[#22] Change parsing engine from regex to PEG (parboiled)

parboiled is PEG (Parsing Expression Grammar) implementation.
PEG is more concise than regex, and regex could not handle recursive
structure well.

GcEventNode is added to access the parsed data easily. Its concrete class is
generated by Auto Value library which utilizes annotation processing. IDE setup
is required and the instruction can be found here:
    google/auto#106

common/src/main/proto/gc_model.proto

@@ -21,4 +21,4 @@ message GcEvent {
     double sys_time = 7;                         // time spent in OS call or waiting for system event
     double real_time = 8;                        // (user_time + sys_tim) / threads# + alpha
     double ref_time = 9;                         // reference processing time
-}
+}

parser/.gitignore

@@ -0,0 +1,2 @@
+src/generated
+src/generated_tests

parser/build.gradle

@@ -1,4 +1,14 @@
+apply plugin: 'idea'
+
 dependencies {
-    compile project(':common')
-    compile 'org.apache.commons:commons-lang3:3.4'
-}
+  compile project(':common')
+  compile 'org.apache.commons:commons-lang3:3.4'
+  compile 'org.parboiled:parboiled-java:1.1.7'
+  compile 'com.google.auto.value:auto-value:1.2'
+}
+
+idea {
+  module {
+    sourceDirs += file("${projectDir}/src/generated")
+  }
+}

