JDBC instrumentation

[felixbarny]

closes #35

[codecov-io]

Codecov Report

Merging #84 into master will decrease coverage by 0.31%.
The diff coverage is 82.89%.

@@             Coverage Diff              @@
##             master      #84      +/-   ##
============================================
- Coverage     74.87%   74.55%   -0.32%     
- Complexity      779      793      +14     
============================================
  Files            87       90       +3     
  Lines          2901     2975      +74     
  Branches        278      283       +5     
============================================
+ Hits           2172     2218      +46     
- Misses          594      615      +21     
- Partials        135      142       +7

Impacted Files	Coverage Δ	Complexity Δ
...java/co/elastic/apm/jdbc/ApmJdbcEventListener.java	52.3% <74.07%> (-9.92%)	8 <4> (-2)
...tic/apm/jdbc/PreparedStatementInstrumentation.java	82.35% <82.35%> (ø)	6 <6> (?)
.../co/elastic/apm/jdbc/StatementInstrumentation.java	87.5% <87.5%> (ø)	7 <7> (?)
...co/elastic/apm/jdbc/ConnectionInstrumentation.java	93.75% <93.75%> (ø)	6 <6> (?)
.../elastic/apm/jdbc/ApmJdbcEventListenerFactory.java	0% <0%> (-100%)	0% <0%> (-3%)
...lastic/apm/servlet/FilterChainInstrumentation.java	27.27% <0%> (-2.73%)	5% <0%> (ø)
...co/elastic/apm/servlet/ServletInstrumentation.java	31.57% <0%> (-0.86%)	5% <0%> (ø)
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 274b42c...f711c8c. Read the comment docs.

[tylerbenson]

Just out of curiosity, why was this needed? Is this a performance optimization to avoid the isSubTypeOf check?

[felixbarny]

See the comment in https://github.com/elastic/apm-agent-java/blob/master/apm-agent-plugins/apm-servlet-plugin/src/main/java/co/elastic/apm/servlet/ServletInstrumentation.java#L69
Basically it’s because the isSubType marcher can be pretty expensive. It causes a lot of type descriptions to be parsed by ByteBuddy. I stagemonitor, I cached TypeDescriptions to improve instrumentation performance (startup times). A user experienced an OOME because of an isSubType matcher. But I should re evaluate with a recent Byte Buddy version.

[felixbarny]

What is your experience in regards to the performance of the isSubType(Class) and hasSuperType(named(String)) matchers?

[felixbarny]

TODO: more integration tests