[Fetch API] Implement Request.formData and Response.formData.

third_party/WebKit/Source/modules/fetch/MultipartParser.cpp

+// Copyright 2016 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "modules/fetch/MultipartParser.h"
+
+#include "platform/HTTPNames.h"
+#include "platform/network/HTTPParsers.h"
+#include "public/platform/Platform.h"
+
+#include <algorithm>
+#include <utility>
+
+namespace blink {
+
+namespace {
+
+constexpr char kCloseDelimiterSuffix[] = "--\r\n";
+constexpr size_t kCloseDelimiterSuffixSize =
+    WTF_ARRAY_LENGTH(kCloseDelimiterSuffix) - 1u;
+constexpr size_t kDashBoundaryOffset = 2u;  // The length of "\r\n".
+constexpr char kDelimiterSuffix[] = "\r\n";
+constexpr size_t kDelimiterSuffixSize = WTF_ARRAY_LENGTH(kDelimiterSuffix) - 1u;
+
+bool octetsNeedDecoding(const HTTPHeaderMap& headerFields) {
+  const AtomicString& encoding =
+      headerFields.get(HTTPNames::Content_Transfer_Encoding);
+
+  // Decoding is not needed if a transfer encoding is not used.
+  if (encoding.isNull())
+    return false;
+
+  // Decoding is not needed if a no-op transfer encoding is used.
+  if (equalIgnoringCase(encoding, "binary") ||
+      equalIgnoringCase(encoding, "7bit") ||
+      equalIgnoringCase(encoding, "8bit")) {
+    return false;
+  }
+
+  return true;
+}
+
+}  // namespace
+
+MultipartParser::Matcher::Matcher() = default;
+
+MultipartParser::Matcher::Matcher(const char* data,
+                                  size_t numMatchedBytes,
+                                  size_t size)
+    : m_data(data), m_numMatchedBytes(numMatchedBytes), m_size(size) {}
+
+bool MultipartParser::Matcher::match(const char* first, const char* last) {
+  while (first < last) {
+    if (!match(*first++))
+      return false;
+  }
+  return true;
+}
+
+void MultipartParser::Matcher::setNumMatchedBytes(size_t numMatchedBytes) {
+  DCHECK_LE(numMatchedBytes, m_size);
+  m_numMatchedBytes = numMatchedBytes;
+}
+
+MultipartParser::MultipartParser(Vector<char> boundary, Client* client)
+    : m_client(client),
+      m_delimiter(std::move(boundary)),
+      m_state(State::ParsingPreamble) {
+  // The delimiter consists of "\r\n" and a dash boundary which consists of
+  // "--" and a boundary.
+  m_delimiter.prepend("\r\n--", 4u);
+  m_matcher = delimiterMatcher(kDashBoundaryOffset);
+}
+
+bool MultipartParser::appendData(const char* bytes, size_t size) {
+  DCHECK_NE(State::Finished, m_state);
+  DCHECK_NE(State::Cancelled, m_state);
+
+  const char* const bytesEnd = bytes + size;
+
+  while (bytes < bytesEnd) {
+    switch (m_state) {
+      case State::ParsingPreamble:
+        // Parse either a preamble and a delimiter or a dash boundary.
+        parseDelimiter(&bytes, bytesEnd);
+        if (!m_matcher.isMatchComplete() && bytes < bytesEnd) {
+          // Parse a preamble data (by ignoring it) and then a delimiter.
+          m_matcher.setNumMatchedBytes(0u);
+          parseDataAndDelimiter(&bytes, bytesEnd);
+        }
+        if (m_matcher.isMatchComplete()) {
+          // Prepare for a delimiter suffix.
+          m_matcher = delimiterSuffixMatcher();
+          m_state = State::ParsingDelimiterSuffix;
+        }
+        break;
+
+      case State::ParsingDelimiterSuffix:
+        // Parse transport padding and "\r\n" after a delimiter.
+        // This state can be reached after either a preamble or part
+        // octets are parsed.
+        if (m_matcher.numMatchedBytes() == 0u)
+          parseTransportPadding(&bytes, bytesEnd);
+        while (bytes < bytesEnd) {
+          if (!m_matcher.match(*bytes++))
+            return false;
+          if (m_matcher.isMatchComplete()) {
+            // Prepare for part header fields.
+            m_state = State::ParsingPartHeaderFields;
+            break;
+          }
+        }
+        break;
+
+      case State::ParsingPartHeaderFields: {
+        // Parse part header fields (which ends with "\r\n") and an empty
+        // line (which also ends with "\r\n").
+        // This state can be reached after a delimiter and a delimiter
+        // suffix after either a preamble or part octets are parsed.
+        HTTPHeaderMap headerFields;
+        if (parseHeaderFields(&bytes, bytesEnd, &headerFields)) {
+          // Decoding is not implemented.
+          if (octetsNeedDecoding(headerFields))

[yhirano, 2017/01/11 08:54:03]

Is this needed? With the current spec we end up ignoring the encoding in any
case, right?

