Modularize Compact Language Detector 2 (CLD2) data sources

components/translate/content/browser/data_file_browser_cld_data_provider.cc

Index: components/translate/content/browser/data_file_browser_cld_data_provider.cc
diff --git a/components/translate/content/browser/data_file_browser_cld_data_provider.cc b/components/translate/content/browser/data_file_browser_cld_data_provider.cc
new file mode 100644
index 0000000000000000000000000000000000000000..42abf80dd36356fe0294d720595c9d1ee729d9b2
--- /dev/null
+++ b/components/translate/content/browser/data_file_browser_cld_data_provider.cc
@@ -0,0 +1,191 @@
+// Copyright 2014 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 "data_file_browser_cld_data_provider.h"
+
+#include "base/basictypes.h"
+#include "base/files/file.h"
+#include "base/lazy_instance.h"
+#include "base/memory/weak_ptr.h"
+#include "base/path_service.h"
+#include "base/synchronization/lock.h"
+#include "base/task_runner.h"
+#include "chrome/common/chrome_constants.h"
+#include "chrome/common/chrome_paths.h"
+#include "components/translate/content/common/data_file_cld_data_provider_messages.h"
+#include "content/public/browser/browser_thread.h"
+#include "content/public/browser/render_process_host.h"
+#include "ipc/ipc_message_macros.h"
+#include "ipc/ipc_platform_file.h"
+
+#if defined(CLD2_IS_COMPONENT)
+#include "chrome/browser/component_updater/cld_component_installer.h"
+#endif
+
+namespace {
+  // The data file,  cached as long as the process stays alive.
+  // We also track the offset at which the data starts, and its length.
+  base::File* s_cached_file_;    // guarded by file_lock_
+  uint64 s_cached_data_offset_;  // guarded by file_lock_
+  uint64 s_cached_data_length_;  // guarded by file_lock_
+
+  // Guards s_cached_file_
+  base::LazyInstance<base::Lock> s_file_lock_;
+}  // namespace
+
+namespace content {
+
+DataFileBrowserCldDataProvider::DataFileBrowserCldDataProvider(
+    content::RenderViewHost* render_view_host)
+    : render_view_host (render_view_host),
+      weak_pointer_factory_(this) {
+}
+
+bool DataFileBrowserCldDataProvider::OnMessageReceived(
+    const IPC::Message& message) {
+  bool handled = true;
+  IPC_BEGIN_MESSAGE_MAP(DataFileBrowserCldDataProvider, message)
+  IPC_MESSAGE_HANDLER(ChromeViewHostMsg_NeedCldDataFile, OnCldDataRequest)
+  IPC_MESSAGE_UNHANDLED(handled = false)
+  IPC_END_MESSAGE_MAP()
+  return handled;
+}
+
+void DataFileBrowserCldDataProvider::OnCldDataRequest() {
+  // Quickly try to read s_cached_file_. If valid, the file handle is
+  // cached and can be used immediately. Else, queue the caching task to the
+  // blocking pool.
+  base::File* handle = NULL;
+  uint64 data_offset = 0;
+  uint64 data_length = 0;
+  {
+    base::AutoLock lock(s_file_lock_.Get());
+    handle = s_cached_file_;
+    data_offset = s_cached_data_offset_;
+    data_length = s_cached_data_length_;
+  }
+
+  if (handle && handle->IsValid()) {
+    // Cached data available. Respond to the request.
+    SendCldDataResponseInternal(handle, data_offset, data_length);
+    return;
+  }
+
+  // Else, we don't have the data file yet. Queue a caching attempt.
+  // The caching attempt happens in the blocking pool because it may involve
+  // arbitrary filesystem access.
+  // After the caching attempt is made, we call MaybeSendCLDDataAvailable
+  // to pass the file handle to the renderer. This only results in an IPC
+  // message if the caching attempt was successful.
