]> git.cameronkatri.com Git - mandoc.git/blobdiff - html.c
clean up markup of .Bd, .D1, .Dl, .Li, and .Ql;
[mandoc.git] / html.c
diff --git a/html.c b/html.c
index 1498f8dbf2fa636813f86c7f5ab1bb6212059432..fcb2e876ceea63c62832925e44b4e043c76ae4b9 100644 (file)
--- a/html.c
+++ b/html.c
@@ -1,4 +1,4 @@
-/*     $Id: html.c,v 1.194 2017/01/17 01:47:51 schwarze Exp $ */
+/*     $Id: html.c,v 1.199 2017/01/19 16:59:30 schwarze Exp $ */
 /*
  * Copyright (c) 2008-2011, 2014 Kristaps Dzonsons <kristaps@bsd.lv>
  * Copyright (c) 2011-2015, 2017 Ingo Schwarze <schwarze@openbsd.org>
 struct htmldata {
        const char       *name;
        int               flags;
-#define        HTML_CLRLINE     (1 << 0)
-#define        HTML_NOSTACK     (1 << 1)
-#define        HTML_AUTOCLOSE   (1 << 2) /* Tag has auto-closure. */
+#define        HTML_NOSTACK     (1 << 0)
+#define        HTML_AUTOCLOSE   (1 << 1)
+#define        HTML_NLBEFORE    (1 << 2)
+#define        HTML_NLBEGIN     (1 << 3)
+#define        HTML_NLEND       (1 << 4)
+#define        HTML_NLAFTER     (1 << 5)
+#define        HTML_NLAROUND    (HTML_NLBEFORE | HTML_NLAFTER)
+#define        HTML_NLINSIDE    (HTML_NLBEGIN | HTML_NLEND)
+#define        HTML_NLALL       (HTML_NLAROUND | HTML_NLINSIDE)
+#define        HTML_INDENT      (1 << 6)
+#define        HTML_NOINDENT    (1 << 7)
 };
 
