dtob-sys 0.1.0

#include "dtob_internal.h"
#include "html_rules.h"
#include <strings.h>

/* ------------------------------------------------------------------ */
/*  XML Reader                                                        */
/* ------------------------------------------------------------------ */

typedef struct {
    const char *buf;
    size_t      pos;
    size_t      len;
    int         html;   /* HTML mode: void elements, raw text, implicit close */
} XmlReader;

static int xr_eof(XmlReader *r) { return r->pos >= r->len; }
static char xr_peek(XmlReader *r) { return xr_eof(r) ? '\0' : r->buf[r->pos]; }
static char xr_consume(XmlReader *r) { return xr_eof(r) ? '\0' : r->buf[r->pos++]; }

static void xr_skip_ws(XmlReader *r)
{
    while (!xr_eof(r) && (r->buf[r->pos] == ' '  || r->buf[r->pos] == '\t' ||
                           r->buf[r->pos] == '\n' || r->buf[r->pos] == '\r'))
        r->pos++;
}

static int xr_match(XmlReader *r, const char *s)
{
    size_t slen = strlen(s);
    if (r->pos + slen > r->len) return 0;
    if (memcmp(r->buf + r->pos, s, slen) != 0) return 0;
    r->pos += slen;
    return 1;
}

static int xr_starts_with(XmlReader *r, const char *s)
{
    size_t slen = strlen(s);
    if (r->pos + slen > r->len) return 0;
    return memcmp(r->buf + r->pos, s, slen) == 0;
}

/* ------------------------------------------------------------------ */
/*  Entity decoding                                                   */
/* ------------------------------------------------------------------ */

/* Append a UTF-8 codepoint to a buffer. Returns number of bytes written. */
static size_t utf8_encode(uint32_t cp, char *out)
{
    if (cp < 0x80) {
        out[0] = (char)cp;
        return 1;
    } else if (cp < 0x800) {
        out[0] = (char)(0xC0 | (cp >> 6));
        out[1] = (char)(0x80 | (cp & 0x3F));
        return 2;
    } else if (cp < 0x10000) {
        out[0] = (char)(0xE0 | (cp >> 12));
        out[1] = (char)(0x80 | ((cp >> 6) & 0x3F));
        out[2] = (char)(0x80 | (cp & 0x3F));
        return 3;
    } else if (cp < 0x110000) {
        out[0] = (char)(0xF0 | (cp >> 18));
        out[1] = (char)(0x80 | ((cp >> 12) & 0x3F));
        out[2] = (char)(0x80 | ((cp >> 6) & 0x3F));
        out[3] = (char)(0x80 | (cp & 0x3F));
        return 4;
    }
    return 0;
}

/* Parse an entity reference starting after '&'. Appends decoded char(s)
 * to *out at *out_pos, growing *out_cap as needed. Returns 1 on success. */
static int xr_decode_entity(XmlReader *r, char **out, size_t *out_pos, size_t *out_cap)
{
    /* ensure space */
    #define ENT_GROW() do { \
        if (*out_pos + 8 >= *out_cap) { \
            *out_cap = *out_cap * 2 + 16; \
            *out = realloc(*out, *out_cap); \
        } \
    } while (0)

    ENT_GROW();

    if (xr_match(r, "amp;"))  { (*out)[(*out_pos)++] = '&';  return 1; }
    if (xr_match(r, "lt;"))   { (*out)[(*out_pos)++] = '<';  return 1; }
    if (xr_match(r, "gt;"))   { (*out)[(*out_pos)++] = '>';  return 1; }
    if (xr_match(r, "quot;")) { (*out)[(*out_pos)++] = '"';  return 1; }
    if (xr_match(r, "apos;")) { (*out)[(*out_pos)++] = '\''; return 1; }

    if (xr_match(r, "#x") || xr_match(r, "#X")) {
        /* hex numeric entity */
        uint32_t cp = 0;
        while (!xr_eof(r) && xr_peek(r) != ';') {
            char c = xr_consume(r);
            if (c >= '0' && c <= '9') cp = cp * 16 + (c - '0');
            else if (c >= 'a' && c <= 'f') cp = cp * 16 + 10 + (c - 'a');
            else if (c >= 'A' && c <= 'F') cp = cp * 16 + 10 + (c - 'A');
            else break;
        }
        if (xr_peek(r) == ';') xr_consume(r);
        *out_pos += utf8_encode(cp, *out + *out_pos);
        return 1;
    }

