HTML cleaner utility.

1 files changed, 265 insertions, 0 deletions

diff --git a/hc.c b/hc.c
new file mode 100644
index 0000000..4c14e04
--- /dev/null
+++ b/hc.c
@@ -0,0 +1,265 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include "hc.h"
+
+
+static int allowed_el_spec[] = {
+  tok_el_a,
+  tok_el_abbr,
+  tok_el_acronym,
+  tok_el_address,
+  /* tok_el_applet, */
+  /* tok_el_area, */
+  tok_el_b,
+  /* tok_el_base, */
+  tok_el_basefont,
+  tok_el_bdo,
+  tok_el_big,
+  tok_el_blockquote,
+  /* tok_el_body, */
+  tok_el_br,
+  /* tok_el_button, */
+  tok_el_caption,
+  tok_el_center,
+  tok_el_cite,
+  tok_el_code,
+  tok_el_col,
+  tok_el_colgroup,
+  tok_el_dd,
+  tok_el_del,
+  tok_el_dfn,
+  tok_el_dir,
+  tok_el_div,
+  tok_el_dl,
+  tok_el_dt,
+  tok_el_em,
+  /* tok_el_fieldset, */
+  tok_el_font,
+  tok_el_form,
+  /* tok_el_frame, */
+  /* tok_el_frameset, */
+  tok_el_h1,
+  tok_el_h2,
+  tok_el_h3,
+  tok_el_h4,
+  tok_el_h5,
+  tok_el_h6,
+  /* tok_el_head, */
+  tok_el_hr,
+  /* tok_el_html, */
+  tok_el_i,
+  /* tok_el_iframe, */
+  tok_el_img,
+  /* tok_el_input, */
+  tok_el_ins,
+  tok_el_kbd,
+  /* tok_el_label, */
+  /* tok_el_legend, */
+  tok_el_li,
+  /* tok_el_link, */
+  /* tok_el_map, */
+  /* tok_el_menu, */
+  /* tok_el_meta, */
+  /* tok_el_noframes, */
+  /* tok_el_noscript, */
+  /* tok_el_object, */
+  tok_el_ol,
+  tok_el_optgroup,
+  /* tok_el_option, */
+  tok_el_p,
+  /* tok_el_param, */
+  tok_el_pre,
+  tok_el_q,
+  tok_el_samp,
+  /* tok_el_script,*/
+  /* tok_el_select, */
+  tok_el_small,
+  tok_el_span,
+  tok_el_strike,
+  tok_el_strong,
+  tok_el_style,
+  tok_el_sub,
+  tok_el_sup,
+  tok_el_table,
+  tok_el_tbody,
+  tok_el_td,
+  /* tok_el_textarea, */
+  tok_el_tfoot,
+  tok_el_th,
+  tok_el_thead,
+  /* tok_el_title, */
+  tok_el_tr,
+  tok_el_tt,
+  tok_el_u,
+  tok_el_ul,
+  /* tok_el_var, */
+  tok_eof,
+};
+
+static int allowed_el[tok_max];
+
+static const token_t blank;
+static token_t pushback;
+
+static void bail()
+{
+  fprintf(stderr, "bad html\n");
+  exit(EXIT_FAILURE);
+}
+
+static token_t mktok(toktype_t type, char *text)
+{
+  token_t tok = { 0, 0, 0, 0 };
+  tok.type = type;
+  tok.lexeme = strdup(text);
+  return tok;
+}
+
+static void deltok(token_t tok)
+{
+  free(tok.lexeme);
+}
+
+static int null(token_t tok)
+{
+  return tok.type == tok_eof;
+}
+
+static token_t gettok(void)
+{
+  if (null(pushback)) {
+    int type = yylex();
+    return mktok(type, yytext);
+  } else {
+    token_t tok = pushback;
+    pushback = blank;
+    return tok;
+  }
+}
+
+static void ungettok(token_t tok)
+{
+  deltok(pushback);
+  pushback = tok;
+}
+
+static token_t printtok(token_t tok)
+{
+  if (!null(tok))
+    fputs(tok.lexeme, stdout);
+  return tok;
+}
+
+static token_t match(int type)
+{
+  token_t tok = gettok();
+  if (tok.type != type)
+    bail();
+  return tok;
+}
+
+static token_t optmatch(int type)
+{
+  token_t tok = gettok();
+  if (tok.type != type) {
+    ungettok(tok);
+    return blank;
+  }
+  return tok;
+}
+
+static token_t lookfor(int type)
+{
+  token_t tok;
+  for (;;) {
+    tok = gettok();
+    if (tok.type == type || null(tok))
+      break;
+    deltok(tok);
+  }
+  return tok;
+}
+
+static token_t printuntil(int type)
+{
+  token_t tok;
+  for (;;) {
+    tok = gettok();
+    if (tok.type == type || null(tok))
+      break;
+    deltok(printtok(tok));
+  }
+  printtok(tok);
+  return tok;
+}
+
+static void parse_element(token_t in)
+{
+  token_t end = optmatch('/');
+  token_t name = gettok();
+
+  switch (name.type) {
+  case '/': 
+    if (!null(end))
+      bail();
+    printtok(in);
+    printtok(name);
+    deltok(printtok(lookfor('>')));
+    goto out; 
+    return;
+  case '>':
+    bail();
+  default:
+    break;
+  }
+
+  if (allowed_el[name.type]) {
+    printtok(in);
+    printtok(end);
+    printtok(name);
+    deltok(printuntil('>'));
+  } else {
+    deltok(lookfor('>'));
+  }
+
+out:
+  deltok(end);
+  deltok(name);
+}
+
+static void parse(void)
+{
+  for (;;) {
+    token_t tok = gettok();
+
+    switch (tok.type) {
+    case '<':
+      parse_element(tok);
+      break;
+    case tok_eof:
+      deltok(tok);
+      return;
+    default:
+      printtok(tok);
+      break;
+    }
+
+    deltok(tok);
+  }
+}
+
+static void init(void)
+{
+  int i;
+
+  for (i = 0; allowed_el_spec[i] != tok_eof; i++)
+    allowed_el[allowed_el_spec[i]] = 1;
+}
+
+int main(void)
+{
+  init();
+  parse();
+  return 0;
+}