 static const struct htmldata htmltags[TAG_MAX] = {
-       {"html",        HTML_CLRLINE}, /* TAG_HTML */
-       {"head",        HTML_CLRLINE}, /* TAG_HEAD */
-       {"body",        HTML_CLRLINE}, /* TAG_BODY */
-       {"meta",        HTML_CLRLINE | HTML_NOSTACK | HTML_AUTOCLOSE}, /* TAG_META */
-       {"title",       HTML_CLRLINE}, /* TAG_TITLE */
-       {"div",         HTML_CLRLINE}, /* TAG_DIV */
-       {"h1",          0}, /* TAG_H1 */
-       {"h2",          0}, /* TAG_H2 */
-       {"span",        0}, /* TAG_SPAN */
-       {"link",        HTML_CLRLINE | HTML_NOSTACK | HTML_AUTOCLOSE}, /* TAG_LINK */
-       {"br",          HTML_CLRLINE | HTML_NOSTACK | HTML_AUTOCLOSE}, /* TAG_BR */
-       {"a",           0}, /* TAG_A */
-       {"table",       HTML_CLRLINE}, /* TAG_TABLE */
-       {"tbody",       HTML_CLRLINE}, /* TAG_TBODY */
-       {"col",         HTML_CLRLINE | HTML_NOSTACK | HTML_AUTOCLOSE}, /* TAG_COL */
-       {"tr",          HTML_CLRLINE}, /* TAG_TR */
-       {"td",          HTML_CLRLINE}, /* TAG_TD */
-       {"li",          HTML_CLRLINE}, /* TAG_LI */
-       {"ul",          HTML_CLRLINE}, /* TAG_UL */
-       {"ol",          HTML_CLRLINE}, /* TAG_OL */
-       {"dl",          HTML_CLRLINE}, /* TAG_DL */
-       {"dt",          HTML_CLRLINE}, /* TAG_DT */
-       {"dd",          HTML_CLRLINE}, /* TAG_DD */
-       {"blockquote",  HTML_CLRLINE}, /* TAG_BLOCKQUOTE */
-       {"pre",         HTML_CLRLINE }, /* TAG_PRE */
-       {"b",           0 }, /* TAG_B */
-       {"i",           0 }, /* TAG_I */
-       {"code",        0 }, /* TAG_CODE */
-       {"small",       0 }, /* TAG_SMALL */
-       {"style",       HTML_CLRLINE}, /* TAG_STYLE */
-       {"math",        HTML_CLRLINE}, /* TAG_MATH */
-       {"mrow",        0}, /* TAG_MROW */
-       {"mi",          0}, /* TAG_MI */
-       {"mo",          0}, /* TAG_MO */
-       {"msup",        0}, /* TAG_MSUP */
-       {"msub",        0}, /* TAG_MSUB */
-       {"msubsup",     0}, /* TAG_MSUBSUP */
-       {"mfrac",       0}, /* TAG_MFRAC */
-       {"msqrt",       0}, /* TAG_MSQRT */
-       {"mfenced",     0}, /* TAG_MFENCED */
-       {"mtable",      0}, /* TAG_MTABLE */
-       {"mtr",         0}, /* TAG_MTR */
-       {"mtd",         0}, /* TAG_MTD */
-       {"munderover",  0}, /* TAG_MUNDEROVER */
-       {"munder",      0}, /* TAG_MUNDER*/
-       {"mover",       0}, /* TAG_MOVER*/
+       {"html",        HTML_NLALL},
+       {"head",        HTML_NLALL | HTML_INDENT},
+       {"body",        HTML_NLALL},
+       {"meta",        HTML_NOSTACK | HTML_AUTOCLOSE | HTML_NLALL},
+       {"title",       HTML_NLAROUND},
+       {"div",         HTML_NLAROUND},
+       {"h1",          HTML_NLAROUND},
+       {"h2",          HTML_NLAROUND},
+       {"span",        0},
+       {"link",        HTML_NOSTACK | HTML_AUTOCLOSE | HTML_NLALL},
+       {"br",          HTML_NOSTACK | HTML_AUTOCLOSE | HTML_NLALL},
+       {"a",           0},
+       {"table",       HTML_NLALL | HTML_INDENT},
+       {"tbody",       HTML_NLALL | HTML_INDENT},
+       {"col",         HTML_NOSTACK | HTML_AUTOCLOSE | HTML_NLALL},
+       {"tr",          HTML_NLALL | HTML_INDENT},
+       {"td",          HTML_NLAROUND},
+       {"li",          HTML_NLAROUND | HTML_INDENT},
+       {"ul",          HTML_NLALL | HTML_INDENT},
+       {"ol",          HTML_NLALL | HTML_INDENT},
+       {"dl",          HTML_NLALL | HTML_INDENT},
+       {"dt",          HTML_NLAROUND},
+       {"dd",          HTML_NLAROUND | HTML_INDENT},
+       {"pre",         HTML_NLALL | HTML_NOINDENT},
+       {"b",           0},
+       {"i",           0},
+       {"code",        0},
+       {"small",       0},
+       {"style",       HTML_NLALL | HTML_INDENT},
+       {"math",        HTML_NLALL | HTML_INDENT},
+       {"mrow",        0},
+       {"mi",          0},
+       {"mo",          0},
+       {"msup",        0},
+       {"msub",        0},
+       {"msubsup",     0},
+       {"mfrac",       0},
+       {"msqrt",       0},
+       {"mfenced",     0},
+       {"mtable",      0},
+       {"mtr",         0},
+       {"mtd",         0},
+       {"munderover",  0},
+       {"munder",      0},
+       {"mover",       0},
 };
 
 static const char      *const roffscales[SCALE_MAX] = {
@@ -106,12 +113,17 @@ static    const char      *const roffscales[SCALE_MAX] = {
 };
 
 static void     a2width(const char *, struct roffsu *);
-static void     bufncat(struct html *, const char *, size_t);
+static void     print_byte(struct html *, char);
+static void     print_endline(struct html *);
+static void     print_endword(struct html *);
+static void     print_indent(struct html *);
+static void     print_word(struct html *, const char *);
+
 static void     print_ctag(struct html *, struct tag *);
-static int      print_escape(char);
-static int      print_encode(struct html *, const char *, int);
+static int      print_escape(struct html *, char);
+static int      print_encode(struct html *, const char *, const char *, int);
+static void     print_href(struct html *, const char *, const char *, int);
 static void     print_metaf(struct html *, enum mandoc_esc);
-static void     print_attr(struct html *, const char *, const char *);
 