    if (xr_match(r, "#")) {
        /* decimal numeric entity */
        uint32_t cp = 0;
        while (!xr_eof(r) && xr_peek(r) != ';') {
            char c = xr_consume(r);
            if (c >= '0' && c <= '9') cp = cp * 10 + (c - '0');
            else break;
        }
        if (xr_peek(r) == ';') xr_consume(r);
        *out_pos += utf8_encode(cp, *out + *out_pos);
        return 1;
    }

    /* named entity table */
    static const struct { const char *name; uint32_t cp; } named[] = {
        {"nbsp;",   0x00A0}, {"iexcl;",  0x00A1}, {"cent;",   0x00A2},
        {"pound;",  0x00A3}, {"curren;", 0x00A4}, {"yen;",    0x00A5},
        {"brvbar;", 0x00A6}, {"sect;",   0x00A7}, {"uml;",    0x00A8},
        {"copy;",   0x00A9}, {"ordf;",   0x00AA}, {"laquo;",  0x00AB},
        {"not;",    0x00AC}, {"shy;",    0x00AD}, {"reg;",    0x00AE},
        {"macr;",   0x00AF}, {"deg;",    0x00B0}, {"plusmn;", 0x00B1},
        {"sup2;",   0x00B2}, {"sup3;",   0x00B3}, {"acute;",  0x00B4},
        {"micro;",  0x00B5}, {"para;",   0x00B6}, {"middot;", 0x00B7},
        {"cedil;",  0x00B8}, {"sup1;",   0x00B9}, {"ordm;",   0x00BA},
        {"raquo;",  0x00BB}, {"frac14;", 0x00BC}, {"frac12;", 0x00BD},
        {"frac34;", 0x00BE}, {"iquest;", 0x00BF},
        {"Agrave;", 0x00C0}, {"Aacute;", 0x00C1}, {"Acirc;",  0x00C2},
        {"Atilde;", 0x00C3}, {"Auml;",   0x00C4}, {"Aring;",  0x00C5},
        {"AElig;",  0x00C6}, {"Ccedil;", 0x00C7}, {"Egrave;", 0x00C8},
        {"Eacute;", 0x00C9}, {"Ecirc;",  0x00CA}, {"Euml;",   0x00CB},
        {"Igrave;", 0x00CC}, {"Iacute;", 0x00CD}, {"Icirc;",  0x00CE},
        {"Iuml;",   0x00CF}, {"ETH;",    0x00D0}, {"Ntilde;", 0x00D1},
        {"Ograve;", 0x00D2}, {"Oacute;", 0x00D3}, {"Ocirc;",  0x00D4},
        {"Otilde;", 0x00D5}, {"Ouml;",   0x00D6}, {"times;",  0x00D7},
        {"Oslash;", 0x00D8}, {"Ugrave;", 0x00D9}, {"Uacute;", 0x00DA},
        {"Ucirc;",  0x00DB}, {"Uuml;",   0x00DC}, {"Yacute;", 0x00DD},
        {"THORN;",  0x00DE}, {"szlig;",  0x00DF},
        {"agrave;", 0x00E0}, {"aacute;", 0x00E1}, {"acirc;",  0x00E2},
        {"atilde;", 0x00E3}, {"auml;",   0x00E4}, {"aring;",  0x00E5},
        {"aelig;",  0x00E6}, {"ccedil;", 0x00E7}, {"egrave;", 0x00E8},
        {"eacute;", 0x00E9}, {"ecirc;",  0x00EA}, {"euml;",   0x00EB},
        {"igrave;", 0x00EC}, {"iacute;", 0x00ED}, {"icirc;",  0x00EE},
        {"iuml;",   0x00EF}, {"eth;",    0x00F0}, {"ntilde;", 0x00F1},
        {"ograve;", 0x00F2}, {"oacute;", 0x00F3}, {"ocirc;",  0x00F4},
        {"otilde;", 0x00F5}, {"ouml;",   0x00F6}, {"divide;", 0x00F7},
        {"oslash;", 0x00F8}, {"ugrave;", 0x00F9}, {"uacute;", 0x00FA},
