add a download slices table into history download db

components/history/core/browser/download_job_info.h

Index: components/history/core/browser/download_job_info.h
diff --git a/components/history/core/browser/download_job_info.h b/components/history/core/browser/download_job_info.h
new file mode 100644
index 0000000000000000000000000000000000000000..ba73cfc3f027d4fb196961e28b6c78d3268ef2e4
--- /dev/null
+++ b/components/history/core/browser/download_job_info.h
@@ -0,0 +1,58 @@
+// Copyright 2017 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.
+
+#ifndef COMPONENTS_HISTORY_CORE_BROWSER_DOWNLOAD_JOB_INFO_H_
+#define COMPONENTS_HISTORY_CORE_BROWSER_DOWNLOAD_JOB_INFO_H_
+
+#include <stdint.h>
+#include <string>
+
+#include "components/history/core/browser/download_types.h"
+
+namespace history {
+
+// Contains the download job information that is stored in the download
+// system's persistent store (or refers to it). Each download item can have
+// multiple download jobs associated with it, with each job having its own
+// start position.
+struct DownloadJobInfo {
+  DownloadJobInfo();
+  DownloadJobInfo(DownloadId download_id,
+                  int job_id,
+                  int64_t start_position,
+                  int64_t length,
+                  int64_t received_bytes,
+                  DownloadState state,
+                  DownloadInterruptReason interrupt_reason);
+  DownloadJobInfo(const DownloadJobInfo& other);
+  ~DownloadJobInfo();
+
+  bool operator==(const DownloadJobInfo&) const;
+
+  // The id of the download in the database. Not changed by UpdateDownload().

[Scott Hess - ex-Googler, 2017/02/08 22:01:40]

"Not changed by UpdateDownload()" seem irrelevant.  It's passed as a const&, so
UpdateDownload() can't change it.

[qinmin, 2017/02/09 01:01:42]

removed

+  DownloadId download_id;
+
+  // Job id of the download.
+  int job_id;

[Scott Hess - ex-Googler, 2017/02/08 22:01:40]

AFAICT, this CL isn't wired in to anything, so I can't really see how you plan
to use it, but based in the tuple (start_position, length, received_bytes), I
would guess you're going to do something like:
 - Setup a bunch of sub-jobs for the overall download.
 - Make progress, updating received_bytes and state.
 - Finish and remove the sub-jobs.

In that case, start_position and length should uniquely identify the job, and,
indeed, start_position alone would too.  In fact, start_position probably must
be unique.  So do you need a job_id, or is start_position sufficient?

[I don't feel super strongly about this, it's a trivial change to the database
space.  I'm actually more concerned about whether the data model fits usage
well.]

[qinmin, 2017/02/09 01:01:42]

Using start position might not be enough. For example, if Chrome always create 4
jobs to evenly handle the data downloading, and the first job finishes 50% while
the other 3 jobs didn't make any progress.  Suppose browser crashes at that
moment, when resuming the download later, we need to update the starting
position for all jobs so that the remaining 7/8 data is splitted evenly between
the 4 jobs. That will cause us to change the start_position for each job. Using
an id to identify the jobs make it easier to achieve such tasks.

[asanka, 2017/02/09 02:19:47]

I think it's worth fleshing out how you are going to use it.

All you really need is a list of slices of the file that have been successfully
downloaded. Those slices can be uniquely identified by download_id and offset as
shess mentioned. The additional bit you need is the count of bytes downloaded
from that offset.

The list of download jobs is an ephemeral thing that each download session needs
to figure out based on the distribution of the holes in the file and network
conditions. I don't think persisting the download jobs to disk is going to help.

+
+  // Start position of the download request.
+  int64_t start_position;
+
+  // Data length to download. If length < 0, download should continue until
+  // the stream reaches the end.
+  int64_t length;
+
+  // The number of bytes received (so far).
+  int64_t received_bytes;
+
+  // The current state of the download.
+  DownloadState state;
+
+  // The reason the download was interrupted, if state == kStateInterrupted.
+  DownloadInterruptReason interrupt_reason;
+};
+
+}  // namespace history
+
+#endif  // COMPONENTS_HISTORY_CORE_BROWSER_DOWNLOAD_JOB_INFO_H_