parser/src/main/java/edu/kaist/algo/parser/CmsGcLogRule.java

@@ -0,0 +1,196 @@
+package edu.kaist.algo.parser;
+
+import org.apache.commons.lang3.StringUtils;
+import org.parboiled.BaseParser;
+import org.parboiled.Rule;
+import org.parboiled.annotations.BuildParseTree;
+import org.parboiled.annotations.Label;
+import org.parboiled.annotations.SuppressSubnodes;
+
+/**
+ * PEG (Parsing Expression Grammar) for the CMS GC log.
+ *
+ * <p>This PEG is for parsing the one line of CMS GC logs. The line should be complete. That is,
+ * it should not be cut off by another thread's interference.
+ *
+ * <p>Beware: CMS-related logs are not supported yet.
+ *
+ * <p>Following options are required:
+ * <ul>
+ * <li>-XX:+UseConcMarkSweepGC</li>
+ * <li>-XX:+UnlockDiagnosticVMOptions</li>
+ * <li>-XX:+LogVMOutput</li>
+ * <li>-XX:+PrintGCDetails</li>
+ * <li>-XX:+PrintGCTimeStamps</li>
+ * </ul>
+ *
+ * <p>PEG (whitespaces are ignored for conciseness):
+ * <pre>
+ * InputLine <- Event UserSysRealTimes
+ * Event <- (Time ': ')? '[' TypeAndDetail (Event)* UsageAndElapsedTime ']'
+ * Time <- Digits '.' Digits ' secs'?
+ * Digits <- [0-9]+
+ * TypeAndDetail <- Type ('(' Detail ')')? ': '?
+ * Type <- 'GC' / 'ParNew' / 'CMS' / 'Full GC' / 'Metaspace' / '1 CMS-initial-mark'
+ *       / 'YG occupancy' / 'Rescan (parallel)' / 'weak refs processing' / 'class unloading'
+ *       / 'scrub symbol table' / 'scrub string table' / '1 CMS-remark'
+ * Detail <- 'System.gc()' / !')'+
+ * UsageAndElapsedTime <- UsageChange? (', ' Event)? (', ' Time)?
+ * UsageChange <- (Size '-&<span>gt;</span>')? UsageWithTotal
+ * UsageWithTotal <- Size '(' Size ')'
+ * Size <- Digits 'K '
+ * UserSysRealTimes <- '[ Times: user=' Time ' sys=' Time ', real=' Time ']'
+ * </pre>
+ */
+@BuildParseTree
+public class CmsGcLogRule extends BaseParser<Object> {
+
+  Rule InputLine() {
+    return Sequence(
+        push(GcEventNode.builder()),
+        Event(),
+        UserSysRealTimes(),
+        push(popAsNode().build())
+    );
+  }
+
+  Rule Event() {
+    return Sequence(
+        Optional(
+            TimeLong(), ": ",
+            swap() && push(popAsNode().timestamp(popAsLong()))
+        ),
+        "[", TypeAndDetail(), " ",
+        ZeroOrMore(
+            push(GcEventNode.builder()),
+            Event(),
+            swap() && push(popAsNode().addChild(popAsNode().build()))
+        ),
+        " ", UsageAndElapsedTime(), "] "
+    );
+  }
+
+  Rule TypeAndDetail() {
+    return Sequence(
+        Type(),
+        push(popAsNode().type(match())),
+        Optional(" ", "(", Detail(), push(popAsNode().detail(match())), ")"),
+        Optional(": ")
+    );
+  }
+
+  @SuppressSubnodes
+  Rule Type() {
+    return FirstOf("GC", "ParNew", "CMS", "Full GC", "Metaspace", "1 CMS-initial-mark",
+        "YG occupancy", "Rescan (parallel)", "weak refs processing", "class unloading",
+        "scrub symbol table", "scrub string table", "1 CMS-remark");
+  }
+
+  @SuppressSubnodes
+  Rule Detail() {
+    return FirstOf("System.gc()", OneOrMore(NoneOf(")")));
+  }
+
+  Rule UsageAndElapsedTime() {
+    return Sequence(
+        Optional(UsageChange()),
+        Optional(", ",
+            push(GcEventNode.builder()),
+            Event(), // Metaspace
+            swap() && push(popAsNode().addChild(popAsNode().build()))
+        ),
+        Optional(", ",
+            TimeDouble(),
+            swap() && push(popAsNode().elapsedTime(popAsDouble()))
+        )
+    );
+  }
+
+  Rule UsageChange() {
+    return Sequence(
+        Optional(
+            Size(), "-&gt;",
+            swap() && push(popAsNode().prevUsage(popAsLong()))
+        ),
+        UsageWithTotal()
+    );
+  }
+
+  Rule UsageWithTotal() {
+    return Sequence(
+        Size(),
+        "(", Size(), ")",
+        swap3() && push(popAsNode().afterUsage(popAsLong())),
+        push(popAsNode().capacity(popAsLong()))
+    );
+  }
+
+  Rule Size() {
+    return Sequence(
+        Digits(),
+        push(Long.valueOf(match())),
+        WhiteSpace(), "K "
+    );
+  }
+
+  Rule UserSysRealTimes() {
+    return Sequence(
+        "[", "Times: ", "user=", TimeDouble(), " sys=", TimeDouble(), ", real=", TimeDouble(), "]",
+        swap4() && push(popAsNode().user(popAsDouble())),
+        push(popAsNode().sys(popAsDouble())),
+        push(popAsNode().real(popAsDouble()))
+    );
+  }
+
+  @Label("Time")
+  @SuppressSubnodes
+  Rule TimeDouble() {
+    return Sequence(
+        Sequence(Digits(), ".", Digits()),
+        push(Double.valueOf(match())),
+        Optional(" secs")
+    );
+  }
+
+  @Label("Time")
+  @SuppressSubnodes
+  Rule TimeLong() {
+    return Sequence(
+        Sequence(Digits(), ".", Digits()),
+        push(Long.valueOf(StringUtils.remove(match(), ".")))
+    );
+  }
+
+  @SuppressSubnodes
+  Rule Digits() {
+    return OneOrMore(Digit());
+  }
+
+  Rule Digit() {
+    return CharRange('0', '9');
+  }
+
+  @SuppressSubnodes
+  Rule WhiteSpace() {
+    return ZeroOrMore(AnyOf(" \t\f"));
+  }
+
+  @Override
+  protected Rule fromStringLiteral(String string) {
+    return string.endsWith(" ")
+        ? Sequence(String(string.substring(0, string.length() - 1)), WhiteSpace())
+        : String(string);
+  }
+
+  protected Double popAsDouble() {
+    return (Double) pop();
+  }
+
+  protected GcEventNode.Builder popAsNode() {
+    return (GcEventNode.Builder) pop();
+  }
+
+  protected Long popAsLong() {
+    return (Long) pop();
+  }
+}