Blob Consolidation & Registry Hookup

content/child/blob_storage/blob_consolidation.cc

Index: content/child/blob_storage/blob_consolidation.cc
diff --git a/content/child/blob_storage/blob_consolidation.cc b/content/child/blob_storage/blob_consolidation.cc
new file mode 100644
index 0000000000000000000000000000000000000000..cd98614f41dcdc169486a41bd84945e0962deb1a
--- /dev/null
+++ b/content/child/blob_storage/blob_consolidation.cc
@@ -0,0 +1,169 @@
+// Copyright 2015 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 <string>
+
+#include "content/child/blob_storage/blob_consolidation.h"
+
+using storage::DataElement;
+using blink::WebBlobData;
+using blink::WebString;
+using blink::WebThreadSafeData;
+using blink::WebURL;
+
+namespace content {
+
+using ReadStatus = BlobConsolidation::ReadStatus;
+
+BlobConsolidation::ConsolidatedItem::ConsolidatedItem()
+    : type(DataElement::TYPE_UNKNOWN),
+      offset(0),
+      length(kuint64max),
+      expected_modification_time(0) {}
+BlobConsolidation::ConsolidatedItem::~ConsolidatedItem() {}
+
+BlobConsolidation::ConsolidatedItem::ConsolidatedItem(DataElement::Type type,
+                                                      uint64_t offset,
+                                                      uint64_t length)
+    : type(type),
+      offset(offset),
+      length(length),
+      expected_modification_time(0) {}
+
+BlobConsolidation::BlobConsolidation()
+    : total_memory_(0), total_memory_read_(0) {}
+
+BlobConsolidation::~BlobConsolidation() {}
+
+void BlobConsolidation::AddDataItem(const WebThreadSafeData& data) {
+  ConsolidatedItem* item;
+  if (consolidated_items_.empty() ||
+      consolidated_items_.back().type != DataElement::TYPE_BYTES) {
+    consolidated_items_.push_back(
+        ConsolidatedItem(DataElement::TYPE_BYTES, 0, 0));
+  }
+  item = &consolidated_items_.back();
+  if (!item->data.empty()) {
+    item->offsets.push_back(static_cast<size_t>(item->length));
+  }
+  item->length += data.size();
+  total_memory_ += data.size();
+  item->data.push_back(WebThreadSafeData());
+  item->data.back().assign(data);
+}
+
+void BlobConsolidation::AddFileItem(const WebString& path, uint64_t offset,
+                                    uint64_t length,
+                                    double expected_modification_time) {
+  consolidated_items_.push_back(
+      ConsolidatedItem(DataElement::TYPE_FILE, offset, length));
+  ConsolidatedItem& item = consolidated_items_.back();
+  item.path.assign(path);
+  item.expected_modification_time = expected_modification_time;
+}
+
+void BlobConsolidation::AddBlobItem(const WebString& uuid, uint64_t offset,
+                                    uint64_t length) {
+  consolidated_items_.push_back(
+      ConsolidatedItem(DataElement::TYPE_BLOB, offset, length));
+  ConsolidatedItem& item = consolidated_items_.back();
+  item.blob_uuid.assign(uuid);
+}
+
+void BlobConsolidation::AddFileSystemItem(const WebURL& url, uint64_t offset,
+                                          uint64_t length,
+                                          double expected_modification_time) {
+  consolidated_items_.push_back(
+      ConsolidatedItem(DataElement::TYPE_FILE_FILESYSTEM, offset, length));
+  ConsolidatedItem& item = consolidated_items_.back();
+  item.filesystem_url = url;
+  item.expected_modification_time = expected_modification_time;
+}
+
+ReadStatus BlobConsolidation::RecordMemoryRead(size_t consolidated_item_index,

[michaeln, 2015/06/17 03:17:55]

please put the method bodies in the .cc in the same order as declared in the .h
file (so this one goes after ReadMemory)

[dmurph, 2015/06/17 18:29:50]

Done.