 
 void *
@@ -157,13 +169,19 @@ print_gen_head(struct html *h)
        /*
         * Print a default style-sheet.
         */
+
        t = print_otag(h, TAG_STYLE, "");
-       print_text(h, "table.head, table.foot { width: 100%; }\n"
-             "td.head-rtitle, td.foot-os { text-align: right; }\n"
-             "td.head-vol { text-align: center; }\n"
-             "table.foot td { width: 50%; }\n"
-             "table.head td { width: 33%; }\n"
-             "div.spacer { margin: 1em 0; }\n");
+       print_text(h, "table.head, table.foot { width: 100%; }");
+       print_endline(h);
+       print_text(h, "td.head-rtitle, td.foot-os { text-align: right; }");
+       print_endline(h);
+       print_text(h, "td.head-vol { text-align: center; }");
+       print_endline(h);
+       print_text(h, "table.foot td { width: 50%; }");
+       print_endline(h);
+       print_text(h, "table.head td { width: 33%; }");
+       print_endline(h);
+       print_text(h, "div.Pp { margin: 1ex 0ex; }");
        print_tagq(h, t);
 
        if (h->style)
@@ -273,27 +291,27 @@ html_strlen(const char *cp)
 }
 
 static int
-print_escape(char c)
+print_escape(struct html *h, char c)
 {
 
        switch (c) {
        case '<':
-               printf("&lt;");
+               print_word(h, "&lt;");
                break;
        case '>':
-               printf("&gt;");
+               print_word(h, "&gt;");
                break;
        case '&':
-               printf("&amp;");
+               print_word(h, "&amp;");
                break;
        case '"':
-               printf("&quot;");
+               print_word(h, "&quot;");
                break;
        case ASCII_NBRSP:
-               printf("&nbsp;");
+               print_word(h, "&nbsp;");
                break;
        case ASCII_HYPH:
-               putchar('-');
+               print_byte(h, '-');
                break;
        case ASCII_BREAK:
                break;
@@ -304,8 +322,9 @@ print_escape(char c)
 }
 
 static int
-print_encode(struct html *h, const char *p, int norecurse)
+print_encode(struct html *h, const char *p, const char *pend, int norecurse)
 {
+       char             numbuf[16];
        size_t           sz;
        int              c, len, nospace;
        const char      *seq;
@@ -313,24 +332,28 @@ print_encode(struct html *h, const char *p, int norecurse)
        static const char rejs[9] = { '\\', '<', '>', '&', '"',
                ASCII_NBRSP, ASCII_HYPH, ASCII_BREAK, '\0' };
 
+       if (pend == NULL)
+               pend = strchr(p, '\0');
+
        nospace = 0;
 
-       while ('\0' != *p) {
+       while (p < pend) {
                if (HTML_SKIPCHAR & h->flags && '\\' != *p) {
                        h->flags &= ~HTML_SKIPCHAR;
                        p++;
                        continue;
                }
 
-               sz = strcspn(p, rejs);
-
-               fwrite(p, 1, sz, stdout);
-               p += (int)sz;
+               for (sz = strcspn(p, rejs); sz-- && p < pend; p++)
+                       if (*p == ' ')
+                               print_endword(h);
+                       else
+                               print_byte(h, *p);
 
-               if ('\0' == *p)
+               if (p >= pend)
                        break;
 
-               if (print_escape(*p++))
+               if (print_escape(h, *p++))
                        continue;
 
                esc = mandoc_escape(&p, &seq, &len);
@@ -389,21 +412,38 @@ print_encode(struct html *h, const char *p, int norecurse)
                if ((c < 0x20 && c != 0x09) ||
                    (c > 0x7E && c < 0xA0))
                        c = 0xFFFD;
-               if (c > 0x7E)
-                       printf("&#%d;", c);
-               else if ( ! print_escape(c))
-                       putchar(c);
+               if (c > 0x7E) {
+                       (void)snprintf(numbuf, sizeof(numbuf), "&#%d;", c);
+                       print_word(h, numbuf);
+               } else if (print_escape(h, c) == 0)
+                       print_byte(h, c);
        }
 
