Add support for description-type package manifest value

6 files changed, 289 insertions, 50 deletions

diff --git a/mod/buildfile b/mod/buildfile
index bd68bd0..ffa9031 100644
--- a/mod/buildfile
+++ b/mod/buildfile
@@ -11,7 +11,9 @@ mod{*}:
   install = libexec/
 }
 
-import libs  = libodb%lib{odb}
+import libs  = libcmark-gfm%lib{cmark-gfm}
+import libs += libcmark-gfm-extensions%lib{cmark-gfm-extensions}
+import libs += libodb%lib{odb}
 import libs += libodb-pgsql%lib{odb-pgsql}
 import libs += libbutl%lib{butl}
 import libs += libbpkg%lib{bpkg}
diff --git a/mod/mod-package-details.cxx b/mod/mod-package-details.cxx
index 28370c2..fa073b5 100644
--- a/mod/mod-package-details.cxx
+++ b/mod/mod-package-details.cxx
@@ -186,13 +186,17 @@ handle (request& rq, response& rs)
     if (const optional<string>& d = pkg->description)
     {
       const string id ("description");
+      const string what (name.string () + " description");
 
       s << (full
-            ? PRE_TEXT (*d, id)
-            : PRE_TEXT (*d,
+            ? DIV_TEXT (*d, *pkg->description_type, id, what, error)
+            : DIV_TEXT (*d,
+                        *pkg->description_type,
                         options_->package_description (),
                         url (!full, squery, page, id),
-                        id));
+                        id,
+                        what,
+                        error));
     }
 
     s << TABLE(CLASS="proplist", ID="package")
@@ -221,7 +225,11 @@ handle (request& rq, response& rs)
     package_db_->query_value<package_count> (
       search_params<package_count> (squery, tenant, name)));
 
-  s << FORM_SEARCH (squery, "q")
+  // Let's disable autofocus in the full page mode since clicking the full or
+  // more link the user most likely intends to read rather than search, while
+  // autofocus scrolls the page to the search field.
+  //
+  s << FORM_SEARCH (squery, "q", !full)
     << DIV_COUNTER (pkg_count, "Version", "Versions");
 
   // Enclose the subsequent tables to be able to use nth-child CSS selector.
diff --git a/mod/mod-package-version-details.cxx b/mod/mod-package-version-details.cxx
index 4e09c80..82221b4 100644
--- a/mod/mod-package-version-details.cxx
+++ b/mod/mod-package-version-details.cxx
@@ -184,13 +184,17 @@ handle (request& rq, response& rs)
   if (const optional<string>& d = pkg->description)
   {
     const string id ("description");
+    const string what (title + " description");
 
     s << (full
-          ? PRE_TEXT (*d, id)
-          : PRE_TEXT (*d,
+          ? DIV_TEXT (*d, *pkg->description_type, id, what, error)
+          : DIV_TEXT (*d,
+                      *pkg->description_type,
                       options_->package_description (),
                       url (!full, id),
-                      id));
+                      id,
+                      what,
+                      error));
   }
 
   const repository_location& rl (pkg->internal_repository.load ()->location);
diff --git a/mod/mod-repository-root.cxx b/mod/mod-repository-root.cxx
index fc13aca..ed170c9 100644
--- a/mod/mod-repository-root.cxx
+++ b/mod/mod-repository-root.cxx
@@ -6,6 +6,8 @@
 
 #include <time.h> // tzset()
 
+#include <cmark-gfm-core-extensions.h>
+
 #include <sstream>
 #include <algorithm> // find()
 
@@ -285,6 +287,10 @@ namespace brep
     // sub-handlers, while handling requests).
     //
     tzset ();
+
+    // To recognize cmark-gfm extensions while parsing Markdown later on.
+    //
+    cmark_gfm_core_extensions_ensure_registered ();
   }
 
   bool repository_root::
diff --git a/mod/page.cxx b/mod/page.cxx
index f780b95..63bb495 100644
--- a/mod/page.cxx
+++ b/mod/page.cxx
@@ -4,6 +4,9 @@
 
