[UMA] Remove redundant bucket_count variable from base::Histogram.

base/metrics/bucket_ranges.h

Index: base/metrics/bucket_ranges.h
diff --git a/base/metrics/bucket_ranges.h b/base/metrics/bucket_ranges.h
index 210b9a7e510a9efea40196248e78c7a5e0502d28..4b07b57eb08c841e22ce700a3c504c0545a20bcc 100644
--- a/base/metrics/bucket_ranges.h
+++ b/base/metrics/bucket_ranges.h
@@ -39,6 +39,12 @@ class BASE_EXPORT BucketRanges {
   uint32 checksum() const { return checksum_; }
   void set_checksum(uint32 checksum) { checksum_ = checksum; }
 
+  // A bucket is defined by a consecutive pair of entries in |ranges|, so there
+  // is one fewer bucket than there are ranges.  For example, if |ranges| is
+  // [0, 1, 3, 7, INT_MAX], then the buckets in this histogram are
+  // [0, 1), [1, 3), [3, 7), and [7, INT_MAX).
+  size_t BucketCount() const { return ranges_.size() - 1; }

[Alexei Svitkine (slow), 2013/06/20 05:53:32]

Drive by nit:

This should probably either use hacker_style or be called GetBucketCount() and
be defined out of line.

[Ilya Sherman, 2013/06/20 05:57:34]

I don't have a strong preference.  Jim, which of these would you prefer?

[jar (doing other things), 2013/06/20 16:41:18]

I was a little tentative about the cost... but I guess size() on a vector should
be super fast, and I shouldn't sweat it.  The place where it *might* be an issue
is when we are looking for corruption, and we use this call to find the max
bound (time and again).  I guess I should wait until we have a profiler telling
me this was a hot spot.


I'm fine with retaining hacker style, and putting this in the header as you did.

[Ilya Sherman, 2013/06/21 04:03:05]

Done.

+
   // Checksum methods to verify whether the ranges are corrupted (e.g. bad
   // memory access).
   uint32 CalculateChecksum() const;