        return nospace;
 }
 
 static void
-print_attr(struct html *h, const char *key, const char *val)
+print_href(struct html *h, const char *name, const char *sec, int man)
 {
-       printf(" %s=\"", key);
-       (void)print_encode(h, val, 1);
-       putchar('\"');
+       const char      *p, *pp;
+
+       pp = man ? h->base_man : h->base_includes;
+       while ((p = strchr(pp, '%')) != NULL) {
+               print_encode(h, pp, p, 1);
+               if (man && p[1] == 'S') {
+                       if (sec == NULL)
+                               print_byte(h, '1');
+                       else
+                               print_encode(h, sec, NULL, 1);
+               } else if ((man && p[1] == 'N') ||
+                   (man == 0 && p[1] == 'I'))
+                       print_encode(h, name, NULL, 1);
+               else
+                       print_encode(h, p, p + 2, 1);
+               pp = p + 2;
+       }
+       if (*pp != '\0')
+               print_encode(h, pp, NULL, 1);
 }
 
 struct tag *
@@ -411,13 +451,18 @@ print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
 {
        va_list          ap;
        struct roffsu    mysu, *su;
+       char             numbuf[16];
        struct tag      *t;
+       const char      *attr;
        char            *s;
-       int              i, have_style;
+       double           v;
+       int              i, have_style, tflags;
+
+       tflags = htmltags[tag].flags;
 
        /* Push this tags onto the stack of open scopes. */
 
-       if ( ! (HTML_NOSTACK & htmltags[tag].flags)) {
+       if ((tflags & HTML_NOSTACK) == 0) {
                t = mandoc_malloc(sizeof(struct tag));
                t->tag = tag;
                t->next = h->tags.head;
@@ -425,16 +470,19 @@ print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
        } else
                t = NULL;
 
-       if ( ! (HTML_NOSPACE & h->flags))
-               if ( ! (HTML_CLRLINE & htmltags[tag].flags)) {
-                       /* Manage keeps! */
-                       if ( ! (HTML_KEEP & h->flags)) {
-                               if (HTML_PREKEEP & h->flags)
-                                       h->flags |= HTML_KEEP;
-                               putchar(' ');
-                       } else
-                               printf("&#160;");
+       if (tflags & HTML_NLBEFORE)
+               print_endline(h);
+       if (h->col == 0)
+               print_indent(h);
+       else if ((h->flags & HTML_NOSPACE) == 0) {
+               if (h->flags & HTML_KEEP)
+                       print_word(h, "&#160;");
+               else {
+                       if (h->flags & HTML_PREKEEP)
+                               h->flags |= HTML_KEEP;
+                       print_endword(h);
                }
+       }
 
        if ( ! (h->flags & HTML_NONOSPACE))
                h->flags &= ~HTML_NOSPACE;
@@ -443,14 +491,15 @@ print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
 
        /* Print out the tag name and attributes. */
 
-       printf("<%s", htmltags[tag].name);
+       print_byte(h, '<');
+       print_word(h, htmltags[tag].name);
 
        va_start(ap, fmt);
 
        have_style = 0;
        while (*fmt != '\0') {
                if (*fmt == 's') {
-                       printf(" style=\"");
+                       print_word(h, " style=\"");
                        have_style = 1;
                        fmt++;
                        break;
@@ -458,20 +507,43 @@ print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
                s = va_arg(ap, char *);
                switch (*fmt++) {
                case 'c':
-                       print_attr(h, "class", s);
+                       attr = "class";
                        break;
                case 'h':
-                       print_attr(h, "href", s);
+                       attr = "href";
                        break;
                case 'i':
-                       print_attr(h, "id", s);
+                       attr = "id";
                        break;
                case '?':
-                       print_attr(h, s, va_arg(ap, char *));
+                       attr = s;
+                       s = va_arg(ap, char *);
                        break;
                default:
                        abort();
                }
+               print_byte(h, ' ');
+               print_word(h, attr);
+               print_byte(h, '=');
+               print_byte(h, '"');
+               switch (*fmt) {
+               case 'M':
+                       print_href(h, s, va_arg(ap, char *), 1);
+                       fmt++;
+                       break;
+               case 'I':
+                       print_href(h, s, NULL, 0);
+                       fmt++;
+                       break;
+               case 'R':
+                       print_byte(h, '#');
+                       fmt++;
+                       /* FALLTHROUGH */
+               default:
+                       print_encode(h, s, NULL, 1);
+                       break;
+               }
+               print_byte(h, '"');
        }
 