 #include <mod/page.hxx>
 
+#include <cmark-gfm.h>
+#include <cmark-gfm-extension_api.h>
+
 #include <set>
 #include <ios>       // hex, uppercase, right
 #include <sstream>
@@ -13,6 +16,7 @@
 #include <libstudxml/serializer.hxx>
 
 #include <web/xhtml.hxx>
+#include <web/xhtml-fragment.hxx>
 #include <web/mime-url-encoding.hxx>
 
 #include <libbrep/package.hxx>
@@ -98,8 +102,12 @@ namespace brep
       <<     TBODY
       <<       TR
       <<         TD(ID="search-txt")
-      <<           *INPUT(TYPE="search", NAME=name_, VALUE=query_,
-                          AUTOFOCUS="")
+      <<           INPUT(TYPE="search", NAME=name_, VALUE=query_);
+
+    if (autofocus_)
+      s << AUTOFOCUS("");
+
+    s <<           ~INPUT
       <<         ~TD
       <<         TD(ID="search-btn")
       <<           *INPUT(TYPE="submit", VALUE="Search")
@@ -805,34 +813,183 @@ namespace brep
 
   // PRE_TEXT
   //
-  void PRE_TEXT::
-  operator() (serializer& s) const
+  static void
+  serialize_pre_text (serializer& s,
+                      const string& text,
+                      size_t length,
+                      const string* url,
+                      const string& id)
   {
-    if (text_.empty ())
+    if (text.empty ())
       return;
 
-    size_t n (text_.find_first_of (" \t\n", length_));
+    size_t n (text.find_first_of (" \t\n", length));
     bool full (n == string::npos); // Text length is below the limit.
 
     // Truncate the text if length exceeds the limit.
     //
-    const string& t (full ? text_ : string (text_, 0, n));
+    const string& t (full ? text : string (text, 0, n));
     s << PRE;
 
-    if (!id_.empty ())
-      s << ID(id_);
+    if (!id.empty ())
+      s << ID(id);
 
     s << t;
 
     if (!full)
     {
-      assert (url_ != nullptr);
-      s << "... " << A(HREF=*url_) << "More" << ~A;
+      assert (url != nullptr);
+      s << "... " << A(HREF=*url) << "More" << ~A;
     }
 
