (formatError's
+     * final replaceAll — g_indent is the property's own level here). */
+    for (size_t i = 0; i < e->message->len; i++) {
+      char c = e->message->data[i];
+      ib_char(&base, c);
+      if (c == '\n') ib_spaces(&base, g_indent);
+    }
+  }
+  ib_char(&base, ']');
+  ScrStr *base_str = ib_take(&base);
+  if (!has_code) return base_str;
+  scr_insp_begin(recurse + 1);
+  {
+    InspBuf entry = {0};
+    ib_cstr(&entry, "code: ");
+    insp_quote_into(&entry, e->code->data, e->code->len);
+    ScrStr *entry_str = ib_take(&entry);
+    scr_insp_entry(entry_str, false);
+    scr_str_release(entry_str);
+  }
+  ScrStr *b0 = scr_str_new("{", 1);
+  ScrStr *b1 = scr_str_new("}", 1);
+  ScrStr *out = scr_insp_end(base_str, b0, b1, recurse + 1, false, false);
+  scr_str_release(b0);
+  scr_str_release(b1);
+  scr_str_release(base_str);
+  return out;
+}
+
+/* ── dyn values (the checked-dynamic JSON DOM) ───────────────────────── */
+
+/* Property-name rendering: bare when the key matches Node's keyStrRegExp
+ * (/^[a-zA-Z_][a-zA-Z_0-9]*$/, '__proto__' excepted), quoted otherwise. */
+static void insp_key_into(InspBuf *b, const char *key, size_t len) {
+  if (len == 9 && memcmp(key, "__proto__", 9) == 0) {
+    ib_cstr(b, "['__proto__']");
+    return;
+  }
+  bool bare = len > 0;
+  for (size_t i = 0; bare && i < len; i++) {
+    char c = key[i];
+    bool head_ok = (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '_';
+    bool tail_ok = head_ok || (c >= '0' && c <= '9');
+    bare = i == 0 ? head_ok : tail_ok;
+  }
+  if (bare) {
+    ib_bytes(b, key, len);
+  } else {
+    insp_quote_into(b, key, len);
+  }
+}
+
+/* Property-name rendering for the synthesized index-signature record
+ * helpers (their keys are runtime strings, so the bare-or-quoted ladder
+ * must run here): insp_key_into over a ScrStr. Borrows; result +1. */
+ScrStr *scr_insp_key(ScrStr *k) {
+  InspBuf b = {0};
+  insp_key_into(&b, k->data, k->len);
+  return ib_take(&b);
+}
+
+/* The full inspect over a dyn tree — the one runtime type whose SHAPE
+ * lives in the value, so the traversal lives here instead of a
+ * synthesized helper. Same engine, same defaults. dyn-boxed bytes render
+ * in the DOM's documented Uint8Array identity (SEMANTICS.md). */
+ScrStr *scr_insp_dyn(ScrDyn *d, double recurse, double depth) {
+  switch (d->kind) {
+    case SCR_DYN_NULL:
+      return scr_str_new("null", 4);
+    case SCR_DYN_UNDEF:
+      return scr_str_new("undefined", 9);
+    case SCR_DYN_BOOL:
+      return d->v.b ? scr_str_new("true", 4) : scr_str_new("false", 5);
+    case SCR_DYN_NUM:
+      return scr_insp_f64(d->v.num);
+    case SCR_DYN_STR:
+      return scr_insp_str(d->v.str);
+    case SCR_DYN_ARR: {
+      if (d->v.arr.len == 0) return scr_str_new("[]", 2);
+      if (recurse > depth) return scr_str_new("[Array]", 7);
+      scr_insp_begin(recurse + 1);
+      size_t shown = d->v.arr.len < 100 ? d->v.arr.len : 100;
+      for (size_t i = 0; i < shown; i++) {
+        ScrDyn *item = d->v.arr.items[i];
+        ScrStr *s = scr_insp_dyn(item, recurse + 1, depth);
+        scr_insp_entry(s, item->kind == SCR_DYN_NUM);
+        scr_str_release(s);
+      }
+      bool more = d->v.arr.len > 100;
+      if (more) {
+        ScrStr *m = scr_insp_more_items((double)(d->v.arr.len - 100));
+        scr_insp_entry(m, d->v.arr.items[100]->kind == SCR_DYN_NUM);
+        scr_str_release(m);
+      }
+      ScrStr *base = scr_str_new("", 0);
+      ScrStr *b0 = scr_str_new("[", 1);
+      ScrStr *b1 = scr_str_new("]", 1);
+      ScrStr *out = scr_insp_end(base, b0, b1, recurse + 1, true, more);
+      scr_str_release(base);
+      scr_str_release(b0);
+      scr_str_release(b1);
+      return out;
+    }
+    case SCR_DYN_OBJ: {
+      if (d->v.obj.len == 0) return scr_str_new("{}", 2);
+      if (recurse > depth) return scr_str_new("[Object]", 8);
+      scr_insp_begin(recurse + 1);
+      for (size_t i = 0; i < d->v.obj.len; i++) {