        /* Print out styles. */
@@ -507,53 +579,76 @@ print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
 
                /* Second letter: style name. */
 
-               bufinit(h);
                switch (*fmt++) {
                case 'b':
-                       bufcat_su(h, "margin-bottom", su);
+                       attr = "margin-bottom";
                        break;
                case 'h':
-                       bufcat_su(h, "height", su);
+                       attr = "height";
                        break;
                case 'i':
-                       bufcat_su(h, "text-indent", su);
+                       attr = "text-indent";
                        break;
                case 'l':
-                       bufcat_su(h, "margin-left", su);
+                       attr = "margin-left";
                        break;
                case 't':
-                       bufcat_su(h, "margin-top", su);
+                       attr = "margin-top";
                        break;
                case 'w':
-                       bufcat_su(h, "width", su);
+                       attr = "width";
                        break;
                case 'W':
-                       bufcat_su(h, "min-width", su);
+                       attr = "min-width";
                        break;
                case '?':
-                       bufcat_style(h, s, va_arg(ap, char *));
-                       break;
+                       print_word(h, s);
+                       print_byte(h, ':');
+                       print_byte(h, ' ');
+                       print_word(h, va_arg(ap, char *));
+                       print_byte(h, ';');
+                       if (*fmt != '\0')
+                               print_byte(h, ' ');
+                       continue;
                default:
                        abort();
                }
-               printf("%s", h->buf);
+               v = su->scale;
+               if (su->unit == SCALE_MM && (v /= 100.0) == 0.0)
+                       v = 1.0;
+               else if (su->unit == SCALE_BU)
+                       v /= 24.0;
+               print_word(h, attr);
+               print_byte(h, ':');
+               print_byte(h, ' ');
+               (void)snprintf(numbuf, sizeof(numbuf), "%.2f", v);
+               print_word(h, numbuf);
+               print_word(h, roffscales[su->unit]);
+               print_byte(h, ';');
+               if (*fmt != '\0')
+                       print_byte(h, ' ');
        }
        if (have_style)
-               putchar('"');
+               print_byte(h, '"');
 
        va_end(ap);
 
        /* Accommodate for "well-formed" singleton escaping. */
 
        if (HTML_AUTOCLOSE & htmltags[tag].flags)
-               putchar('/');
+               print_byte(h, '/');
 
-       putchar('>');
+       print_byte(h, '>');
 
-       h->flags |= HTML_NOSPACE;
+       if (tflags & HTML_NLBEGIN)
+               print_endline(h);
+       else
+               h->flags |= HTML_NOSPACE;
 
-       if ((HTML_AUTOCLOSE | HTML_CLRLINE) & htmltags[tag].flags)
-               putchar('\n');
+       if (tflags & HTML_INDENT)
+               h->indent++;
+       if (tflags & HTML_NOINDENT)
+               h->noindent++;
 
        return t;
 }
@@ -561,6 +656,7 @@ print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
 static void
 print_ctag(struct html *h, struct tag *tag)
 {
+       int      tflags;
 
        /*
         * Remember to close out and nullify the current
@@ -571,11 +667,21 @@ print_ctag(struct html *h, struct tag *tag)
        if (tag == h->tblt)
                h->tblt = NULL;
 
-       printf("</%s>", htmltags[tag->tag].name);
-       if (HTML_CLRLINE & htmltags[tag->tag].flags) {
-               h->flags |= HTML_NOSPACE;
-               putchar('\n');
-       }
+       tflags = htmltags[tag->tag].flags;
+
+       if (tflags & HTML_INDENT)
+               h->indent--;
+       if (tflags & HTML_NOINDENT)
+               h->noindent--;
+       if (tflags & HTML_NLEND)
+               print_endline(h);
+       print_indent(h);
+       print_byte(h, '<');
+       print_byte(h, '/');
+       print_word(h, htmltags[tag->tag].name);
+       print_byte(h, '>');
+       if (tflags & HTML_NLAFTER)
+               print_endline(h);
 
        h->tags.head = tag->next;
        free(tag);
@@ -584,22 +690,20 @@ print_ctag(struct html *h, struct tag *tag)
 void
 print_gen_decls(struct html *h)
 {
-
-       puts("<!DOCTYPE html>");
+       print_word(h, "<!DOCTYPE html>");
+       print_endline(h);
 }
 