        {"ucirc;",  0x00FB}, {"uuml;",   0x00FC}, {"yacute;", 0x00FD},
        {"thorn;",  0x00FE}, {"yuml;",   0x00FF},
        {"OElig;",  0x0152}, {"oelig;",  0x0153}, {"Scaron;", 0x0160},
        {"scaron;", 0x0161}, {"Yuml;",   0x0178},
        {"fnof;",   0x0192}, {"circ;",   0x02C6}, {"tilde;",  0x02DC},
        {"ensp;",   0x2002}, {"emsp;",   0x2003}, {"thinsp;", 0x2009},
        {"zwnj;",   0x200C}, {"zwj;",    0x200D}, {"lrm;",    0x200E},
        {"rlm;",    0x200F}, {"ndash;",  0x2013}, {"mdash;",  0x2014},
        {"lsquo;",  0x2018}, {"rsquo;",  0x2019}, {"sbquo;",  0x201A},
        {"ldquo;",  0x201C}, {"rdquo;",  0x201D}, {"bdquo;",  0x201E},
        {"dagger;", 0x2020}, {"Dagger;", 0x2021}, {"bull;",   0x2022},
        {"hellip;", 0x2026}, {"permil;", 0x2030}, {"lsaquo;", 0x2039},
        {"rsaquo;", 0x203A}, {"euro;",   0x20AC}, {"trade;",  0x2122},
        {"larr;",   0x2190}, {"uarr;",   0x2191}, {"rarr;",   0x2192},
        {"darr;",   0x2193}, {"harr;",   0x2194}, {"lArr;",   0x21D0},
        {"rArr;",   0x21D2}, {"hArr;",   0x21D4},
        {"forall;", 0x2200}, {"part;",   0x2202}, {"exist;",  0x2203},
        {"empty;",  0x2205}, {"nabla;",  0x2207}, {"isin;",   0x2208},
        {"notin;",  0x2209}, {"ni;",     0x220B}, {"prod;",   0x220F},
        {"sum;",    0x2211}, {"minus;",  0x2212}, {"lowast;", 0x2217},
        {"radic;",  0x221A}, {"prop;",   0x221D}, {"infin;",  0x221E},
        {"ang;",    0x2220}, {"and;",    0x2227}, {"or;",     0x2228},
        {"cap;",    0x2229}, {"cup;",    0x222A}, {"int;",    0x222B},
        {"there4;", 0x2234}, {"sim;",    0x223C}, {"cong;",   0x2245},
        {"asymp;",  0x2248}, {"ne;",     0x2260}, {"equiv;",  0x2261},
        {"le;",     0x2264}, {"ge;",     0x2265}, {"sub;",    0x2282},
        {"sup;",    0x2283}, {"sube;",   0x2286}, {"supe;",   0x2287},
        {"oplus;",  0x2295}, {"otimes;", 0x2297}, {"perp;",   0x22A5},
        {"sdot;",   0x22C5}, {"lceil;",  0x2308}, {"rceil;",  0x2309},
        {"lfloor;", 0x230A}, {"rfloor;", 0x230B}, {"lang;",   0x2329},
        {"rang;",   0x232A}, {"loz;",    0x25CA}, {"spades;", 0x2660},
        {"clubs;",  0x2663}, {"hearts;", 0x2665}, {"diams;",  0x2666},
        {NULL, 0}
    };
    for (int ni = 0; named[ni].name; ni++) {
        if (xr_match(r, named[ni].name)) {
            ENT_GROW();
            *out_pos += utf8_encode(named[ni].cp, *out + *out_pos);
            return 1;
        }
    }

    /* truly unknown — emit '&' and let the rest parse as text */
    (*out)[(*out_pos)++] = '&';
    return 1;
    #undef ENT_GROW
}