+                                               size_t memory_size) {
+  if (consolidated_item_index >= consolidated_items_.size()) {
+    return ReadStatus::ERROR_OUT_OF_BOUNDS;
+  }
+  if (total_memory_read_ + memory_size > total_memory_) {
+    LOG(ERROR) << "Memory for blob marked as read more than once.";
+    return ReadStatus::ERROR;
+  }
+  const ConsolidatedItem& item = consolidated_items_[consolidated_item_index];
+  if (item.type != DataElement::TYPE_BYTES) {
+    LOG(ERROR) << "Cannot read bytes from item with type " << item.type;
+    return ReadStatus::ERROR_WRONG_TYPE;
+  }
+
+  // Keep track of memory read.
+  total_memory_read_ += memory_size;
+  return total_memory_read_ < total_memory_ ? ReadStatus::BLOB_BYTES_PENDING
+                                            : ReadStatus::DONE;
+}
+
+ReadStatus BlobConsolidation::ReadMemory(size_t consolidated_item_index,
+                                         size_t consolidated_offset,
+                                         size_t consolidated_size,
+                                         char* memory_out) {
+  if (consolidated_item_index >= consolidated_items_.size()) {
+    return ReadStatus::ERROR_OUT_OF_BOUNDS;
+  }
+  const ConsolidatedItem& item = consolidated_items_[consolidated_item_index];
+  if (item.type != DataElement::TYPE_BYTES) {
+    return ReadStatus::ERROR_WRONG_TYPE;
+  }
+  if (consolidated_size + consolidated_offset > item.length) {
+    LOG(ERROR) << "Invalid consolidated size " << consolidated_size
+               << " and offset " << consolidated_offset << " vs item length of "
+               << item.length;
+    return ReadStatus::ERROR_OUT_OF_BOUNDS;
+  }
+
+  // We do a binary search to find the correct data to start with in the data

[michaeln, 2015/06/17 03:17:55]

fun ;)

+  // elements.  This is slightly customized due to our unique storage and
+  // constraints
+  size_t mid = 0;
+  size_t offset_from_mid = consolidated_offset;
+  size_t num_items = item.data.size();
+  if (!item.offsets.empty()) {
+    size_t low = 0;
+    size_t high = num_items - 1;
+    while (true) {
+      mid = (high + low) / 2;
+      // Note: we don't include the implicit '0' for the first item in offsets,
+      // so we do (index - 1).

[michaeln, 2015/06/17 03:17:55]

i'd say you don't need what's after the , in the comment. Nice to keep it one
line. If you keep it, maybe say [mid - 1].

[dmurph, 2015/06/17 18:29:50]

Done.

+      size_t item_offset = (mid == 0 ? 0 : item.offsets[mid - 1]);
+      offset_from_mid = consolidated_offset - item_offset;
+      size_t next_item_offset = (mid + 1 == num_items ? 0 : item.offsets[mid]);
+      if (item_offset == consolidated_offset) {
+        // found exact match.
+        break;
+      } else if (item_offset > consolidated_offset) {
+        high = mid - 1;
+      } else if (mid + 1 == num_items ||
+                 next_item_offset > consolidated_offset) {
+        // We are at the last item or the next offset it greater than the one
+        // we want, so the current item wins.
+        break;
+      } else {
+        low = mid + 1;
+      }
+    }
+  }
+
+  DCHECK_LT(offset_from_mid, item.data[mid].size());
+  // read starting from 'mid' and 'offset_from_mid'

[michaeln, 2015/06/17 03:17:55]

nit: cap "R" and a period at the end

[dmurph, 2015/06/17 18:29:50]

Done.

+  for (size_t memory_read = 0;
+       mid < num_items && memory_read < consolidated_size; mid++) {
+    size_t read_size = std::min(item.data[mid].size() - offset_from_mid,
+                                consolidated_size - memory_read);
+    memmove(memory_out + memory_read, item.data[mid].data() + offset_from_mid,
+            read_size);
+    offset_from_mid = 0;
+    memory_read += read_size;
+  }
+  return ReadStatus::DONE;
+}
+
+}  // namespace content