org.graylog.plugins.pipelineprocessor.parser.PipelineRuleParser Maven / Gradle / Ivy
/*
* Copyright (C) 2020 Graylog, Inc.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the Server Side Public License, version 1,
* as published by MongoDB, Inc.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* Server Side Public License for more details.
*
* You should have received a copy of the Server Side Public License
* along with this program. If not, see
* .
*/
package org.graylog.plugins.pipelineprocessor.parser;
import com.google.common.collect.ImmutableList;
import com.google.common.collect.ImmutableSortedSet;
import com.google.common.collect.Lists;
import com.google.common.collect.Maps;
import com.google.common.collect.Sets;
import com.swrve.ratelimitedlogger.RateLimitedLog;
import org.antlr.v4.runtime.ANTLRInputStream;
import org.antlr.v4.runtime.BaseErrorListener;
import org.antlr.v4.runtime.CommonTokenStream;
import org.antlr.v4.runtime.DefaultErrorStrategy;
import org.antlr.v4.runtime.ParserRuleContext;
import org.antlr.v4.runtime.RecognitionException;
import org.antlr.v4.runtime.Recognizer;
import org.antlr.v4.runtime.RuleContext;
import org.antlr.v4.runtime.Token;
import org.antlr.v4.runtime.tree.ParseTreeProperty;
import org.antlr.v4.runtime.tree.ParseTreeWalker;
import org.apache.commons.lang3.StringEscapeUtils;
import org.apache.mina.util.IdentityHashSet;
import org.graylog.plugins.pipelineprocessor.ast.Pipeline;
import org.graylog.plugins.pipelineprocessor.ast.Rule;
import org.graylog.plugins.pipelineprocessor.ast.Stage;
import org.graylog.plugins.pipelineprocessor.ast.exceptions.PrecomputeFailure;
import org.graylog.plugins.pipelineprocessor.ast.expressions.AdditionExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.AndExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.ArrayLiteralExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.BinaryExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.BooleanExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.BooleanValuedFunctionWrapper;
import org.graylog.plugins.pipelineprocessor.ast.expressions.ComparisonExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.DoubleExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.EqualityExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.Expression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.FieldAccessExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.FieldRefExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.FunctionExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.IndexedAccessExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.LogicalExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.LongExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.MapLiteralExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.MessageRefExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.MultiplicationExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.NotExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.OrExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.SignedExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.StringExpression;
import org.graylog.plugins.pipelineprocessor.ast.expressions.VarRefExpression;
import org.graylog.plugins.pipelineprocessor.ast.functions.Function;
import org.graylog.plugins.pipelineprocessor.ast.functions.FunctionArgs;
import org.graylog.plugins.pipelineprocessor.ast.functions.FunctionDescriptor;
import org.graylog.plugins.pipelineprocessor.ast.functions.ParameterDescriptor;
import org.graylog.plugins.pipelineprocessor.ast.statements.FunctionStatement;
import org.graylog.plugins.pipelineprocessor.ast.statements.Statement;
import org.graylog.plugins.pipelineprocessor.ast.statements.VarAssignStatement;
import org.graylog.plugins.pipelineprocessor.parser.errors.IncompatibleArgumentType;
import org.graylog.plugins.pipelineprocessor.parser.errors.IncompatibleIndexType;
import org.graylog.plugins.pipelineprocessor.parser.errors.IncompatibleType;
import org.graylog.plugins.pipelineprocessor.parser.errors.IncompatibleTypes;
import org.graylog.plugins.pipelineprocessor.parser.errors.InvalidFunctionArgument;
import org.graylog.plugins.pipelineprocessor.parser.errors.InvalidOperation;
import org.graylog.plugins.pipelineprocessor.parser.errors.MissingRequiredParam;
import org.graylog.plugins.pipelineprocessor.parser.errors.NonIndexableType;
import org.graylog.plugins.pipelineprocessor.parser.errors.OptionalParametersMustBeNamed;
import org.graylog.plugins.pipelineprocessor.parser.errors.ParseError;
import org.graylog.plugins.pipelineprocessor.parser.errors.SyntaxError;
import org.graylog.plugins.pipelineprocessor.parser.errors.UndeclaredFunction;
import org.graylog.plugins.pipelineprocessor.parser.errors.UndeclaredVariable;
import org.graylog.plugins.pipelineprocessor.parser.errors.WrongNumberOfArgs;
import org.graylog2.plugin.Message;
import org.joda.time.DateTime;
import org.joda.time.Duration;
import org.joda.time.Period;
import javax.inject.Inject;
import java.util.ArrayDeque;
import java.util.HashMap;
import java.util.List;
import java.util.Locale;
import java.util.Map;
import java.util.Objects;
import java.util.Set;
import java.util.concurrent.atomic.AtomicLong;
import static com.google.common.collect.ImmutableSortedSet.orderedBy;
import static java.util.Comparator.comparingInt;
import static java.util.stream.Collectors.toList;
import static org.graylog.plugins.pipelineprocessor.processors.PipelineInterpreter.getRateLimitedLog;
public class PipelineRuleParser {
private final FunctionRegistry functionRegistry;
private static AtomicLong uniqueId = new AtomicLong(0);
@Inject
public PipelineRuleParser(FunctionRegistry functionRegistry) {
this.functionRegistry = functionRegistry;
}
private static final RateLimitedLog log = getRateLimitedLog(PipelineRuleParser.class);
public static final ParseTreeWalker WALKER = ParseTreeWalker.DEFAULT;
public Rule parseRule(String rule, boolean silent) throws ParseException {
return parseRule("dummy" + uniqueId.getAndIncrement(), rule, silent);
}
/**
* Parses the given rule source and optionally generates a Java class for it if the classloader is not null.