/* ------------------------------------------------------------------ */
/*  Skip constructs we discard                                        */
/* ------------------------------------------------------------------ */

static void xr_skip_comment(XmlReader *r)
{
    /* already consumed '<!--' */
    while (!xr_eof(r)) {
        if (xr_match(r, "-->")) return;
        r->pos++;
    }
}

static void xr_skip_pi(XmlReader *r)
{
    /* already consumed '<?' */
    while (!xr_eof(r)) {
        if (xr_match(r, "?>")) return;
        r->pos++;
    }
}

static void xr_skip_doctype(XmlReader *r)
{
    /* already consumed '<!DOCTYPE' or '<!doctype' etc */
    int depth = 1;
    while (!xr_eof(r) && depth > 0) {
        char c = xr_consume(r);
        if (c == '<') depth++;
        else if (c == '>') depth--;
    }
}

/* ------------------------------------------------------------------ */
/*  CDATA                                                             */
/* ------------------------------------------------------------------ */

static DtobValue *xr_parse_cdata(XmlReader *r)
{
    /* already consumed '<![CDATA[' */
    size_t start = r->pos;
    while (!xr_eof(r)) {
        if (xr_starts_with(r, "]]>")) {
            size_t end = r->pos;
            r->pos += 3;
            return dtob_raw((const uint8_t *)r->buf + start, end - start);
        }
        r->pos++;
    }
    /* unterminated CDATA — return what we have */
    return dtob_raw((const uint8_t *)r->buf + start, r->len - start);
}

/* ------------------------------------------------------------------ */
/*  Text node                                                         */
/* ------------------------------------------------------------------ */

static DtobValue *xr_parse_text(XmlReader *r)
{
    size_t cap = 64;
    size_t pos = 0;
    char *buf = malloc(cap);

    while (!xr_eof(r) && xr_peek(r) != '<') {
        if (xr_peek(r) == '&') {
            xr_consume(r);
            xr_decode_entity(r, &buf, &pos, &cap);
        } else {
            if (pos + 1 >= cap) {
                cap *= 2;
                buf = realloc(buf, cap);
            }
            buf[pos++] = xr_consume(r);
        }
    }

    DtobValue *v = json_string(buf, pos);
    free(buf);
    return v;
}

/* ------------------------------------------------------------------ */
/*  Tag name                                                          */
/* ------------------------------------------------------------------ */

static int is_name_char(char c)
{
    return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
           (c >= '0' && c <= '9') || c == '-' || c == '_' || c == '.' || c == ':';
}

static char *xr_parse_name(XmlReader *r, size_t *out_len)
{
    size_t start = r->pos;
    while (!xr_eof(r) && is_name_char(xr_peek(r)))
        r->pos++;
    size_t len = r->pos - start;
    if (len == 0) return NULL;
    char *name = malloc(len + 1);
    memcpy(name, r->buf + start, len);
    name[len] = '\0';
    if (out_len) *out_len = len;
    return name;
}

/* ------------------------------------------------------------------ */
/*  Attribute value                                                   */
/* ------------------------------------------------------------------ */

static char *xr_parse_attr_value(XmlReader *r, size_t *out_len)
{
    char quote = xr_consume(r); /* '"' or '\'' */
    if (quote != '"' && quote != '\'') return NULL;

    size_t cap = 64;
    size_t pos = 0;
    char *buf = malloc(cap);

    while (!xr_eof(r) && xr_peek(r) != quote) {
        if (xr_peek(r) == '&') {
            xr_consume(r);
            xr_decode_entity(r, &buf, &pos, &cap);
        } else {
            if (pos + 1 >= cap) {
                cap *= 2;
                buf = realloc(buf, cap);
            }
            buf[pos++] = xr_consume(r);
        }
    }
    if (xr_peek(r) == quote) xr_consume(r); /* closing quote */

    *out_len = pos;
    return buf;
}

/* HTML unquoted attribute value: everything until whitespace or '>' */
static char *xr_parse_unquoted_attr_value(XmlReader *r, size_t *out_len)
{
    size_t start = r->pos;
    while (!xr_eof(r)) {
        char c = xr_peek(r);
        if (c == ' ' || c == '\t' || c == '\n' || c == '\r' ||
            c == '>' || c == '/') break;
        r->pos++;
    }
    size_t len = r->pos - start;
    if (len == 0) return NULL;
    char *buf = malloc(len + 1);
    memcpy(buf, r->buf + start, len);
    buf[len] = '\0';
    *out_len = len;
    return buf;
}