     s << ~PRE;
   }
 
+  void PRE_TEXT::
+  operator() (serializer& s) const
+  {
+    serialize_pre_text (s, text_, length_, url_, id_);
+  }
+
+  // DIV_TEXT
+  //
+  void DIV_TEXT::
+  operator() (serializer& s) const
+  {
+    switch (type_)
+    {
+    case text_type::plain:
+      {
+        // To keep things regular we wrap the preformatted text into <div>.
+        //
+        s << DIV(ID=id_, CLASS="plain");
+        serialize_pre_text (s, text_, length_, url_, "" /* id */);
+        s << ~DIV;
+        break;
+      }
+    case text_type::common_mark:
+    case text_type::github_mark:
+      {
+        // Convert Markdown into XHTML wrapping it into the <div> element.
+        //
+        auto print_error = [&s, this] (const string& e)
+        {
+          s << DIV(ID=id_, CLASS="markdown")
+          <<   SPAN(CLASS="error") << e << ~SPAN
+          << ~DIV;
+        };
+
+        // Note that the only possible reason for the following cmark API
+        // calls to fail is the inability to allocate memory. Unfortunately,
+        // instead of reporting the failure to the caller, the API issues
+        // diagnostics to stderr and aborts the process. Let's decrease the
+        // probability of such an event by limiting the text size to 64K.
+        //
+        if (text_.size () > 64 * 1024)
+        {
+          print_error (what_ + " is too long");
+          return;
+        }
+
+        string html;
+        {
+          char* r;
+          {
+            // Parse Markdown into the AST.
+            //
+            unique_ptr<cmark_parser, void (*)(cmark_parser*)> parser (
+              cmark_parser_new (CMARK_OPT_DEFAULT | CMARK_OPT_VALIDATE_UTF8),
+              [] (cmark_parser* p) {cmark_parser_free (p);});
+
+            // Enable GitHub extensions in the parser, if requested.
+            //
+            if (type_ == text_type::github_mark)
+            {
+              auto add = [&parser] (const char* ext)
+                {
+                  cmark_syntax_extension* e (
+                    cmark_find_syntax_extension (ext));
+
+                  // Built-in extension is only expected.
+                  //
+                  assert (e != nullptr);
+
+                  cmark_parser_attach_syntax_extension (parser.get (), e);
+                };
+
+              add ("table");
+              add ("strikethrough");
+              add ("autolink");
+
+              // Somehow feels unsafe (there are some nasty warnings when
+              // upstream's tasklist.c is compiled), so let's disable for now.
+              //
+              // add ("tasklist");
+            }
+
+            cmark_parser_feed (parser.get (), text_.c_str (), text_.size ());
+
+            unique_ptr<cmark_node, void (*)(cmark_node*)> doc (
+              cmark_parser_finish (parser.get ()),
+              [] (cmark_node* n) {cmark_node_free (n);});
+
+            // Render the AST into an XHTML fragment.
+            //
+            // Note that unlike GitHub we follow the default API behavior and
+            // don't allow the raw HTML in Markdown (omitting the
+            // CMARK_OPT_UNSAFE flag). This way we can assume the rendered
+            // HTML is a well-formed XHTML fragment, which we rely upon for
+            // truncation (see below). Note that by default the renderer
+            // suppresses any HTML-alike markup and unsafe URLs (javascript:,
+            // etc).
+            //
+            r = cmark_render_html (doc.get (),
+                                   CMARK_OPT_DEFAULT,
+                                   nullptr /* extensions */);
+          }
+
+          unique_ptr<char, void (*)(char*)> deleter (
+            r,
+            [] (char* s) {cmark_get_default_mem_allocator ()->free (s);});
+
+          html = r;
+        }
+
+        // From the CommonMark Spec it follows that the resulting HTML can be
+        // assumed a well-formed XHTML fragment with all the elements having
+        // closing tags. But let's not assume this being the case (due to some
+        // library bug or similar) and handle the xml::parsing exception.
+        //
+        try
+        {
+          fragment f (html, "gfm-html", url_ == nullptr ? 0 : length_);
+
+          s << DIV(ID=id_, CLASS="markdown");
+
+          // Disable indentation not to introduce unwanted spaces.
+          //
+          s.suspend_indentation ();
+          s << f;
+          s.resume_indentation ();
+
+          if (f.truncated)
+            s << DIV(CLASS="more") << A(HREF=*url_) << "More" << ~A << ~DIV;
+
+          s << ~DIV;
+        }
+        catch (const xml::parsing& e)
+        {
+          string error ("unable to parse " + what_ + " XHTML fragment: " +
+                        e.what ());
+          diag_ << error;
+          print_error (error);
+        }
+
+        break;
+      }
+    }
+  }
+
   // DIV_PAGER
   //
   DIV_PAGER::
diff --git a/mod/page.hxx b/mod/page.hxx
index b5c62c1..1138a80 100644
--- a/mod/page.hxx
+++ b/mod/page.hxx
@@ -17,6 +17,7 @@
 #include <libbrep/build.hxx>
 #include <libbrep/package.hxx>
 
+#include <mod/diagnostics.hxx>
 #include <mod/options-types.hxx> // page_menu
 
 namespace brep
@@ -24,7 +25,7 @@ namespace brep
   // Page common building blocks.
   //
 
-  // Generates CSS link elements.
+  // Generate CSS link elements.
   //
   class CSS_LINKS
   {
@@ -39,7 +40,7 @@ namespace brep
     const dir_path& root_;
   };
 
-  // Generates page header element.
+  // Generate page header element.
   //
   class DIV_HEADER
   {
@@ -60,13 +61,16 @@ namespace brep
     const string& tenant_;
   };
 