 void
 print_text(struct html *h, const char *word)
 {
-
-       if ( ! (HTML_NOSPACE & h->flags)) {
-               /* Manage keeps! */
+       if (h->col && (h->flags & HTML_NOSPACE) == 0) {
                if ( ! (HTML_KEEP & h->flags)) {
                        if (HTML_PREKEEP & h->flags)
                                h->flags |= HTML_KEEP;
-                       putchar(' ');
+                       print_endword(h);
                } else
-                       printf("&#160;");
+                       print_word(h, "&#160;");
        }
 
        assert(NULL == h->metaf);
@@ -615,11 +719,12 @@ print_text(struct html *h, const char *word)
                print_otag(h, TAG_I, "");
                break;
        default:
+               print_indent(h);
                break;
        }
 
        assert(word);
-       if ( ! print_encode(h, word, 0)) {
+       if ( ! print_encode(h, word, NULL, 0)) {
                if ( ! (h->flags & HTML_NONOSPACE))
                        h->flags &= ~HTML_NOSPACE;
                h->flags &= ~HTML_NONEWLINE;
@@ -663,150 +768,135 @@ print_paragraph(struct html *h)
 {
        struct tag      *t;
 
-       t = print_otag(h, TAG_DIV, "c", "spacer");
+       t = print_otag(h, TAG_DIV, "c", "Pp");
        print_tagq(h, t);
 }
 
 
+/***********************************************************************
+ * Low level output functions.
+ * They implement line breaking using a short static buffer.
+ ***********************************************************************/
+
 /*
- * Calculate the scaling unit passed in a `-width' argument.  This uses
- * either a native scaling unit (e.g., 1i, 2m) or the string length of
- * the value.
+ * Buffer one HTML output byte.
+ * If the buffer is full, flush and deactivate it and start a new line.
+ * If the buffer is inactive, print directly.
  */
 static void
-a2width(const char *p, struct roffsu *su)
-{
-       if (a2roffsu(p, su, SCALE_MAX) < 2) {
-               su->unit = SCALE_EN;
-               su->scale = html_strlen(p);
-       } else if (su->scale < 0.0)
-               su->scale = 0.0;
-}
-
-void
-bufinit(struct html *h)
-{
-
-       h->buf[0] = '\0';
-       h->buflen = 0;
-}
-
-void
-bufcat_style(struct html *h, const char *key, const char *val)
+print_byte(struct html *h, char c)
 {
+       if ((h->flags & HTML_BUFFER) == 0) {
+               putchar(c);
+               h->col++;
+               return;
+       }
 
-       bufcat(h, key);
-       bufcat(h, ":");
-       bufcat(h, val);
-       bufcat(h, ";");
-}
-
-void
-bufcat(struct html *h, const char *p)
-{
-
-       /*
-        * XXX This is broken and not easy to fix.
-        * When using the -Oincludes option, buffmt_includes()
-        * may pass in strings overrunning BUFSIZ, causing a crash.
-        */
-
-       h->buflen = strlcat(h->buf, p, BUFSIZ);
-       assert(h->buflen < BUFSIZ);
-}
-
-void
-bufcat_fmt(struct html *h, const char *fmt, ...)
-{
-       va_list          ap;
+       if (h->col + h->bufcol < sizeof(h->buf)) {
+               h->buf[h->bufcol++] = c;
+               return;
+       }
 
-       va_start(ap, fmt);
-       (void)vsnprintf(h->buf + (int)h->buflen,
-           BUFSIZ - h->buflen - 1, fmt, ap);
-       va_end(ap);
-       h->buflen = strlen(h->buf);
+       putchar('\n');
+       h->col = 0;
+       print_indent(h);
+       putchar(' ');
+       putchar(' ');
+       fwrite(h->buf, h->bufcol, 1, stdout);
+       putchar(c);
+       h->col = (h->indent + 1) * 2 + h->bufcol + 1;
+       h->bufcol = 0;
+       h->flags &= ~HTML_BUFFER;
 }
 
+/*
+ * If something was printed on the current output line, end it.
+ * Not to be called right after print_indent().
+ */
 static void
-bufncat(struct html *h, const char *p, size_t sz)
+print_endline(struct html *h)
 {
+       if (h->col == 0)
+               return;
 