/* ------------------------------------------------------------------ */
/*  Raw text content (for <style>, <script>, <xmp>)                   */
/* ------------------------------------------------------------------ */

static DtobValue *xr_parse_raw_content(XmlReader *r, const char *tag)
{
    /* scan for </tag> case-insensitively */
    size_t tag_len = strlen(tag);
    size_t start = r->pos;

    while (!xr_eof(r)) {
        if (r->buf[r->pos] == '<' && r->pos + 2 + tag_len < r->len &&
            r->buf[r->pos + 1] == '/' &&
            strncasecmp(r->buf + r->pos + 2, tag, tag_len) == 0) {
            char after = r->buf[r->pos + 2 + tag_len];
            if (after == '>' || after == ' ' || after == '\t' ||
                after == '\n' || after == '\r') {
                size_t end = r->pos;
                r->pos += 2 + tag_len;
                while (!xr_eof(r) && xr_peek(r) != '>') r->pos++;
                if (xr_peek(r) == '>') xr_consume(r);
                return json_string(r->buf + start, end - start);
            }
        }
        r->pos++;
    }
    return json_string(r->buf + start, r->len - start);
}

/* Escapable raw text (for <textarea>, <title>): entities decoded, no child elements */
static DtobValue *xr_parse_escapable_raw_content(XmlReader *r, const char *tag)
{
    size_t tag_len = strlen(tag);
    size_t cap = 64;
    size_t pos = 0;
    char *buf = malloc(cap);

    while (!xr_eof(r)) {
        /* check for </tag> */
        if (r->buf[r->pos] == '<' && r->pos + 2 + tag_len < r->len &&
            r->buf[r->pos + 1] == '/' &&
            strncasecmp(r->buf + r->pos + 2, tag, tag_len) == 0) {
            char after = r->buf[r->pos + 2 + tag_len];
            if (after == '>' || after == ' ' || after == '\t' ||
                after == '\n' || after == '\r') {
                r->pos += 2 + tag_len;
                while (!xr_eof(r) && xr_peek(r) != '>') r->pos++;
                if (xr_peek(r) == '>') xr_consume(r);
                break;
            }
        }
        if (xr_peek(r) == '&') {
            xr_consume(r);
            xr_decode_entity(r, &buf, &pos, &cap);
        } else {
            if (pos + 1 >= cap) {
                cap *= 2;
                buf = realloc(buf, cap);
            }
            buf[pos++] = xr_consume(r);
        }
    }

    DtobValue *v = json_string(buf, pos);
    free(buf);
    return v;
}

/* ------------------------------------------------------------------ */
/*  Peek at upcoming tag name (for implicit closing)                  */
/* ------------------------------------------------------------------ */

/* Peek at the tag name after '<' without consuming. Returns static pointer
 * into a thread-local buffer. Returns NULL if not a valid tag start. */
static const char *xr_peek_tag_name(XmlReader *r)
{
    static char peek_buf[64];
    if (xr_peek(r) != '<') return NULL;
    size_t p = r->pos + 1;
    /* skip '</' — we only care about opening tags for implicit close */
    if (p < r->len && r->buf[p] == '/') return NULL;
    size_t start = p;
    while (p < r->len && is_name_char(r->buf[p])) p++;
    size_t len = p - start;
    if (len == 0 || len >= sizeof(peek_buf)) return NULL;
    memcpy(peek_buf, r->buf + start, len);
    peek_buf[len] = '\0';
    return peek_buf;
}

