Method getContextFeatures

src/org/antlr/codebuff/Trainer.java:546–611

Get the token type and tree ancestor features. These are computed the same for both training and formatting.

(Corpus corpus,
	                                       Map<Token, TerminalNode> tokenToNodeMap,
	                                       InputDocument doc,
	                                       int i)

	 *  the same for both training and formatting.
	 */
	public static int[] getContextFeatures(Corpus corpus,
	                                       Map<Token, TerminalNode> tokenToNodeMap,
	                                       InputDocument doc,
	                                       int i)
	{
		int[] features = new int[NUM_FEATURES];
		CodeBuffTokenStream tokens = doc.tokens;
		TerminalNode node = tokenToNodeMap.get(tokens.get(i));
		if ( node==null ) {
			System.err.println("### No node associated with token "+tokens.get(i));
			return features;
		}
		Token curToken = node.getSymbol();

		// Get context information for previous token
		Token prevToken = tokens.getPreviousRealToken(i);
		TerminalNode prevNode = tokenToNodeMap.get(prevToken);

		ParserRuleContext prevEarliestRightAncestor = earliestAncestorEndingWithToken(prevNode);
		int prevEarliestAncestorRuleIndex = prevEarliestRightAncestor.getRuleIndex();
		int prevEarliestAncestorRuleAltNum = prevEarliestRightAncestor.getAltNumber();

		// Get context information for current token
		ParserRuleContext earliestLeftAncestor = earliestAncestorStartingWithToken(node);

		ParserRuleContext earliestLeftAncestorParent  =
			earliestLeftAncestor!=null ? earliestLeftAncestor.getParent() : null;

		ParserRuleContext earliestLeftAncestorParent2 =
			earliestLeftAncestorParent!=null ? earliestLeftAncestorParent.getParent() : null;

		ParserRuleContext earliestLeftAncestorParent3 =
			earliestLeftAncestorParent2!=null ? earliestLeftAncestorParent2.getParent() : null;

		ParserRuleContext earliestLeftAncestorParent4 =
			earliestLeftAncestorParent3!=null ? earliestLeftAncestorParent3.getParent() : null;

		ParserRuleContext earliestLeftAncestorParent5 =
			earliestLeftAncestorParent4!=null ? earliestLeftAncestorParent4.getParent() : null;

		features[INDEX_PREV_TYPE]                     = prevToken.getType();
		features[INDEX_PREV_EARLIEST_RIGHT_ANCESTOR]  = rulealt(prevEarliestAncestorRuleIndex,prevEarliestAncestorRuleAltNum);
		features[INDEX_CUR_TOKEN_TYPE]                = curToken.getType();
		features[INDEX_CUR_TOKEN_CHILD_INDEX]         = getChildIndexOrListMembership(node);
		features[INDEX_EARLIEST_LEFT_ANCESTOR]        = rulealt(earliestLeftAncestor);
		features[INDEX_ANCESTORS_CHILD_INDEX]         = getChildIndexOrListMembership(earliestLeftAncestor);
		features[INDEX_ANCESTORS_PARENT_RULE]         = earliestLeftAncestorParent!=null ? rulealt(earliestLeftAncestorParent) : -1;
		features[INDEX_ANCESTORS_PARENT_CHILD_INDEX]  = getChildIndexOrListMembership(earliestLeftAncestorParent);
		features[INDEX_ANCESTORS_PARENT2_RULE]        = earliestLeftAncestorParent2!=null ? rulealt(earliestLeftAncestorParent2) : -1;
		features[INDEX_ANCESTORS_PARENT2_CHILD_INDEX] = getChildIndexOrListMembership(earliestLeftAncestorParent2);
		features[INDEX_ANCESTORS_PARENT3_RULE]        = earliestLeftAncestorParent3!=null ? rulealt(earliestLeftAncestorParent3) : -1;
		features[INDEX_ANCESTORS_PARENT3_CHILD_INDEX] = getChildIndexOrListMembership(earliestLeftAncestorParent3);
		features[INDEX_ANCESTORS_PARENT4_RULE]        = earliestLeftAncestorParent4!=null ? rulealt(earliestLeftAncestorParent4) : -1;
		features[INDEX_ANCESTORS_PARENT4_CHILD_INDEX] = getChildIndexOrListMembership(earliestLeftAncestorParent4);
		features[INDEX_ANCESTORS_PARENT5_RULE]        = earliestLeftAncestorParent5!=null ? rulealt(earliestLeftAncestorParent5) : -1;
		features[INDEX_ANCESTORS_PARENT5_CHILD_INDEX] = getChildIndexOrListMembership(earliestLeftAncestorParent5);

		features[INDEX_MATCHING_TOKEN_STARTS_LINE] = getMatchingSymbolStartsLine(corpus, doc, node);