Reduce instrumentation overhead

This reduces the overhead of the method instrumentation code primarily
by reducing the number of method calls. There are also some other small
optimisations such as not casting timing values to Floats (there's no
particular need for this), using Symbols for method call metric names,
and reducing the number of Hash lookups for instrumented methods.

The exact impact depends on the code being executed. For example, for a
method that's only called once the difference won't be very noticeable.
However, for methods that are called many times the difference can be
more significant.

For example, the loading time of a large commit
(nrclark/dummy_project@81ebdea5df2fb42e59257cb3eaad671a5c53ca36)
was reduced from around 19 seconds to around 15 seconds using these
changes.

1 files changed, 4 insertions, 12 deletions

diff --git a/spec/lib/gitlab/metrics/transaction_spec.rb b/spec/lib/gitlab/metrics/transaction_spec.rb
index 3b1c67a2147..f1a191d9410 100644
--- a/spec/lib/gitlab/metrics/transaction_spec.rb
+++ b/spec/lib/gitlab/metrics/transaction_spec.rb
@@ -46,19 +46,11 @@ describe Gitlab::Metrics::Transaction do
     end
   end
 
-  describe '#measure_method' do
-    it 'adds a new method if it does not exist already' do
-      transaction.measure_method('Foo#bar') { 'foo' }
+  describe '#method_call_for' do
+    it 'returns a MethodCall' do
+      method = transaction.method_call_for('Foo#bar')
 
-      expect(transaction.methods['Foo#bar']).
-        to be_an_instance_of(Gitlab::Metrics::MethodCall)
-    end
-
-    it 'adds timings to an existing method call' do
-      transaction.measure_method('Foo#bar') { 'foo' }
-      transaction.measure_method('Foo#bar') { 'foo' }
-
-      expect(transaction.methods['Foo#bar'].call_count).to eq(2)
+      expect(method).to be_an_instance_of(Gitlab::Metrics::MethodCall)
     end
   end