*
* @param id the id of the rule, necessary to generate code
* @param rule rule source code
* @param silent don't emit status messages during parsing
* @return the parse rule
* @throws ParseException if a one or more parse errors occur
*/
public Rule parseRule(String id, String rule, boolean silent) throws ParseException {
final ParseContext parseContext = new ParseContext(silent);
final SyntaxErrorListener errorListener = new SyntaxErrorListener(parseContext);
final RuleLangLexer lexer = new RuleLangLexer(new ANTLRInputStream(rule));
lexer.removeErrorListeners();
lexer.addErrorListener(errorListener);
final RuleLangParser parser = new RuleLangParser(new CommonTokenStream(lexer));
parser.setErrorHandler(new DefaultErrorStrategy());
parser.removeErrorListeners();
parser.addErrorListener(errorListener);
final RuleLangParser.RuleDeclarationContext ruleDeclaration = parser.ruleDeclaration();
// parsing stages:
// 1. build AST nodes, checks for invalid var, function refs
// 2. type annotator: infer type information from var refs, func refs
// 3. checker: static type check w/ coercion nodes
// 4. optimizer: TODO
WALKER.walk(new RuleAstBuilder(parseContext), ruleDeclaration);
WALKER.walk(new RuleTypeAnnotator(parseContext), ruleDeclaration);
WALKER.walk(new RuleTypeChecker(parseContext), ruleDeclaration);
if (parseContext.getErrors().isEmpty()) {
return parseContext.getRules().get(0).withId(id);
}
throw new ParseException(parseContext.getErrors());
}
public List parsePipelines(String pipelines) throws ParseException {
final ParseContext parseContext = new ParseContext(false);
final SyntaxErrorListener errorListener = new SyntaxErrorListener(parseContext);
final RuleLangLexer lexer = new RuleLangLexer(new ANTLRInputStream(pipelines));
lexer.removeErrorListeners();
lexer.addErrorListener(errorListener);
final RuleLangParser parser = new RuleLangParser(new CommonTokenStream(lexer));
parser.setErrorHandler(new DefaultErrorStrategy());
parser.removeErrorListeners();
parser.addErrorListener(errorListener);
final RuleLangParser.PipelineDeclsContext pipelineDeclsContext = parser.pipelineDecls();
WALKER.walk(new PipelineAstBuilder(parseContext), pipelineDeclsContext);
if (parseContext.getErrors().isEmpty()) {
return parseContext.pipelines;
}
throw new ParseException(parseContext.getErrors());
}
public Pipeline parsePipeline(String id, String source) {
final ParseContext parseContext = new ParseContext(false);
final SyntaxErrorListener errorListener = new SyntaxErrorListener(parseContext);
final RuleLangLexer lexer = new RuleLangLexer(new ANTLRInputStream(source));
lexer.removeErrorListeners();
lexer.addErrorListener(errorListener);
final RuleLangParser parser = new RuleLangParser(new CommonTokenStream(lexer));
parser.setErrorHandler(new DefaultErrorStrategy());
parser.removeErrorListeners();
parser.addErrorListener(errorListener);
final RuleLangParser.PipelineContext pipelineContext = parser.pipeline();
WALKER.walk(new PipelineAstBuilder(parseContext), pipelineContext);
if (parseContext.getErrors().isEmpty()) {
final Pipeline pipeline = parseContext.pipelines.get(0);
return pipeline.withId(id);
}
throw new ParseException(parseContext.getErrors());
}
public static String unquote(String string, char quoteChar) {
if (string.length() >= 2 &&
string.charAt(0) == quoteChar && string.charAt(string.length() - 1) == quoteChar) {
return string.substring(1, string.length() - 1);
}
return string;
}
public static String unescape(String string) {
return StringEscapeUtils.unescapeJava(string);
}
private static class SyntaxErrorListener extends BaseErrorListener {
private final ParseContext parseContext;
public SyntaxErrorListener(ParseContext parseContext) {
this.parseContext = parseContext;
}
@Override
public void syntaxError(Recognizer recognizer,
Object offendingSymbol,
int line,
int charPositionInLine,
String msg,
RecognitionException e) {
parseContext.addError(new SyntaxError(offendingSymbol, line, charPositionInLine, msg, e));
}
}
private class RuleAstBuilder extends RuleLangBaseListener {
private final ParseContext parseContext;
private final ParseTreeProperty> args;
private final ParseTreeProperty> argsList;
private final ParseTreeProperty exprs;
private final Set definedVars = Sets.newHashSet();
// this is true for nested field accesses
private ArrayDeque isIdIsFieldAccess = new ArrayDeque<>();
public RuleAstBuilder(ParseContext parseContext) {
this.parseContext = parseContext;
args = parseContext.arguments();