/* ------------------------------------------------------------------ */
/*  Element parser                                                    */
/* ------------------------------------------------------------------ */

static DtobValue *xr_parse_element(XmlReader *r);

static DtobValue *xr_parse_node(XmlReader *r)
{
    /* HTML: only skip whitespace if followed by a tag '<'.  If followed
     * by text, the whitespace is inter-word spacing and must be preserved
     * (it gets collapsed to a single space by the caller). */
    if (r->html) {
        size_t saved = r->pos;
        xr_skip_ws(r);
        if (xr_eof(r)) return NULL;
        if (xr_peek(r) != '<') r->pos = saved; /* rewind — part of text */
    } else {
        xr_skip_ws(r);
        if (xr_eof(r)) return NULL;
    }

    if (xr_peek(r) == '<') {
        /* check what kind of <...> */
        if (xr_starts_with(r, "<!--")) {
            r->pos += 4;
            xr_skip_comment(r);
            return xr_parse_node(r); /* skip and continue */
        }
        if (xr_starts_with(r, "<?")) {
            r->pos += 2;
            xr_skip_pi(r);
            return xr_parse_node(r);
        }
        if (!r->html && xr_starts_with(r, "<![CDATA[")) {
            r->pos += 9;
            return xr_parse_cdata(r);
        }
        if (xr_starts_with(r, "<!")) {
            r->pos += 2;
            xr_skip_doctype(r);
            return xr_parse_node(r);
        }
        if (xr_starts_with(r, "</")) {
            return NULL; /* closing tag — caller handles */
        }
        return xr_parse_element(r);
    }

    /* text node */
    return xr_parse_text(r);
}

/* Consume a closing tag only if it matches the current element (HTML mode).
 * In XML mode, always consume (assumes well-formed input). */
static void xr_consume_close_tag(XmlReader *r, const char *tag)
{
    /* r is positioned at '</' */
    size_t saved = r->pos;
    r->pos += 2;
    size_t name_start = r->pos;
    while (!xr_eof(r) && is_name_char(xr_peek(r))) r->pos++;
    size_t name_len = r->pos - name_start;

    if (r->html) {
        if (name_len == strlen(tag) &&
            strncasecmp(r->buf + name_start, tag, name_len) == 0) {
            /* matching close tag — consume it */
            while (!xr_eof(r) && xr_peek(r) != '>') r->pos++;
            if (xr_peek(r) == '>') xr_consume(r);
        } else {
            /* mismatched close tag — don't consume, let parent handle */
            r->pos = saved;
        }
    } else {
        /* XML: always consume */
        while (!xr_eof(r) && xr_peek(r) != '>') r->pos++;
        if (xr_peek(r) == '>') xr_consume(r);
    }
}

static DtobValue *xr_parse_element(XmlReader *r)
{
    /* consume '<' */
    xr_consume(r);

    /* tag name */
    size_t tag_len;
    char *tag = xr_parse_name(r, &tag_len);
    if (!tag) {
        fprintf(stderr, "xml: expected tag name\n");
        return NULL;
    }

    /* HTML: normalize tag to lowercase */
    if (r->html) html_normalize_tag(tag, tag_len);

    /* attributes */
    DtobValue *attrs = dtob_kvset();
    xr_skip_ws(r);
    while (!xr_eof(r) && xr_peek(r) != '>' && xr_peek(r) != '/') {
        size_t aname_len;
        char *aname = xr_parse_name(r, &aname_len);
        if (!aname) break;

        /* HTML: normalize attribute name to lowercase */
        if (r->html) html_normalize_tag(aname, aname_len);

        xr_skip_ws(r);
        if (xr_peek(r) == '=') {
            xr_consume(r); /* '=' */
            xr_skip_ws(r);
            size_t vlen;
            char *val;
            char next = xr_peek(r);
            if (r->html && next != '"' && next != '\'') {
                val = xr_parse_unquoted_attr_value(r, &vlen);
            } else {
                val = xr_parse_attr_value(r, &vlen);
            }
            if (val) {
                const char *key = aname;
                if (!r->html) {
                    /* XML: strip namespace prefix */
                    char *colon = strchr(aname, ':');
                    if (colon) key = colon + 1;
                }
                dtob_kvset_put(attrs, key, json_string(val, vlen));
                free(val);
            }
        } else {
            /* boolean attribute (no value) */
            dtob_kvset_put(attrs, aname, json_true());
        }
        free(aname);
        xr_skip_ws(r);
    }

    /* strip namespace prefix from tag (XML only — HTML tags don't have ns) */
    const char *tag_clean = tag;
    if (!r->html) {
        char *colon = strchr(tag, ':');
        if (colon) tag_clean = colon + 1;
    }