[e_hakkinen, 2017/05/02 23:38:56]

Yes, let's remove this.

+            return false;
+          // Prepare for part octets.
+          m_matcher = delimiterMatcher();
+          m_state = State::ParsingPartOctets;
+          m_client->partHeaderFieldsInMultipartReceived(headerFields);
+        }
+        break;
+      }
+
+      case State::ParsingPartOctets: {
+        // Parse part octets and a delimiter.
+        // This state can be reached only after part header fields are
+        // parsed.
+        const size_t numInitiallyMatchedBytes = m_matcher.numMatchedBytes();
+        const char* octetsBegin = bytes;
+        parseDelimiter(&bytes, bytesEnd);
+        if (!m_matcher.isMatchComplete() && bytes < bytesEnd) {
+          if (m_matcher.numMatchedBytes() >= numInitiallyMatchedBytes &&
+              numInitiallyMatchedBytes > 0u) {
+            // Since the matched bytes did not form a complete
+            // delimiter, the matched bytes turned out to be octet
+            // bytes instead of being delimiter bytes. Additionally,
+            // some of the matched bytes are from the previous call and
+            // are therefore not in the range [octetsBegin, bytesEnd[.
+            m_client->partDataInMultipartReceived(m_matcher.data(),
+                                                  m_matcher.numMatchedBytes());
+            if (m_state != State::ParsingPartOctets)
+              break;
+            octetsBegin = bytes;
+          }
+          m_matcher.setNumMatchedBytes(0u);
+          parseDataAndDelimiter(&bytes, bytesEnd);
+          const char* const octetsEnd = bytes - m_matcher.numMatchedBytes();
+          if (octetsBegin < octetsEnd) {
+            m_client->partDataInMultipartReceived(
+                octetsBegin, static_cast<size_t>(octetsEnd - octetsBegin));
+            if (m_state != State::ParsingPartOctets)
+              break;
+          }
+        }
+        if (m_matcher.isMatchComplete()) {
+          m_state = State::ParsingDelimiterOrCloseDelimiterSuffix;
+          m_client->partDataInMultipartFullyReceived();
+        }
+        break;
+      }
+
+      case State::ParsingDelimiterOrCloseDelimiterSuffix:
+        // Determine whether this is a delimiter suffix or a close
+        // delimiter suffix.
+        // This state can be reached only after part octets are parsed.
+        if (*bytes == '-') {
+          // Prepare for a close delimiter suffix.
+          m_matcher = closeDelimiterSuffixMatcher();
+          m_state = State::ParsingCloseDelimiterSuffix;
+        } else {
+          // Prepare for a delimiter suffix.
+          m_matcher = delimiterSuffixMatcher();
+          m_state = State::ParsingDelimiterSuffix;
+        }
+        break;
+
+      case State::ParsingCloseDelimiterSuffix:
+        // Parse "--", transport padding and "\r\n" after a delimiter
+        // (a delimiter and "--" constitute a close delimiter).
+        // This state can be reached only after part octets are parsed.
+        for (;;) {
+          if (m_matcher.numMatchedBytes() == 2u)
+            parseTransportPadding(&bytes, bytesEnd);
+          if (bytes >= bytesEnd)
+            break;
+          if (!m_matcher.match(*bytes++))
+            return false;
+          if (m_matcher.isMatchComplete()) {
+            // Prepare for an epilogue.
+            m_state = State::ParsingEpilogue;
+            break;
+          }
+        }
+        break;
+
+      case State::ParsingEpilogue:
+        // Parse an epilogue (by ignoring it).
+        // This state can be reached only after a delimiter and a close
+        // delimiter suffix after part octets are parsed.
+        return true;
+
+      case State::Cancelled:
+      case State::Finished:
+        // The client changed the state.
+        return false;
+    }
+  }
+
+  DCHECK_EQ(bytesEnd, bytes);
+
+  return true;
+}
+
+void MultipartParser::cancel() {
+  m_state = State::Cancelled;
+}
+
+bool MultipartParser::finish() {
+  DCHECK_NE(State::Cancelled, m_state);
+  DCHECK_NE(State::Finished, m_state);
+
+  const State initialState = m_state;
+  m_state = State::Finished;
+
+  switch (initialState) {
+    case State::ParsingPartOctets:
+      if (m_matcher.numMatchedBytes() > 0u) {
+        // Since the matched bytes did not form a complete delimiter,
+        // the matched bytes turned out to be octet bytes instead of being
+        // delimiter bytes.
