Video capture: Refactor GetBestMatchedFormat from Win to OS independent

media/video/capture/video_capture_types.cc

Index: media/video/capture/video_capture_types.cc
diff --git a/media/video/capture/video_capture_types.cc b/media/video/capture/video_capture_types.cc
index ce5c354e290eed22fef06f69a067b4cf2d01265c..56671df9f2972a99d39ebcea0a5917566032b7ba 100644
--- a/media/video/capture/video_capture_types.cc
+++ b/media/video/capture/video_capture_types.cc
@@ -4,6 +4,8 @@
 
 #include "media/video/capture/video_capture_types.h"
 
+#include <cmath>
+
 #include "base/logging.h"
 #include "base/strings/stringprintf.h"
 #include "media/base/limits.h"
@@ -69,4 +71,17 @@ std::string VideoCaptureFormat::PixelFormatToString(VideoPixelFormat format) {
 
 VideoCaptureParams::VideoCaptureParams()
     : resolution_change_policy(RESOLUTION_POLICY_FIXED) {}
+
+int DiffVideoCaptureFormat(const VideoCaptureFormat& lhs,
+                           const VideoCaptureFormat& rhs) {
+  // Distinguish between rational frames rates by multiplying with
+  // kFrameRatePrecision. Multiply frame size as well to make it have
+  // precedence over frame rate.

[mcasas, 2014/09/23 12:34:57]

This comment doesn't read clearly to me, can you please rewrite?
(specially if you take the next comment in)

+  return kFrameRatePrecision *
+             (std::abs(lhs.frame_size.width() - rhs.frame_size.width()) +
+              std::abs(lhs.frame_size.height() - rhs.frame_size.height())) +
+         static_cast<int>(kFrameRatePrecision *
+                          std::fabs(lhs.frame_rate - rhs.frame_rate));

[mcasas, 2014/09/23 12:34:57]

what about:

return kFrameRatePrecision * 
    (std::abs(lhs.frame_size.width() - rhs.frame_size.width()) +
    std::abs(lhs.frame_size.height() - rhs.frame_size.height()) + 
    static_cast<int>(std::fabs(lhs.frame_rate - rhs.frame_rate)));

[magjed_chromium, 2014/09/23 14:39:54]

The reason for not factoring kFrameRatePrecision is that we have both a float
multiplication and an int multiplication. Your suggestion does not work because
you have already lost the precision in the frame rate and are just blowing up an
integer. The alternative is to convert everything to float, then multiply, then
convert to int. I implemented that change.

+}
+
 }  // namespace media