From 8f9a80a9ac8f353ce2cdafa23f0e5163d30d5800 Mon Sep 17 00:00:00 2001
From: Karen Arutyunov <karen@codesynthesis.com>
Date: Wed, 1 May 2019 22:32:11 +0300
Subject: Add support for description-type package manifest value

---
 web/xhtml-fragment.cxx | 49 +++++++++++++++++++++++++++++++++++++++----------
 web/xhtml-fragment.hxx | 19 ++++++++++++-------
 2 files changed, 51 insertions(+), 17 deletions(-)

(limited to 'web')

diff --git a/web/xhtml-fragment.cxx b/web/xhtml-fragment.cxx
index 4d1c178..fe8a0a7 100644
--- a/web/xhtml-fragment.cxx
+++ b/web/xhtml-fragment.cxx
@@ -20,25 +20,38 @@ namespace web
   namespace xhtml
   {
     fragment::
-    fragment (const string& text, const string& name)
+    fragment (const string& text, const string& name, size_t length)
     {
       // To parse the fragment make it a valid xml document, wrapping with the
-      // root element.
+      // root element. If requested, truncate the fragment before the
+      // first-level element when the content length limit is exceeded.
       //
       string doc ("<d>" + text + "</d>");
 
-      parser p (
-        doc.c_str (),
-        doc.size (),
-        name,
-        parser::receive_elements | parser::receive_characters |
-        parser::receive_attributes_event);
+      parser p (doc.c_str (),
+                doc.size (),
+                name,
+                parser::receive_elements   |
+                parser::receive_characters |
+                parser::receive_attributes_event);
+
+      size_t len (0);
+      size_t level (0);
 
       for (parser::event_type e: p)
       {
         switch (e)
         {
         case parser::start_element:
+          {
+            truncated = length != 0 && level == 1 && len >= length;
+
+            if (truncated)
+              break;
+
+            ++level;
+          }
+          // Fall through.
         case parser::start_attribute:
           {
             const auto& n (p.qname ());
@@ -51,6 +64,10 @@ namespace web
             break;
           }
         case parser::end_element:
+          {
+            --level;
+          }
+          // Fall through.
         case parser::end_attribute:
           {
             events_.emplace_back (e, "");
@@ -58,12 +75,25 @@ namespace web
           }
         case parser::characters:
           {
-            events_.emplace_back (e, p.value ());
+            string& s (p.value ());
+
+            assert (!events_.empty ()); // Contains root element start.
+
+            if (events_.back ().first != parser::start_attribute)
+              len += s.size ();
+
+            events_.emplace_back (e, move (s));
             break;
           }
         default:
           assert (false);
         }
+
+        if (truncated)
+        {
+          events_.emplace_back (parser::end_element, ""); // Close root.
+          break;
+        }
       }
 
       // Unwrap the fragment removing the root element events.
@@ -85,7 +115,6 @@ namespace web
             s.start_element (xmlns, e.second);
             break;
           }
-
         case parser::start_attribute:
           {
             s.start_attribute (e.second);
diff --git a/web/xhtml-fragment.hxx b/web/xhtml-fragment.hxx
index 14b9a21..fd41967 100644
--- a/web/xhtml-fragment.hxx
+++ b/web/xhtml-fragment.hxx
@@ -22,16 +22,21 @@ namespace web
     class fragment
     {
     public:
+      bool truncated = false;
+
+    public:
       fragment () = default;
 
-      // Parse string as XHTML document fragment. The fragment should be
-      // complete, in the sense that all elements should have closing tags.
-      // Elements and attributes are considered to be in the namespace of the
-      // entire XHTML document, so no namespace should be specified for them.
-      // Do not validate against XHTML vocabulary. Can throw xml::parsing
-      // exception.
+      // Parse string as an XHTML document fragment, truncating it if
+      // requested. The fragment should be complete, in the sense that all
+      // elements should have closing tags. Elements and attributes are
+      // considered to be in the namespace of the entire XHTML document, so no
+      // namespace should be specified for them. Do not validate against XHTML
+      // vocabulary. Can throw xml::parsing exception.
       //
-      fragment (const std::string& xhtml, const std::string& input_name);
+      fragment (const std::string& xhtml,
+                const std::string& input_name,
+                size_t length = 0);
 
       void
       operator() (xml::serializer&) const;
-- 
cgit v1.1