+        m_client->partDataInMultipartReceived(m_matcher.data(),
+                                              m_matcher.numMatchedBytes());
+      }
+      return false;
+    case State::ParsingCloseDelimiterSuffix:
+      // Require a full close delimiter consisting of a delimiter and "--"
+      // but ignore missing or partial "\r\n" after that.
+      return m_matcher.numMatchedBytes() >= 2u;
+    case State::ParsingEpilogue:
+      return true;
+    default:
+      return false;
+  }
+}
+
+MultipartParser::Matcher MultipartParser::closeDelimiterSuffixMatcher() const {
+  return Matcher(kCloseDelimiterSuffix, 0u, kCloseDelimiterSuffixSize);
+}
+
+MultipartParser::Matcher MultipartParser::delimiterMatcher(
+    size_t numAlreadyMatchedBytes) const {
+  return Matcher(m_delimiter.data(), numAlreadyMatchedBytes,
+                 m_delimiter.size());
+}
+
+MultipartParser::Matcher MultipartParser::delimiterSuffixMatcher() const {
+  return Matcher(kDelimiterSuffix, 0u, kDelimiterSuffixSize);
+}
+
+void MultipartParser::parseDataAndDelimiter(const char** bytesPointer,
+                                            const char* bytesEnd) {
+  DCHECK_EQ(0u, m_matcher.numMatchedBytes());
+
+  // Search for a complete delimiter within the bytes.
+  const char* delimiterBegin = std::search(
+      *bytesPointer, bytesEnd, m_delimiter.begin(), m_delimiter.end());
+  if (delimiterBegin != bytesEnd) {
+    // A complete delimiter was found. The bytes before that are octet
+    // bytes.
+    const char* const delimiterEnd = delimiterBegin + m_delimiter.size();
+    const bool matched = m_matcher.match(delimiterBegin, delimiterEnd);
+    DCHECK(matched);
+    DCHECK(m_matcher.isMatchComplete());
+    *bytesPointer = delimiterEnd;
+  } else {
+    // Search for a partial delimiter in the end of the bytes.
+    const size_t size = static_cast<size_t>(bytesEnd - *bytesPointer);
+    for (delimiterBegin = bytesEnd - std::min(m_delimiter.size() - 1u, size);
+         delimiterBegin < bytesEnd; ++delimiterBegin) {
+      if (m_matcher.match(delimiterBegin, bytesEnd))
+        break;
+      m_matcher.setNumMatchedBytes(0u);
+    }
+    // If a partial delimiter was found in the end of bytes, the bytes
+    // before the partial delimiter are definitely octets bytes and
+    // the partial delimiter bytes are buffered for now.
+    // If a partial delimiter was not found in the end of bytes, all bytes
+    // are definitely octets bytes.
+    // In all cases, all bytes are parsed now.
+    *bytesPointer = bytesEnd;
+  }
+
+  DCHECK(m_matcher.isMatchComplete() || *bytesPointer == bytesEnd);
+}
+
+void MultipartParser::parseDelimiter(const char** bytesPointer,
+                                     const char* bytesEnd) {
+  DCHECK(!m_matcher.isMatchComplete());
+  while (*bytesPointer < bytesEnd && m_matcher.match(*(*bytesPointer))) {
+    ++(*bytesPointer);
+    if (m_matcher.isMatchComplete())
+      break;
+  }
+}
+
+bool MultipartParser::parseHeaderFields(const char** bytesPointer,
+                                        const char* bytesEnd,
+                                        HTTPHeaderMap* headerFields) {
+  // Combine the current bytes with buffered header bytes if needed.
+  const char* headerBytes = *bytesPointer;
+  size_t headerSize = static_cast<size_t>(bytesEnd - *bytesPointer);
+  if (!m_bufferedHeaderBytes.isEmpty()) {
+    m_bufferedHeaderBytes.append(headerBytes, headerSize);
+    headerBytes = m_bufferedHeaderBytes.data();
+    headerSize = m_bufferedHeaderBytes.size();
+  }
+
+  size_t end = 0u;
+  if (!parseMultipartFormHeadersFromBody(headerBytes, headerSize, headerFields,
+                                         &end)) {
+    // Store the current header bytes for the next call unless that has
+    // already been done.
+    if (m_bufferedHeaderBytes.isEmpty())
+      m_bufferedHeaderBytes.append(headerBytes, headerSize);
+    *bytesPointer = bytesEnd;
+    return false;
+  }
+  m_bufferedHeaderBytes.clear();
+  *bytesPointer = bytesEnd - (headerSize - end);
+
+  return true;
+}
+
+void MultipartParser::parseTransportPadding(const char** bytesPointer,
+                                            const char* bytesEnd) const {
+  while (*bytesPointer < bytesEnd &&
+         (*(*bytesPointer) == '\t' || *(*bytesPointer) == ' '))
+    ++(*bytesPointer);
+}
+
+DEFINE_TRACE(MultipartParser) {
+  visitor->trace(m_client);
+}
+
+}  // namespace blink