Introduce EventSourceParser

third_party/WebKit/Source/core/page/EventSourceParser.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 "core/page/EventSourceParser.h"
+
+#include "core/EventTypeNames.h"
+#include "core/page/EventSource.h"
+#include "wtf/ASCIICType.h"
+#include "wtf/NotFound.h"
+#include "wtf/StdLibExtras.h"
+#include "wtf/text/TextCodec.h"
+#include "wtf/text/TextEncoding.h"
+#include "wtf/text/TextEncodingRegistry.h"
+
+namespace blink {
+
+namespace {
+
+String fromUTF8(const char* bytes, size_t size)
+{
+    return newTextCodec(UTF8Encoding())->decode(bytes, size, WTF::DataEOF);
+}
+
+} // namespace
+
+void EventSourceParser::addBytes(const char* bytes, size_t size)
+{
+    size_t start = 0;

[tyoshino (SeeGerritForStatus), 2016/01/28 07:31:43]

explain that start is the start position of non-BOM bytes in |bytes| which are
not yet processed.

[yhirano, 2016/01/28 08:45:36]

Done.

+    const char kBOM[] = {0xef, 0xbb, 0xbf};
+    for (size_t i = 0; i < size && !m_isStopped; ++i) {

[tyoshino (SeeGerritForStatus), 2016/01/28 07:31:43]

explain that kBOM doesn't contain either CL or LF and therefore we can scan BOM
and line in parallel.

[yhirano, 2016/01/28 08:45:36]

Done.

+        if (m_isRecognizingBOM && m_line.size() + (i - start) == WTF_ARRAY_LENGTH(kBOM)) {
+            Vector<char> line = m_line;
+            line.append(&bytes[start], i - start);
+            ASSERT(line.size() == WTF_ARRAY_LENGTH(kBOM));
+            m_isRecognizingBOM = false;
+            if (strncmp(line.data(), kBOM, WTF_ARRAY_LENGTH(kBOM)) == 0) {
+                start = i;
+                m_line.clear();
+                continue;
+            }
+        }
+        if (m_isRecognizingCRLF && bytes[i] == '\n') {
+            // This is the latter part of "\r\n".
+            m_isRecognizingCRLF = false;
+            ++start;
+            continue;
+        }
+        m_isRecognizingCRLF = false;
+        if (bytes[i] == '\r' || bytes[i] == '\n') {
+            m_line.append(&bytes[start], i - start);
+            parseLine();
+            m_line.clear();
+            start = i + 1;
+            m_isRecognizingCRLF = bytes[i] == '\r';
+            m_isRecognizingBOM = false;
+        }
+    }
+    if (m_isStopped)
+        return;
+    m_line.append(&bytes[start], size - start);
+}
+
+void EventSourceParser::parseLine()
+{
+    if (m_line.size() == 0) {
+        // We dispatch an event when seeing an empty line.
+        if (!m_data.isEmpty()) {
+            ASSERT(m_data[m_data.size() - 1] == '\n');
+            m_lastEventId = m_id;
+            String data = fromUTF8(m_data.data(), m_data.size() - 1);
+            m_client->onMessageEvent(m_event.isEmpty() ? EventTypeNames::message : m_event, data, m_lastEventId);
+            m_data.clear();
+        }
+        m_event = AtomicString();

[tyoshino (SeeGerritForStatus), 2016/01/28 07:31:43]

use nullAtom?

[yhirano, 2016/01/28 08:45:36]

Done.

+        return;
+    }
+    size_t fieldNameEnd = m_line.find(':');
+    size_t fieldValueStart;
+    if (fieldNameEnd == WTF::kNotFound) {
+        fieldNameEnd = m_line.size();
+        fieldValueStart = fieldNameEnd;
+    } else {
+        fieldValueStart = fieldNameEnd + 1;
+        if (fieldValueStart < m_line.size() && m_line[fieldValueStart] == ' ') {
+            ++fieldValueStart;
+        }
+    }
+    size_t fieldValueSize = m_line.size() - fieldValueStart;
+    String fieldName = fromUTF8(m_line.data(), fieldNameEnd);
+    if (fieldName == "event") {
+        m_event = AtomicString(fromUTF8(m_line.data() + fieldValueStart, fieldValueSize));
+    } else if (fieldName == "data") {
+        m_data.append(m_line.data() + fieldValueStart, fieldValueSize);
+        m_data.append('\n');
+    } else if (fieldName == "id") {
+        m_id = AtomicString(fromUTF8(m_line.data() + fieldValueStart, fieldValueSize));
+    } else if (fieldName == "retry") {
+        bool hasOnlyDigits = true;
+        for (size_t i = fieldValueStart; i < m_line.size() && hasOnlyDigits; ++i)
+            hasOnlyDigits = isASCIIDigit(m_line[i]);
+        if (fieldValueStart == m_line.size()) {
+            m_client->onReconnectionTimeSet(EventSource::defaultReconnectDelay);
+        } else if (hasOnlyDigits) {
+            bool ok;
+            auto reconnectionTime = fromUTF8(m_line.data() + fieldValueStart, fieldValueSize).toUInt64(&ok);

[tyoshino (SeeGerritForStatus), 2016/01/28 07:31:43]

can we use charactersToUInt64() directly?

[yhirano, 2016/01/28 08:45:36]

charactersToUint64 expects LChar* or UChar* and we have char*.

+            if (ok)
+                m_client->onReconnectionTimeSet(reconnectionTime);
+        }
+    } else {
+        // Unrecognized field name. Ignore!
+    }
+}
+
+DEFINE_TRACE(EventSourceParser)
+{
+    visitor->trace(m_client);
+}
+
+} // namespace blink
+