Enter uriEncode()

src/tools/stringTools.cpp

@@ -208,6 +208,43 @@ bool isHarmlessUriChar(char c)
   return false;
 }
 
+bool mustBeUriEncodedFor(kiwix::URIComponentKind target, char c)
+{
+  if (isHarmlessUriChar(c))
+    return false;
+
+  switch (c) {
+  case '/': // There is no reason to encode the path separator in the general
+            // case. It must be encoded only in a path component when its
+            // semantics of a path separator has to be suppressed.
+            return false;
+
+  case '@': // In a relative URL of the form abc@def/xyz (with no / in abc)
+            // a non-encoded @ will make "abc" and "def" to be interpreted as
+            // username and host components, respectively
+            return target == kiwix::URIComponentKind::PATH;
+
+  case ':': // In a relative URL of the form abc:def/xyz (with no / in abc)
+            // a non-encoded : will make "abc" and "def" to be interpreted as
+            // host and port components, respectively
+            return target == kiwix::URIComponentKind::PATH;
+
+  case '?': // A non-encoded '?' acts as a separator between the path
+            // and query components
+            return target == kiwix::URIComponentKind::PATH;
+
+  case '&': return target == kiwix::URIComponentKind::QUERY;
+  case '=': return target == kiwix::URIComponentKind::QUERY;
+  case '+': return target == kiwix::URIComponentKind::QUERY;
+
+  case '#': // A non-encoded '#' in either path or query-component
+            // would mark the beginning of the fragment component
+            return true;
+  }
+
+  return true;
+}
+
 int hexToInt(char c) {
   switch (c) {
   case '0': return 0;
@@ -247,6 +284,26 @@ std::string kiwix::urlEncode(const std::string& value)
   return os.str();
 }
 
+namespace kiwix
+{
+
+std::string uriEncode(URIComponentKind target, const std::string& value)
+{
+  std::ostringstream os;
+  os << std::hex << std::uppercase;
+  for (const char c : value) {
+    if ( mustBeUriEncodedFor(target, c)  ) {
+      const unsigned int charVal = static_cast<unsigned char>(c);
+      os << '%' << std::setw(2) << std::setfill('0') << charVal;
+    } else {
+      os << c;
+    }
+  }
+  return os.str();
+}
+
+} // namespace kiwix
+
 std::string kiwix::urlDecode(const std::string& value, bool component)
 {
   std::ostringstream os;

src/tools/stringTools.h

@@ -60,6 +60,17 @@ private:
 std::string urlEncode(const std::string& value);
 std::string urlDecode(const std::string& value, bool component = false);
 
+// Only URI components that are of interest to libkiwix
+// are included in the below enumeration type
+enum class URIComponentKind
+{
+  PATH,
+  QUERY
+};
+
+// Encode 'value' for usage in a URI componenet specified by 'target'
+std::string uriEncode(URIComponentKind target, const std::string& value);
+
 std::string join(const std::vector<std::string>& list, const std::string& sep);
 
 std::string ucAll(const std::string& word);

test/stringTools.cpp

@@ -163,4 +163,35 @@ TEST(stringTools, urlDecode)
   EXPECT_EQ(urlDecode(encodedUriDelimSymbols, false), encodedUriDelimSymbols);
 }
 
-};
+TEST(stringTools, uriEncode)
+{
+  const char letters[] = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ";
+  EXPECT_EQ(uriEncode(URIComponentKind::PATH,  letters), letters);
+  EXPECT_EQ(uriEncode(URIComponentKind::QUERY, letters), letters);
+
+  const char digits[] = "0123456789";
+  EXPECT_EQ(uriEncode(URIComponentKind::PATH,  digits), digits);
+  EXPECT_EQ(uriEncode(URIComponentKind::QUERY, digits), digits);
+
+  const char nonEncodableSymbols[] = ".-_~()*!/";
+  EXPECT_EQ(uriEncode(URIComponentKind::PATH,  nonEncodableSymbols), nonEncodableSymbols);
+  EXPECT_EQ(uriEncode(URIComponentKind::QUERY, nonEncodableSymbols), nonEncodableSymbols);
+
+  const char uriDelimSymbols[] = ":@?=+&#$;,";
+  EXPECT_EQ(uriEncode(URIComponentKind::PATH,  uriDelimSymbols), "%3A%40%3F=+&%23%24%3B%2C");
+  EXPECT_EQ(uriEncode(URIComponentKind::QUERY, uriDelimSymbols), ":@?%3D%2B%26%23%24%3B%2C");
+
+  const char otherSymbols[] = R"(`%^[]{}\|"<>)";
+  EXPECT_EQ(uriEncode(URIComponentKind::PATH, otherSymbols), "%60%25%5E%5B%5D%7B%7D%5C%7C%22%3C%3E");
+  EXPECT_EQ(uriEncode(URIComponentKind::PATH, otherSymbols), uriEncode(URIComponentKind::QUERY, otherSymbols));
+
+  const char whitespace[] = " \n\t\r";
+  EXPECT_EQ(uriEncode(URIComponentKind::PATH, whitespace), "%20%0A%09%0D");
+  EXPECT_EQ(uriEncode(URIComponentKind::PATH, whitespace), uriEncode(URIComponentKind::QUERY, whitespace));
+
+  const char someNonASCIIChars[] = "Σ♂♀ツ";
+  EXPECT_EQ(uriEncode(URIComponentKind::PATH, someNonASCIIChars), "%CE%A3%E2%99%82%E2%99%80%E3%83%84");
+  EXPECT_EQ(uriEncode(URIComponentKind::PATH, someNonASCIIChars), uriEncode(URIComponentKind::QUERY, someNonASCIIChars));
+}
+
+} // unnamed namespace