+  content::BrowserThread::PostBlockingPoolTaskAndReply(
+      FROM_HERE,
+      base::Bind(&DataFileBrowserCldDataProvider::OnCldDataRequestInternal,
+                 weak_pointer_factory_.GetWeakPtr()),
+      base::Bind(&DataFileBrowserCldDataProvider::SendCldDataResponse,
+                 weak_pointer_factory_.GetWeakPtr()));
+
+}
+
+void DataFileBrowserCldDataProvider::SendCldDataResponse() {
+  base::File* handle = NULL;
+  uint64 data_offset = 0;
+  uint64 data_length = 0;
+  {
+    base::AutoLock lock(s_file_lock_.Get());
+    handle = s_cached_file_;
+    data_offset = s_cached_data_offset_;
+    data_length = s_cached_data_length_;
+  }
+
+  if (handle && handle->IsValid())
+    SendCldDataResponseInternal(handle, data_offset, data_length);
+}
+
+void DataFileBrowserCldDataProvider::SendCldDataResponseInternal(
+    const base::File* handle,
+    const uint64 data_offset,
+    const uint64 data_length) {
+  // Data available, respond to the request.
+  IPC::PlatformFileForTransit ipc_platform_file =
+      IPC::GetFileHandleForProcess(
+          handle->GetPlatformFile(),
+          render_view_host->GetProcess()->GetHandle(),
+          false);
+  // In general, sending a response from within the code path that is processing
+  // a request is discouraged because there is potential for deadlock (if the
+  // methods are sent synchronously) or loops (if the response can trigger a
+  // new request). Neither of these concerns is relevant in this code, so
+  // sending the response from within the code path of the request handler is
+  // safe.
+  render_view_host->Send(new ChromeViewMsg_CldDataFileAvailable(
+      render_view_host->GetRoutingID(),
+      ipc_platform_file,
+      data_offset,
+      data_length));
+}
+
+void DataFileBrowserCldDataProvider::OnCldDataRequestInternal() {
+  // Because this function involves arbitrary file system access, it must run
+  // on the blocking pool.
+  DCHECK(!content::BrowserThread::CurrentlyOn(content::BrowserThread::IO));
+  DCHECK(!content::BrowserThread::CurrentlyOn(content::BrowserThread::UI));
+
+  {
+    base::AutoLock lock(s_file_lock_.Get());
+    if (s_cached_file_)
+      return;  // Already done, duplicate request
+  }
+
+#if defined(CLD2_IS_COMPONENT)
+  base::FilePath path = component_updater::GetLatestCldDataFile();
+  if (path.empty())
+    return;
+#else  // CLD2 data is at a well-known file path
+  base::FilePath path;
+  if (!PathService::Get(chrome::DIR_USER_DATA, &path)) {
+    LOG(WARNING) << "Unable to locate user data directory";
+    return;  // Chrome isn't properly installed.
+  }
+  path = path.Append(chrome::kCLDDataFilename);
+#endif
+
+  // If the file exists, we can send an IPC-safe construct back to the
+  // renderer process immediately; otherwise, nothing to do here.
+  if (!base::PathExists(path))
+    return;
+
+  // Attempt to open the file for reading.
+  scoped_ptr<base::File> file(
+      new base::File(path, base::File::FLAG_OPEN | base::File::FLAG_READ));
+  if (!file->IsValid()) {
+    LOG(WARNING) << "CLD data file exists but cannot be opened";
+    return;
+  }
+
+  base::File::Info file_info;
+  if (!file->GetInfo(&file_info)) {
+    LOG(WARNING) << "CLD data file exists but cannot be inspected";
+    return;
+  }
+
+  // For now, our offset and length are simply 0 and the length of the file,
+  // respectively. If we later decide to include the CLD2 data file inside of
+  // a larger binary context, these params can be twiddled appropriately.
+  const uint64 data_offset = 0;
+  const uint64 data_length = file_info.size;
+
+  {
+    base::AutoLock lock(s_file_lock_.Get());
+    if (s_cached_file_) {
+      // Idempotence: Racing another request on the blocking pool, abort.
+    } else {
+      // Else, this request has taken care of it all. Cache all info.
+      s_cached_file_ = file.release();
+      s_cached_data_offset_ = data_offset;
+      s_cached_data_length_ = data_length;
+    }
+  }
+}
+
+}  // namespace content