-  // Generates package search form element with the specified query input
+  // Generate package search form element with the specified query input
   // element name.
   //
   class FORM_SEARCH
   {
   public:
-    FORM_SEARCH (const string& q, const string& n): query_ (q), name_ (n) {}
+    FORM_SEARCH (const string& q, const string& n, bool a = true)
+        : query_ (q), name_ (n), autofocus_ (a)
+    {
+    }
 
     void
     operator() (xml::serializer&) const;
@@ -74,9 +78,10 @@ namespace brep
   private:
     const string& query_;
     const string& name_;
+    bool autofocus_;
   };
 
-  // Generates counter element.
+  // Generate counter element.
   //
   // It could be redunant to distinguish between singular and plural word forms
   // if it wouldn't be so cheap in English, and phrase '1 Packages' wouldn't
@@ -97,7 +102,7 @@ namespace brep
     const char* plural_;
   };
 
-  // Generates table row element, that has the 'label: value' layout.
+  // Generate table row element, that has the 'label: value' layout.
   //
   class TR_VALUE
   {
@@ -113,7 +118,7 @@ namespace brep
     const string& value_;
   };
 
-  // Generates table row element, that has the 'label: <input type="text"/>'
+  // Generate table row element, that has the 'label: <input type="text"/>'
   // layout.
   //
   class TR_INPUT
@@ -139,7 +144,7 @@ namespace brep
     bool autofocus_;
   };
 
-  // Generates table row element, that has the 'label: <select></select>'
+  // Generate table row element, that has the 'label: <select></select>'
   // layout. Option elements are represented as a list of value/inner-text
   // pairs.
   //
@@ -162,7 +167,7 @@ namespace brep
     const vector<pair<string, string>>& options_;
   };
 
-  // Generates tenant id element.
+  // Generate tenant id element.
   //
   // Displays a link to the service page for the specified tenant.
   //
@@ -185,7 +190,7 @@ namespace brep
     const string& tenant_;
   };
 
-  // Generates package name element with an optional search criteria. The
+  // Generate package name element with an optional search criteria. The
   // search string should be url-encoded, if specified.
   //
   class TR_NAME
@@ -207,7 +212,7 @@ namespace brep
     const string& tenant_;
   };
 
-  // Generates package version element.
+  // Generate package version element.
   //
   class TR_VERSION
   {
@@ -248,7 +253,7 @@ namespace brep
     const string* tenant_;
   };
 
-  // Generates package project name element.
+  // Generate package project name element.
   //
   // Displays a link to the package search page with the project name
   // specified as a keyword.
@@ -268,7 +273,7 @@ namespace brep
     const string& tenant_;
   };
 
-  // Generates package summary element.
+  // Generate package summary element.
   //
   class TR_SUMMARY
   {
@@ -282,7 +287,7 @@ namespace brep
     const string& summary_;
   };
 
-  // Generates package license alternatives element.
+  // Generate package license alternatives element.
   //
   class TR_LICENSE
   {
@@ -296,7 +301,7 @@ namespace brep
     const license_alternatives& licenses_;
   };
 
-  // Generates package license alternatives elements. Differs from TR_LICENSE
+  // Generate package license alternatives elements. Differs from TR_LICENSE
   // by producing multiple rows instead of a single one.
   //
   class TR_LICENSES
@@ -311,7 +316,7 @@ namespace brep
     const license_alternatives& licenses_;
   };
 
-  // Generates package tags element.
+  // Generate package tags element.
   //
   class TR_TAGS
   {
@@ -340,7 +345,7 @@ namespace brep
     const string& tenant_;
   };
 
-  // Generates package dependencies element.
+  // Generate package dependencies element.
   //
   class TR_DEPENDS
   {
@@ -357,7 +362,7 @@ namespace brep
     const string& tenant_;
   };
 
-  // Generates package requirements element.
+  // Generate package requirements element.
   //
   class TR_REQUIRES
   {
@@ -371,7 +376,7 @@ namespace brep
     const requirements& requirements_;
   };
 
