From: Pavel Charvat <pavel.charvat@netcentrum.cz>
Date: Thu, 24 Apr 2008 07:07:20 +0000 (+0200)
Subject: XML: Implemented detection and validation of ignorable whitespace
X-Git-Tag: holmes-import~449^2~8
X-Git-Url: http://mj.ucw.cz/gitweb/?a=commitdiff_plain;ds=sidebyside;h=150d10693b285ce5ff296571100b23b884f30560;p=libucw.git

XML: Implemented detection and validation of ignorable whitespace
---

diff --git a/sherlock/xml/parse.c b/sherlock/xml/parse.c
index 0d62112a..67ef1076 100644
--- a/sherlock/xml/parse.c
+++ b/sherlock/xml/parse.c
@@ -421,9 +421,16 @@ xml_flush_chars(struct xml_context *ctx)
   uns len = xml_end_chars(ctx, &text), rlen;
   if (len)
     {
+      if (ctx->flags & XML_NO_CHARS)
+        {
+          if ((ctx->flags & XML_REPORT_CHARS) && ctx->h_ignorable)
+            ctx->h_ignorable(ctx, text, len);
+	  mp_restore(ctx->pool, &ctx->chars_state);
+	  return 0;
+	}
       if ((ctx->flags & XML_REPORT_CHARS) && ctx->h_block && (rlen = xml_report_chars(ctx, &rtext)))
 	ctx->h_block(ctx, rtext, rlen);
-      if (!(ctx->flags & XML_ALLOC_CHARS) && !(ctx->flags & XML_REPORT_CHARS) && !ctx->h_chars)
+      if (!(ctx->flags & XML_ALLOC_CHARS) && !(ctx->flags & XML_REPORT_CHARS))
         {
 	  mp_restore(ctx->pool, &ctx->chars_state);
 	  return 0;
@@ -450,25 +457,51 @@ xml_append_chars(struct xml_context *ctx)
 {
   TRACE(ctx, "append_chars");
   struct fastbuf *out = &ctx->chars;
-  while (xml_get_char(ctx) != '<')
-    if (xml_last_char(ctx) == '&')
-      {
-	xml_inc(ctx);
-        xml_parse_ref(ctx);
-      }
-    else
-      bput_utf8_32(out, xml_last_char(ctx));
+  if (ctx->flags & XML_NO_CHARS)
+    while (xml_get_char(ctx) != '<')
+      if (xml_last_cat(ctx) & XML_CHAR_WHITE)
+	bput_utf8_32(out, xml_last_char(ctx));
+      else
+        {
+	  xml_error(ctx, "This element must not contain character data");
+	  while (xml_get_char(ctx) != '<');
+	  break;
+	}
+  else
+    while (xml_get_char(ctx) != '<')
+      if (xml_last_char(ctx) == '&')
+        {
+	  xml_inc(ctx);
+          xml_parse_ref(ctx);
+        }
+      else
+        bput_utf8_32(out, xml_last_char(ctx));
   xml_unget_char(ctx);
 }
 
 /*** CDATA sections ***/
 
+static void
+xml_skip_cdata(struct xml_context *ctx)
+{
+  TRACE(ctx, "skip_cdata");
+  xml_parse_seq(ctx, "CDATA[");
+  while (xml_get_char(ctx) != ']' || xml_get_char(ctx) != ']' || xml_get_char(ctx) != '>');
+  xml_dec(ctx);
+}
+
 static void
 xml_append_cdata(struct xml_context *ctx)
 {
   /* CDSect :== '<![CDATA[' (Char* - (Char* ']]>' Char*)) ']]>'
    * Already parsed: '<![' */
   TRACE(ctx, "append_cdata");
+  if (ctx->flags & XML_NO_CHARS)
+    {
+      xml_error(ctx, "This element must not contain CDATA");
+      xml_skip_cdata(ctx);
+      return;
+    }
   xml_parse_seq(ctx, "CDATA[");
   struct fastbuf *out = &ctx->chars;
   uns rlen;
@@ -493,15 +526,6 @@ xml_append_cdata(struct xml_context *ctx)
   xml_dec(ctx);
 }
 
-static void UNUSED
-xml_skip_cdata(struct xml_context *ctx)
-{
-  TRACE(ctx, "skip_cdata");
-  xml_parse_seq(ctx, "CDATA[");
-  while (xml_get_char(ctx) != ']' || xml_get_char(ctx) != ']' || xml_get_char(ctx) != '>');
-  xml_dec(ctx);
-}
-
 /*** Attribute values ***/
 
 char *
@@ -669,6 +693,11 @@ xml_push_element(struct xml_context *ctx)
     xml_error(ctx, "Undefined element <%s>", e->name);
   else
     {
+      if (e->dtd->type == XML_DTD_ELEM_MIXED)
+        ctx->flags &= ~XML_NO_CHARS;
+      else
+	ctx->flags |= XML_NO_CHARS;
+
       // FIXME: validate regular expressions
     }
   while (1)
diff --git a/sherlock/xml/xml-test.c b/sherlock/xml/xml-test.c
index 972a339b..0c685d8f 100644
--- a/sherlock/xml/xml-test.c
+++ b/sherlock/xml/xml-test.c
@@ -23,6 +23,7 @@ enum {
   WANT_IGNORE_COMMENTS,
   WANT_IGNORE_PIS,
   WANT_REPORT_BLOCKS,
+  WANT_REPORT_IGNORABLE,
   WANT_FILE_ENTITIES,
 };
 