    /* self-closing? */
    int self_closing = 0;
    if (xr_match(r, "/>")) {
        self_closing = 1;
    } else if (xr_peek(r) == '>') {
        xr_consume(r);
    }

    /* build result kv_set */
    DtobValue *elem = dtob_kvset();
    dtob_kvset_put(elem, "t", json_string(tag_clean, strlen(tag_clean)));
    dtob_kvset_put(elem, "a", attrs);

    DtobValue *children = dtob_array();

    if (!self_closing && r->html && html_is_void(tag_clean)) {
        /* void element — no children, no closing tag */
    } else if (!self_closing && r->html && html_is_raw_text(tag_clean)) {
        DtobValue *raw = xr_parse_raw_content(r, tag_clean);
        if (raw) dtob_array_push(children, raw);
    } else if (!self_closing && r->html && html_is_escapable_raw_text(tag_clean)) {
        DtobValue *raw = xr_parse_escapable_raw_content(r, tag_clean);
        if (raw) dtob_array_push(children, raw);
    } else if (!self_closing) {
        /* parse children until closing tag */
        while (!xr_eof(r)) {
            /* HTML: do not skip whitespace here — let it parse as text nodes.
             * The whitespace-only text node filter below handles cleanup. */

            /* check for closing tag */
            if (xr_starts_with(r, "</")) {
                xr_consume_close_tag(r, tag_clean);
                break;
            }

            /* HTML: check if upcoming opening tag implicitly closes us */
            if (r->html) {
                const char *upcoming = xr_peek_tag_name(r);
                if (upcoming && html_implicitly_closes(tag_clean, upcoming))
                    break; /* don't consume — parent or sibling handles it */
            }

            DtobValue *child = xr_parse_node(r);
            if (!child) break;

            /* HTML: collapse whitespace-only text nodes to a single space.
             * This preserves inter-word spacing (e.g. "<em>x</em> y")
             * while normalizing multi-line formatting whitespace. */
            if (r->html && child->type == JSON_STRING) {
                int all_ws = 1;
                for (size_t i = 0; i < child->data_len; i++) {
                    char c = (char)child->data[i];
                    if (c != ' ' && c != '\t' && c != '\n' && c != '\r') {
                        all_ws = 0;
                        break;
                    }
                }
                if (all_ws) {
                    if (child->data_len > 0) {
                        /* collapse to single space */
                        child->data[0] = ' ';
                        child->data_len = 1;
                    } else {
                        dtob_free(child);
                        continue;
                    }
                }
            }

            dtob_array_push(children, child);
        }
    }

    /* HTML: the loop may have exited via `child == NULL` (xr_parse_node saw
     * '</' after skipping whitespace/comments) before the top-of-loop check
     * had a chance to consume the close tag.  Consume it now if it matches. */
    if (!self_closing && r->html && xr_starts_with(r, "</"))
        xr_consume_close_tag(r, tag_clean);

    dtob_kvset_put(elem, "c", children);
    free(tag);
    return elem;
}

/* ------------------------------------------------------------------ */
/*  Public entry points                                               */
/* ------------------------------------------------------------------ */

static DtobValue *parse_markup(const char *xml, int html)
{
    XmlReader r = { xml, 0, strlen(xml), html };

    /* skip leading whitespace and preamble (comments, PIs, doctypes) */
    while (!xr_eof(&r)) {
        xr_skip_ws(&r);
        if (xr_eof(&r)) break;

        if (xr_starts_with(&r, "<!--")) {
            r.pos += 4;
            xr_skip_comment(&r);
        } else if (xr_starts_with(&r, "<?")) {
            r.pos += 2;
            xr_skip_pi(&r);
        } else if (xr_starts_with(&r, "<!")) {
            r.pos += 2;
            xr_skip_doctype(&r);
        } else {
            break;
        }
    }

    if (xr_eof(&r) || xr_peek(&r) != '<') {
        fprintf(stderr, "xml: no root element found\n");
        return NULL;
    }

    return xr_parse_element(&r);
}

DtobValue *dtob_from_xml(const char *xml)
{
    return parse_markup(xml, 0);
}

DtobValue *dtob_from_html(const char *html)
{
    return parse_markup(html, 1);
}