-  // Generates url element.
+  // Generate url element.
   //
   class TR_URL
   {
@@ -386,7 +391,7 @@ namespace brep
     const char* label_;
   };
 
-  // Generates email element.
+  // Generate email element.
   //
   class TR_EMAIL
   {
@@ -402,7 +407,7 @@ namespace brep
     const char* label_;
   };
 
-  // Generates package version priority element.
+  // Generate package version priority element.
   //
   class TR_PRIORITY
   {
@@ -416,7 +421,7 @@ namespace brep
     const priority& priority_;
   };
 
-  // Generates repository name element.
+  // Generate repository name element.
   //
   class TR_REPOSITORY
   {
@@ -433,7 +438,7 @@ namespace brep
     const string& tenant_;
   };
 
-  // Generates repository location element.
+  // Generate repository location element.
   //
   class TR_LOCATION
   {
@@ -447,7 +452,7 @@ namespace brep
     const repository_location& location_;
   };
 
-  // Generates package download URL element.
+  // Generate package download URL element.
   //
   class TR_DOWNLOAD
   {
@@ -461,7 +466,7 @@ namespace brep
     const string& url_;
   };
 
-  // Generates sha256sum element.
+  // Generate sha256sum element.
   //
   class TR_SHA256SUM
   {
@@ -475,7 +480,7 @@ namespace brep
     const string& sha256sum_;
   };
 
-  // Generates build results element.
+  // Generate build results element.
   //
   class TR_BUILD_RESULT
   {
@@ -492,7 +497,7 @@ namespace brep
     const dir_path& root_;
   };
 
-  // Generates comment element.
+  // Generate comment element.
   //
   class SPAN_COMMENT
   {
@@ -506,7 +511,7 @@ namespace brep
     const string& comment_;
   };
 
-  // Generates package build result status element.
+  // Generate package build result status element.
   //
   class SPAN_BUILD_RESULT_STATUS
   {
@@ -520,7 +525,7 @@ namespace brep
     const bbot::result_status& status_;
   };
 
-  // Generates paragraph elements converting a plain text into XHTML5 applying
+  // Generate paragraph elements converting a plain text into XHTML5 applying
   // some heuristics (see implementation for details). Truncate the text if
   // requested.
   //
@@ -551,7 +556,7 @@ namespace brep
     string id_;
   };
 
-  // Generates pre-formatted text element. Truncate the text if requested.
+  // Generate pre-formatted text element. Truncate the text if requested.
   //
   class PRE_TEXT
   {
@@ -579,7 +584,64 @@ namespace brep
     string id_;
   };
 
-  // Generates paging element.
+  // Generate a typed text element truncating it if requested. On the
+  // underlying parsing/rendering error, log it and generate the error
+  // description element instead. Note that such an error indicates an issue
+  // with the implementation, rather than with the specified text.
+  //
+  class DIV_TEXT
+  {
+  public:
+    // Generate a full text element.
+    //
+    DIV_TEXT (const string& t,
+              text_type tp,
+              const string& id,
+              const string& what,
+              const basic_mark& diag)
+        : text_ (t),
+          type_ (tp),
+          length_ (t.size ()),
+          url_ (nullptr),
+          id_ (id),
+          what_ (what),
+          diag_ (diag)
+    {
+    }
+
+    // Generate a brief text element.
+    //
+    DIV_TEXT (const string& t,
+              text_type tp,
+              size_t l,
+              const string& u,
+              const string& id,
+              const string& what,
+              const basic_mark& diag)
+        : text_ (t),
+          type_ (tp),
+          length_ (l),
+          url_ (&u),
+          id_ (id),
+          what_ (what),
+          diag_ (diag)
+    {
+    }
+
+    void
+    operator() (xml::serializer&) const;
+
+  private:
+    const string& text_;
+    text_type type_;
+    size_t length_;
+    const string* url_; // Full page url.
+    string id_;
+    const string& what_;
+    const basic_mark& diag_;
+  };
+
+  // Generate paging element.
   //
   class DIV_PAGER
   {