+        ScrDynEntry *ent = &d->v.obj.entries[i];
+        ScrStr *val = scr_insp_dyn(ent->value, recurse + 1, depth);
+        InspBuf eb = {0};
+        insp_key_into(&eb, ent->key, ent->key_len);
+        ib_cstr(&eb, ": ");
+        ib_bytes(&eb, val->data, val->len);
+        scr_str_release(val);
+        ScrStr *entry = ib_take(&eb);
+        scr_insp_entry(entry, false);
+        scr_str_release(entry);
+      }
+      ScrStr *base = scr_str_new("", 0);
+      ScrStr *b0 = scr_str_new("{", 1);
+      ScrStr *b1 = scr_str_new("}", 1);
+      ScrStr *out = scr_insp_end(base, b0, b1, recurse + 1, false, false);
+      scr_str_release(base);
+      scr_str_release(b0);
+      scr_str_release(b1);
+      return out;
+    }
+    case SCR_DYN_FUNC: {
+      /* Node's function rendering: [Function: name] / [Function
+       * (anonymous)]. The boxed name is the compiler's best-effort static
+       * spelling (SEMANTICS.md — reference-site naming). */
+      const char *name = d->v.fn.name;
+      InspBuf out = {0};
+      if (name && name[0]) {
+        ib_cstr(&out, "[Function: ");
+        ib_cstr(&out, name);
+        ib_char(&out, ']');
+      } else {
+        ib_cstr(&out, "[Function (anonymous)]");
+      }
+      return ib_take(&out);
+    }
+    case SCR_DYN_HANDLE: {
+      /* Node prints the class's full property dump (ServerResponse {
+       * ... }) — internals we do not model. Fence loudly instead of a
+       * silent-wrong shape (SEMANTICS.md). */
+      InspBuf out = {0};
+      ib_cstr(&out, "util.inspect of a dynamic ");
+      ib_cstr(&out, scr_dyn_handle_cls(d));
+      ib_cstr(&out, " is not supported yet");
+      scr_throw_error(SCR_ERR_ERROR, ib_take(&out));
+      return scr_str_new("", 0);
+    }
+    case SCR_DYN_PROMISE: {
+      /* Node renders Promise { <pending> } / Promise { value } — the
+       * settled-value rendering pulls in the whole inspector; fence
+       * loudly instead of a silent-wrong shape (the handle stance). */
+      InspBuf out = {0};
+      ib_cstr(&out, "util.inspect of a promise value is not supported yet");
+      scr_throw_error(SCR_ERR_ERROR, ib_take(&out));
+      return scr_str_new("", 0);
+    }
+    case SCR_DYN_BYTES: {
+      ScrBytes *b = d->v.bytes;
+      if (d->buffer) {
+        /* Buffer-flavored (a stream chunk): Node's <Buffer ..> form. */
+        return scr_insp_buffer(b);
+      }
+      char prefix[48];
+      int pn = snprintf(prefix, sizeof prefix, "Uint8Array(%zu) [", b->len);
+      if (b->len == 0) {
+        InspBuf out = {0};
+        ib_bytes(&out, prefix, (size_t)pn);
+        ib_char(&out, ']');
+        return ib_take(&out);
+      }
+      if (recurse > depth) return scr_str_new("[Uint8Array]", 12);
+      scr_insp_begin(recurse + 1);
+      size_t shown = b->len < 100 ? b->len : 100;
+      for (size_t i = 0; i < shown; i++) {
+        ScrStr *s = scr_insp_f64((double)((const unsigned char *)b->data)[i]);
+        scr_insp_entry(s, true);
+        scr_str_release(s);
+      }
+      bool more = b->len > 100;
+      if (more) {
+        ScrStr *m = scr_insp_more_items((double)(b->len - 100));
+        scr_insp_entry(m, true);
+        scr_str_release(m);
+      }
+      ScrStr *base = scr_str_new("", 0);
+      ScrStr *b0 = scr_str_new(prefix, (size_t)pn);
+      ScrStr *b1 = scr_str_new("]", 1);
+      ScrStr *out = scr_insp_end(base, b0, b1, recurse + 1, true, more);
+      scr_str_release(base);
+      scr_str_release(b0);
+      scr_str_release(b1);
+      return out;
+    }
+  }
+  return scr_str_new("undefined", 9); /* unreachable */
+}
+
+/* util.format's %s over a dyn value: strings pass VERBATIM (Node's
+ * `typeof arg === 'string' ? arg : inspect(arg)` — the %s depth is 0,
+ * the rest-args depth 2); everything else inspects at the given depth. */
+ScrStr *scr_insp_dyn_s(ScrDyn *d, double depth) {
+  if (d->kind == SCR_DYN_STR) return scr_str_retain(d->v.str);
+  return scr_insp_dyn(d, 0, depth);
+}