[PPAPI] API definition for video media stream artifacts

ppapi/api/ppb_video_frame.idl

+/* Copyright (c) 2013 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.
+ */
+
+/**
+ * Defines the <code>PPB_VideoFrame</code> interface.
+ */
+label Chrome {
+  M33 = 0.1
+};
+
+enum PP_VideoFrame_Format {
+  /**
+   * Unknown format value.
+   */
+  PP_VIDEOFRAME_FORMAT_UNKNOWN = 0,
+
+  /**
+   * 12bpp YVU planar 1x1 Y, 2x2 VU samples
+   */
+  PP_VIDEOFRAME_FORMAT_YV12 = 1,
+
+  /**
+   * 16bpp YVU planar 1x1 Y, 2x1 VU samples
+   */
+  PP_VIDEOFRAME_FORMAT_YV16 = 2,
+
+  /**
+   * 12bpp YVU planar 1x1 Y, 2x2 VU samples
+   */
+  PP_VIDEOFRAME_FORMAT_I420 = 3,
+
+  /**
+   * 20bpp YVU planar 1x1 Y, 2x2 VU, 1x1 A samples
+   */
+  PP_VIDEOFRAME_FORMAT_YV12A = 4,
+
+  /**
+   * JPEG color range version of YV12
+   */
+  PP_VIDEOFRAME_FORMAT_YV12J = 7

[dmichael (off chromium), 2013/12/20 17:42:52]

Why is this 7?


Are all of these uncompressed, and so a frame has fixed size? (pardon my
ignorance)

[yzshen1, 2013/12/20 17:52:26]

I can see that you want to match the enum values of media/base/video_frame.h.
But maybe it is better to not leave holes and do explicit conversion?

[Peng, 2013/12/20 18:49:17]

I think it is uncompressed too. Just the color range is slight different.

+};
+
+interface PPB_VideoFrame {
+  /**
+   * Determines if a resource is a VideoFrame resource.
+   *
+   * @param[in] resource The <code>PP_Resource</code> to test.
+   *
+   * @return A <code>PP_Bool</code> with <code>PP_TRUE</code> if the given
+   * resource is a VideoFrame resource or <code>PP_FALSE</code> otherwise.
+   */
+  PP_Bool IsVideoFrame([in] PP_Resource resource);
+
+  /**
+   * Get timestamp of the video frame. Given in seconds since the

[yzshen1, 2013/12/20 17:52:26]

Get*s* (Here and similar issues below.)

[Peng, 2013/12/20 20:55:31]

Done.

+   * start of the containing video stream.

[yzshen1, 2013/12/20 17:52:26]

You could use @param[in] @return to be consistent.

[Peng, 2013/12/20 20:55:31]

Done.

+   */
+  PP_TimeDelta GetTimestamp([in] PP_Resource resource);

[yzshen1, 2013/12/20 17:52:26]

Please change all |resource| below to |video_frame|.

[Peng, 2013/12/20 20:55:31]

Done.

+
+  /**
+   * Set the timestamp of the video frame. Given in seconds since the
+   * start of the containing video stream.
+   */
+  void SetTimestamp([in] PP_Resource resource, [in] PP_TimeDelta timestamp);

[dmichael (off chromium), 2013/12/20 17:42:52]

Hmm, you know, any Pepper calls we make require acquiring the proxy lock, which
might be heavily contested. Would it make sense at all to instead make the
timestamp a parameter to the PutFrame function (or whatever it will be called)?

[Peng, 2013/12/20 18:49:17]

What about the GetTimestamp() and other Get functions? Maybe we should use a
VideoFrameInfo c structure, and then PutFrame(), GetFrame(), etc will look like:

PutFrame(const struct VideoFrameInfo* info, mem_t pixels);
GetFrame(struct VideoFrameInfo* info, mem_t* pixels);
AllocFrame(VideoFrameInfo* info, mem_t* pixels);
RecycleFrame(mem_t pixels);

Thought?

[yzshen1, 2013/12/20 23:37:43]

IMO, the current PP_Resource way seems cleaner.

I think the performance won't be too bad, we have many PP_Resources that have
getter/setters. What do you think?

On 2013/12/20 18:49:17, Peng wrote:

+
+  /**
+   * Get format of the video frame.
+   */
+  PP_VideoFrame_Format GetFormat([in] PP_Resource resource);
+
+  /**
+   * Get size of the video frame.
+   */
+  PP_Bool GetSize([in] PP_Resource resource, [out] PP_Size size);
+
+  /**
+   * Get data buffer for video frame pixels.
+   */
+  mem_t GetDataBuffer([in] PP_Resource resource);
+
+  /**
+   * Get size of data buffer.
+   */
+  uint32_t GetDataBufferSize([in] PP_Resource resource);
+};