Refine parser

unittest/CMakeLists.txt

@@ -2,7 +2,7 @@
 
 set(UT_SRCS
     base64_unittest.cc
-    parser_unittest.cc
+    request_parser_unittest.cc
     rest_service_manager_unittest.cc
     url_unittest.cc
     )

unittest/request_parser_unittest.cc

@@ -2,10 +2,6 @@
 
 #include "webcc/request.h"
 #include "webcc/request_parser.h"
-#include "webcc/response.h"
-#include "webcc/response_parser.h"
-
-#include <iostream>
 
 // -----------------------------------------------------------------------------
 
@@ -30,7 +26,7 @@ protected:
     EXPECT_EQ("Close", request_.GetHeader("Connection"));
 
     EXPECT_EQ("", request_.content());
-    EXPECT_EQ(0, request_.content_length());
+    EXPECT_EQ(webcc::kInvalidLength, request_.content_length());
   }
 
   std::string payload_;
@@ -39,7 +35,7 @@ protected:
   webcc::RequestParser parser_;
 };
 
-TEST_F(GetRequestParserTest, ParseFullDataOnce) {
+TEST_F(GetRequestParserTest, ParseOnce) {
   bool ok = parser_.Parse(payload_.data(), payload_.size());
 
   EXPECT_TRUE(ok);
@@ -125,7 +121,7 @@ protected:
   webcc::RequestParser parser_;
 };
 
-TEST_F(PostRequestParserTest, ParseFullDataOnce) {
+TEST_F(PostRequestParserTest, ParseOnce) {
   bool ok = parser_.Parse(payload_.data(), payload_.size());
 
   EXPECT_TRUE(ok);
@@ -213,7 +209,7 @@ protected:
   webcc::RequestParser parser_;
 };
 
-TEST_F(MultipartRequestParserTest, ParseFullDataOnce) {
+TEST_F(MultipartRequestParserTest, ParseOnce) {
   bool ok = parser_.Parse(payload_.data(), payload_.size());
 
   EXPECT_TRUE(ok);

webcc/message.h

@@ -34,6 +34,10 @@ public:
     return content_length_;
   }
 
+  void set_content_length(std::size_t content_length) {
+    content_length_ = content_length;
+  }
+
   const std::string& content() const {
     return content_;
   }

webcc/parser.cc

@@ -139,54 +139,40 @@ bool Parser::GetNextLine(std::size_t off, std::string* line, bool erase) {
 bool Parser::ParseHeaderLine(const std::string& line) {
   Header header;
   if (!Split2(line, ':', &header.first, &header.second)) {
+    LOG_ERRO("Invalid header: %s", line.c_str());
     return false;
   }
 
-  do {
-    if (!chunked_ && !content_length_parsed_) {
-      if (boost::iequals(header.first, headers::kContentLength)) {
-        content_length_parsed_ = true;
+  if (boost::iequals(header.first, headers::kContentLength)) {
+    content_length_parsed_ = true;
 
-        if (!StringToSizeT(header.second, 10, &content_length_)) {
-          LOG_ERRO("Invalid content length: %s.", header.second.c_str());
-          return false;
-        }
-
-        LOG_INFO("Content length: %u.", content_length_);
-
-        // Reserve memory to avoid frequent reallocation when append.
-        try {
-          content_.reserve(content_length_);
-        } catch (const std::exception& e) {
-          LOG_ERRO("Failed to reserve content memory: %s.", e.what());
-          return false;
-        }
-
-        break;
-      }
+    if (!StringToSizeT(header.second, 10, &content_length_)) {
+      LOG_ERRO("Invalid content length: %s.", header.second.c_str());
+      return false;
     }
 
-    // TODO: Replace `!chunked_` with <TransferEncodingParsed>.
-    if (!chunked_ && !content_length_parsed_) {
-      if (boost::iequals(header.first, headers::kTransferEncoding)) {
-        if (header.second == "chunked") {
-          // The content is chunked.
-          chunked_ = true;
-        }
+    LOG_INFO("Content length: %u.", content_length_);
 
-        break;
-      }
+    // Reserve memory to avoid frequent reallocation when append.
+    try {
+      content_.reserve(content_length_);
+    } catch (const std::exception& e) {
+      LOG_ERRO("Failed to reserve content memory: %s.", e.what());
+      return false;
     }
-  } while (false);
-
-  // Parse Content-Type.
-  if (boost::iequals(header.first, headers::kContentType)) {
+  } else if (boost::iequals(header.first, headers::kContentType)) {
     ContentType content_type(header.second);
     if (!content_type.Valid()) {
       LOG_ERRO("Invalid content-type header: %s", header.second.c_str());
       return false;
+    } else {
+      message_->SetContentType(content_type);
+    }
+  } else if (boost::iequals(header.first, headers::kTransferEncoding)) {
+    if (header.second == "chunked") {
+      // The content is chunked.
+      chunked_ = true;
     }
-    message_->SetContentType(content_type);
   }
 
   message_->SetHeader(std::move(header));
@@ -220,7 +206,7 @@ bool Parser::ParseFixedContent(const char* data, std::size_t length) {
     pending_data_.clear();
   }
 