@@ -36,7 +37,8 @@ static struct option longopts[] = {
   { "hide-errors",	0, 0, WANT_HIDE_ERRORS },
   { "ignore-comments",	0, 0, WANT_IGNORE_COMMENTS },
   { "ignore-pis",	0, 0, WANT_IGNORE_PIS },
-  { "reports-blocks",	0, 0, WANT_REPORT_BLOCKS },
+  { "report-blocks",	0, 0, WANT_REPORT_BLOCKS },
+  { "report-ignorable",	0, 0, WANT_REPORT_IGNORABLE },
   { "file-entities",	0, 0, WANT_FILE_ENTITIES },
   { NULL,		0, 0, 0 }
 };
@@ -58,6 +60,7 @@ CF_USAGE
     --ignore-comments   Ignore comments\n\
     --ignore-pis        Ignore processing instructions\n\
     --report-blocks	Report blocks or characters and CDATA sections\n\
+    --report-ignorable  Report ignorable whitespace\n\
     --file-entities     Resolve file external entities (not fully normative)\n\
 \n", stderr);
   exit(1);
@@ -71,6 +74,7 @@ static uns want_hide_errors;
 static uns want_ignore_comments;
 static uns want_ignore_pis;
 static uns want_report_blocks;
+static uns want_report_ignorable;
 static uns want_file_entities;
 
 static struct fastbuf *out;
@@ -206,6 +210,12 @@ h_cdata(struct xml_context *ctx UNUSED, char *text, uns len UNUSED)
   bprintf(out, "SAX:  cdata text='%s'\n", text);
 }
 
+static void
+h_ignorable(struct xml_context *ctx UNUSED, char *text, uns len UNUSED)
+{
+  bprintf(out, "SAX:  ignorable text='%s'\n", text);
+}
+
 static void
 h_dtd_start(struct xml_context *ctx UNUSED)
 {
@@ -257,6 +267,9 @@ main(int argc, char **argv)
 	case WANT_REPORT_BLOCKS:
 	  want_report_blocks++;
 	  break;
+	case WANT_REPORT_IGNORABLE:
+	  want_report_ignorable++;
+	  break;
 	case WANT_FILE_ENTITIES:
 	  want_file_entities++;
 	  break;
@@ -287,6 +300,8 @@ main(int argc, char **argv)
           ctx.h_block = h_block;
           ctx.h_cdata = h_cdata;
 	}
+      if (want_report_ignorable)
+        ctx.h_ignorable = h_ignorable;
       ctx.h_dtd_start = h_dtd_start;
       ctx.h_dtd_end = h_dtd_end;
     }
diff --git a/sherlock/xml/xml.h b/sherlock/xml/xml.h
index ac9ebefb..01b115da 100644
--- a/sherlock/xml/xml.h
+++ b/sherlock/xml/xml.h
@@ -80,6 +80,7 @@ enum xml_flags {
   /* Other parameters */
   XML_VALIDATING =			0x00000100,	/* Validate everything (not fully implemented!) */
   XML_PARSE_DTD =			0x00000200,	/* Enable parsing of DTD */
+  XML_NO_CHARS =			0x00000400,	/* The current element must not contain character data (filled automaticaly if using DTD) */
 
   /* Internals, do not change! */
   XML_EMPTY_ELEM_TAG =			0x00010000,	/* The current element match EmptyElemTag */
@@ -193,6 +194,7 @@ struct xml_context {
   void (*h_chars)(struct xml_context *ctx);		/* Called after some characters (only with XML_REPORT_CHARS) */
   void (*h_block)(struct xml_context *ctx, char *text, uns len);	/* Called for each continuous block of characters not reported by h_cdata() (only with XML_REPORT_CHARS) */
   void (*h_cdata)(struct xml_context *ctx, char *text, uns len);	/* Called for each CDATA section (only with XML_REPORT_CHARS) */
+  void (*h_ignorable)(struct xml_context *ctx, char *text, uns len);	/* Called for ignorable whitespace (content in tags without #PCDATA) */
   void (*h_dtd_start)(struct xml_context *ctx);		/* Called just after the DTD structure is initialized */
   void (*h_dtd_end)(struct xml_context *ctx);		/* Called after DTD subsets subsets */
   struct xml_dtd_entity *(*h_find_entity)(struct xml_context *ctx, char *name);		/* Called when needed to resolve a general entity */