-       assert(h->buflen + sz + 1 < BUFSIZ);
-       strncat(h->buf, p, sz);
-       h->buflen += sz;
+       if (h->bufcol) {
+               putchar(' ');
+               fwrite(h->buf, h->bufcol, 1, stdout);
+               h->bufcol = 0;
+       }
+       putchar('\n');
+       h->col = 0;
+       h->flags |= HTML_NOSPACE;
+       h->flags &= ~HTML_BUFFER;
 }
 
-void
-buffmt_includes(struct html *h, const char *name)
+/*
+ * Flush the HTML output buffer.
+ * If it is inactive, activate it.
+ */
+static void
+print_endword(struct html *h)
 {
-       const char      *p, *pp;
-
-       pp = h->base_includes;
+       if (h->noindent) {
+               print_byte(h, ' ');
+               return;
+       }
 
-       bufinit(h);
-       while (NULL != (p = strchr(pp, '%'))) {
-               bufncat(h, pp, (size_t)(p - pp));
-               switch (*(p + 1)) {
-               case 'I':
-                       bufcat(h, name);
-                       break;
-               default:
-                       bufncat(h, p, 2);
-                       break;
-               }
-               pp = p + 2;
+       if ((h->flags & HTML_BUFFER) == 0) {
+               h->col++;
+               h->flags |= HTML_BUFFER;
+       } else if (h->bufcol) {
+               putchar(' ');
+               fwrite(h->buf, h->bufcol, 1, stdout);
+               h->col += h->bufcol + 1;
        }
-       if (pp)
-               bufcat(h, pp);
+       h->bufcol = 0;
 }
 
-void
-buffmt_man(struct html *h, const char *name, const char *sec)
+/*
+ * If at the beginning of a new output line,
+ * perform indentation and mark the line as containing output.
+ * Make sure to really produce some output right afterwards,
+ * but do not use print_otag() for producing it.
+ */
+static void
+print_indent(struct html *h)
 {
-       const char      *p, *pp;
+       size_t   i;
 
-       pp = h->base_man;
+       if (h->col)
+               return;
 
-       bufinit(h);
-       while (NULL != (p = strchr(pp, '%'))) {
-               bufncat(h, pp, (size_t)(p - pp));
-               switch (*(p + 1)) {
-               case 'S':
-                       bufcat(h, sec ? sec : "1");
-                       break;
-               case 'N':
-                       bufcat_fmt(h, "%s", name);
-                       break;
-               default:
-                       bufncat(h, p, 2);
-                       break;
-               }
-               pp = p + 2;
+       if (h->noindent == 0) {
+               h->col = h->indent * 2;
+               for (i = 0; i < h->col; i++)
+                       putchar(' ');
        }
-       if (pp)
-               bufcat(h, pp);
+       h->flags &= ~HTML_NOSPACE;
 }
 
-void
-bufcat_su(struct html *h, const char *p, const struct roffsu *su)
+/*
+ * Print or buffer some characters
+ * depending on the current HTML output buffer state.
+ */
+static void
+print_word(struct html *h, const char *cp)
 {
-       double           v;
-
-       v = su->scale;
-       if (SCALE_MM == su->unit && 0.0 == (v /= 100.0))
-               v = 1.0;
-       else if (SCALE_BU == su->unit)
-               v /= 24.0;
-
-       bufcat_fmt(h, "%s: %.2f%s;", p, v, roffscales[su->unit]);
+       while (*cp != '\0')
+               print_byte(h, *cp++);
 }
 
-void
-bufcat_id(struct html *h, const char *src)
+/*
+ * Calculate the scaling unit passed in a `-width' argument.  This uses
+ * either a native scaling unit (e.g., 1i, 2m) or the string length of
+ * the value.
+ */
+static void
+a2width(const char *p, struct roffsu *su)
 {
-
-       /* Cf. <http://www.w3.org/TR/html5/dom.html#the-id-attribute>. */
-
-       for (; '\0' != *src; src++)
-               bufncat(h, *src == ' ' ? "_" : src, 1);
+       if (a2roffsu(p, su, SCALE_MAX) < 2) {
+               su->unit = SCALE_EN;
+               su->scale = html_strlen(p);
+       } else if (su->scale < 0.0)
+               su->scale = 0.0;
 }