Add net::GzipSourceStream

net/filter/gzip_source_stream.h

+// 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.
+
+#ifndef NET_FILTER_GZIP_SOURCE_STREAM_H_
+#define NET_FILTER_GZIP_SOURCE_STREAM_H_
+
+#include <memory>
+
+#include "base/macros.h"
+#include "base/memory/ref_counted.h"
+#include "net/base/io_buffer.h"
+#include "net/base/net_export.h"
+#include "net/filter/filter_source_stream.h"
+#include "net/filter/gzip_header.h"
+
+typedef struct z_stream_s z_stream;
+
+namespace net {
+
+class IOBuffer;
+
+// GZipSourceStream applies gzip and deflate content encoding/decoding to a data
+// stream. As specified by HTTP 1.1, with gzip encoding the content is
+// wrapped with a gzip header, and with deflate encoding the content is in
+// a raw, headerless DEFLATE stream.
+//
+// Internally GZipSourceStream uses zlib inflate to do decoding.
+//
+class NET_EXPORT_PRIVATE GzipSourceStream : public FilterSourceStream {
+ public:
+  enum GzipSourceStreamMode {
+    GZIP_SOURCE_STREAM_DEFLATE,
+    GZIP_SOURCE_STREAM_GZIP,
+    // Same as GZIP_SOURCE_STREAM_GZIP, but allow fallback to plain text.
+    GZIP_SOURCE_STREAM_GZIP_WITH_FALLBACK,
+  };
+
+  ~GzipSourceStream() override;
+
+  // Creates a GzipSourceStream. Return nullptr if initialization fails.
+  static std::unique_ptr<GzipSourceStream> Create(
+      std::unique_ptr<SourceStream> previous,
+      GzipSourceStreamMode mode);
+
+ private:
+  GzipSourceStream(std::unique_ptr<SourceStream> previous,
+                   GzipSourceStreamMode mode);
+  bool Init();

[Randy Smith (Not in Mondays), 2016/09/12 20:54:01]

Documentation as to the return value of Init()?

[xunjieli, 2016/09/14 16:44:01]

Done.

+
+  // SourceStream implementation
+  std::string GetTypeAsString() const override;
+  int FilterData(IOBuffer* output_buffer,
+                 int output_buffer_size,
+                 DrainableIOBuffer* input_buffer) override;
+
+  // Synchronous decompressor. This function consumes bytes from |input_buffer|
+  // and decompresses them into |output_buffer| until it consumes all of
+  // |input_buffer|. This decompressor will decompress a zlib stream (either
+  // gzip or deflate) until the zlib reaches EOF, then will pass any further
+  // input through untouched.
+  int Decompress(IOBuffer* output_buffer,
+                 size_t output_buffer_size,
+                 DrainableIOBuffer* input_buffer);
+
+  // Returns how many bytes are drained from |input_buffer|.

[Randy Smith (Not in Mondays), 2016/09/12 20:54:01]

I don't understand based on the comment what this function does?  (I assume it
copies directly from input to output, but it should say that.)

[xunjieli, 2016/09/14 16:44:01]

Done.

[Randy Smith (Not in Mondays), 2016/09/14 22:25:23]

Not quite what I meant.  The documentation is still documenting only the return
value, not what the function *does*.  If it purely copied just saying "Copies
bytes from input_buffer to output_buffer, returning the number of bytes copied"
would be fine.  But my memory of the implementation is that the function is more
complex and does something with gzip headers.  What does the function do?

[xunjieli, 2016/09/19 13:57:02]

Done. It additionally drains footer bytes from the stream. I've adjusted the
documentation. Let me know if that's clearer.

+  size_t Passthrough(char* output_buffer,
+                     size_t output_buffer_size,
+                     DrainableIOBuffer* input_buffer);
+
+  // Inserts a zlib header to the data stream before calling zlib inflate.
+  // This is used to work around server bugs.
+  // The function returns true on success and false otherwise.
+  bool InsertZlibHeader();
+
+  // Returns whether an invalid GZip header has been observed. This method
+  // returns true as soon as an invalid GZip header is observed; it returns
+  // false if a complete, valid header has been observed, or not enough bytes
+  // have been seen to be sure. This method pulls its input from |buffer_|.

[Randy Smith (Not in Mondays), 2016/09/12 20:54:01]

I presume this can be called multiple times with multiple input_buffers?  And
that in the case of returning false for a complete, valid header, it doesn't
drain any more bytes?

[xunjieli, 2016/09/14 16:44:01]

Done. Yes, I believe you are right. I will add this to the doc.

+  bool IsGzipHeaderInvalid(DrainableIOBuffer* input_buffer);
+
+  // Returns whether this stream looks like it could be plain text (ie, not
+  // actually gzipped). Right now this uses an extremely simple heuristic; see
+  // the source for details. This method checks the input in |buffer_|, but does
+  // not drain from it.
+  bool ShouldFallbackToPlain(DrainableIOBuffer* input_buffer);
+
+  // Skips gzip footer bytes if necessary. Might drain input from |buffer_| if
+  // there are still footer bytes to read.
+  void SkipGzipFooterIfNeeded(DrainableIOBuffer* input_buffer);
+
+  // Indicates the type of content decoding the stream source is performing.
+  // This variable is set only once by Init.
+  GzipSourceStreamMode mode_;

[Randy Smith (Not in Mondays), 2016/09/12 20:54:01]

It looks to me as if this is set in the constructor, not in Init().  If that's
true, can it be made const?

[Randy Smith (Not in Mondays), 2016/09/12 20:54:01]

nit, suggestion: I prefer whitespace before explanatory comments, to make it
easier to visually group a comment with the code it applies to.

[xunjieli, 2016/09/14 16:44:01]

Done.

[xunjieli, 2016/09/14 16:44:01]

Done.

+  // The control block of zlib which actually does the decoding.
+  // This data structure is initialized by Init and updated only by
+  // Decompress, with InsertZlibHeader being the exception as a workaround.
+  std::unique_ptr<z_stream> zlib_stream_;
+  // Whether end of stream has been reached.
+  bool zlib_eof_;
+  // A flag used by Decompress to record whether we've successfully added
+  // a zlib header to this stream.
+  bool zlib_header_added_;
+  // Used to parse the gzip header in gzip stream.
+  // It is used when the decoding mode is GZIP_SOURCE_STREAM_GZIP.
+  GZipHeader gzip_header_;
+  // Whether gzip header should be checked for validity.
+  bool should_check_gzip_header_;
+  // Tracks how many bytes of gzip footer are yet to be filtered.
+  size_t gzip_footer_bytes_left_;
+
+  DISALLOW_COPY_AND_ASSIGN(GzipSourceStream);
+};
+
+}  // namespace net
+
+#endif  // NET_FILTER_GZIP_SOURCE_STREAM_H__