-  // NOTE: Don't have to firstly put the data to the pending data.
+  // Don't have to firstly put the data to the pending data.
   AppendContent(data, length);
 
   if (IsContentFull()) {
@@ -232,13 +218,8 @@ bool Parser::ParseFixedContent(const char* data, std::size_t length) {
 }
 
 bool Parser::ParseChunkedContent(const char* data, std::size_t length) {
-  // Append the new data to the pending data.
-  // NOTE: It's more difficult to avoid this than fixed-length content.
   pending_data_.append(data, length);
 
-  LOG_VERB("Parse chunked content (pending data size: %u).",
-           pending_data_.size());
-
   while (true) {
     // Read chunk-size if necessary.
     if (chunk_size_ == kInvalidLength) {
@@ -321,6 +302,9 @@ bool Parser::Finish() {
     return true;
   }
 
+  // Could be kInvalidLength when chunked.
+  message_->set_content_length(content_length_);
+
   if (!IsContentCompressed()) {
     message_->SetContent(std::move(content_), false);
     return true;

webcc/parser.h

@@ -59,6 +59,7 @@ protected:
   void AppendContent(const char* data, std::size_t count);
   void AppendContent(const std::string& data);
 
+  // TODO: Rename to IsFixedContentFull.
   bool IsContentFull() const;
 
   // Check header Content-Encoding to see if the content is compressed.

webcc/request_parser.cc

@@ -48,13 +48,8 @@ bool RequestParser::ParseContent(const char* data, std::size_t length) {
 
 bool RequestParser::ParseMultipartContent(const char* data,
                                           std::size_t length) {
-  // Append the new data to the pending data.
-  // NOTE: It's more difficult to avoid this than normal fixed-length content.
   pending_data_.append(data, length);
 
-  LOG_VERB("Parse multipart content (3pending data size: %u).",
-           pending_data_.size());
-
   if (!content_length_parsed_ || content_length_ == kInvalidLength) {
     // Invalid content length (syntax error).
     return false;
@@ -71,7 +66,7 @@ bool RequestParser::ParseMultipartContent(const char* data,
       if (!GetNextLine(0, &line, true)) {
         break;  // Not enough data
       }
-      if (!IsBoundary(line)) {
+      if (!IsBoundary(line, 0, line.size())) {
         LOG_ERRO("Invalid boundary: %s", line.c_str());
         return false;
       }
@@ -105,12 +100,13 @@ bool RequestParser::ParseMultipartContent(const char* data,
       std::size_t off = 0;
       std::size_t count = 0;
       bool ended = false;
+      // TODO: Remember last CRLF position.
       if (!GetNextBoundaryLine(&off, &count, &ended)) {
         // Wait until next boundary.
         break;
       }
 
-      LOG_INFO("Next boundary has been found.");
+      LOG_INFO("Next boundary found.");
 
       // This part has ended.
       if (off > 2) {
@@ -218,40 +214,40 @@ bool RequestParser::GetNextBoundaryLine(std::size_t* b_off,
       continue;  // Empty line
     }
 
-    // TODO: Avoid temp string.
-    std::string line = pending_data_.substr(off, count);
-
-    if (IsBoundary(line)) {
+    if (IsBoundary(pending_data_, off, count, ended)) {
       *b_off = off;
       *b_count = count;
       return true;
     }
 
-    if (IsBoundaryEnd(line)) {
-      *b_off = off;
-      *b_count = count;
-      *ended = true;
-      return true;
-    }
-
     off = pos + 2;
   }
 
   return false;
 }
 
-bool RequestParser::IsBoundary(const std::string& line) const {
-  if (line == "--" + request_->content_type().boundary()) {
-    return true;
+bool RequestParser::IsBoundary(const std::string& str, std::size_t off,
+                               std::size_t count, bool* end) const {
+  const std::string& boundary = request_->content_type().boundary();
+
+  if (count != boundary.size() + 2 && count != boundary.size() + 4) {
+    return false;
+  }
+
+  if (str[off] != '-' || str[off + 1] != '-') {
+    return false;
   }
-  return false;
-}
 
-bool RequestParser::IsBoundaryEnd(const std::string& line) const {
-  if (line == "--" + request_->content_type().boundary() + "--") {
-    return true;
+  if (count == boundary.size() + 4) {
+    if (str[off + count - 1] != '-' || str[off + count - 2] != '-') {
+      return false;
+    }
+    if (end != nullptr) {
+      *end = true;
+    }
   }
-  return false;
+  
+  return strncmp(boundary.c_str(), &str[off + 2], boundary.size()) == 0;
 }
 
 }  // namespace webcc

webcc/request_parser.h

@@ -29,8 +29,10 @@ private:
   bool ParsePartHeaders(bool* need_more_data);
   bool GetNextBoundaryLine(std::size_t* b_off, std::size_t* b_count,
                            bool* ended);
-  bool IsBoundary(const std::string& line) const;
-  bool IsBoundaryEnd(const std::string& line) const;
+
+  // Check if the str.substr(off, count) is a boundary.
+  bool IsBoundary(const std::string& str, std::size_t off,
+                  std::size_t count, bool* end = nullptr) const;
 
 private:
   Request* request_;