argsList = parseContext.argumentLists();
exprs = parseContext.expressions();
isIdIsFieldAccess.push(false); // top of stack
}
@Override
public void exitRuleDeclaration(RuleLangParser.RuleDeclarationContext ctx) {
final Rule.Builder ruleBuilder = Rule.builder();
ruleBuilder.name(unquote(ctx.name == null ? "" : ctx.name.getText(), '"'));
final Expression expr = exprs.get(ctx.condition);
LogicalExpression condition;
if (expr instanceof LogicalExpression) {
condition = (LogicalExpression) expr;
} else if (expr != null && expr.getType().equals(Boolean.class)) {
condition = new BooleanValuedFunctionWrapper(ctx.getStart(), expr);
} else {
condition = new BooleanExpression(ctx.getStart(), false);
log.debug("Unable to create condition, replacing with 'false'");
}
ruleBuilder.when(condition);
ruleBuilder.then(parseContext.statements);
final Rule rule = ruleBuilder.build();
parseContext.addRule(rule);
log.trace("Declaring rule {}", rule);
}
@Override
public void exitFuncStmt(RuleLangParser.FuncStmtContext ctx) {
final Expression expr = exprs.get(ctx.functionCall());
final FunctionStatement functionStatement = new FunctionStatement(expr);
parseContext.statements.add(functionStatement);
}
@Override
public void exitVarAssignStmt(RuleLangParser.VarAssignStmtContext ctx) {
final String name = unquote(ctx.varName.getText(), '`');
final Expression expr = exprs.get(ctx.expression());
parseContext.defineVar(name, expr);
definedVars.add(name);
parseContext.statements.add(new VarAssignStatement(name, expr));
}
@Override
public void exitFunctionCall(RuleLangParser.FunctionCallContext ctx) {
final String name = ctx.funcName.getText();
Map argsMap = this.args.get(ctx.arguments());
final List positionalArgs = this.argsList.get(ctx.arguments());
final Function function = functionRegistry.resolve(name);
if (function == null) {
parseContext.addError(new UndeclaredFunction(ctx));
} else {
final ImmutableList params = function.descriptor().params();
final boolean hasOptionalParams = params.stream().anyMatch(ParameterDescriptor::optional);
if (argsMap != null) {
// check for the right number of arguments to the function if the function only has required params
if (!hasOptionalParams && params.size() != argsMap.size()) {
parseContext.addError(new WrongNumberOfArgs(ctx, function, argsMap.size()));
} else {
// there are optional parameters, check that all required ones are present
final Map givenArguments = argsMap;
final List missingParams =
params.stream()
.filter(p -> !p.optional())
.map(p -> givenArguments.containsKey(p.name()) ? null : p)
.filter(Objects::nonNull)
.collect(toList());
for (ParameterDescriptor param : missingParams) {
parseContext.addError(new MissingRequiredParam(ctx, function, param));
}
}
} else if (positionalArgs != null) {
// use descriptor to turn positional arguments into a map
argsMap = Maps.newHashMap();
// if we only have required parameters and the number doesn't match, complain
if (!hasOptionalParams && positionalArgs.size() != params.size()) {
parseContext.addError(new WrongNumberOfArgs(ctx, function, positionalArgs.size()));
}
// if optional parameters precede any required ones, the function must used named parameters
boolean hasError = false;
if (hasOptionalParams) {
// find the index of the first optional parameter
// then check if any non-optional come after it, if so, complain
int firstOptional = Integer.MAX_VALUE;
boolean requiredAfterOptional = false;
int i = 0;
for (ParameterDescriptor param : params) {
i++;
if (param.optional()) {
firstOptional = Math.min(firstOptional, i);
} else {
if (i > firstOptional) {
requiredAfterOptional = true;
}
}
}
if (requiredAfterOptional) {
parseContext.addError(new OptionalParametersMustBeNamed(ctx, function));
hasError = true;
} else {
final long numberRequiredParams = params.stream()
.filter(p -> !p.optional())
.count();
if (numberRequiredParams > positionalArgs.size()) {
parseContext.addError(new WrongNumberOfArgs(ctx, function, positionalArgs.size()));
hasError = true;
}
}
}
if (!hasError) {
// only try to assign params if we didn't encounter a problem with position optional params above
int i = 0;
for (ParameterDescriptor p : params) {
if (i >= positionalArgs.size()) {
// avoid index out of bounds, we've added an error anyway
// the remaining parameters were optional, so we can safely skip them
break;
}
final Expression argExpr = positionalArgs.get(i);
argsMap.put(p.name(), argExpr);
i++;
}
}
} else if (!params.stream().allMatch(ParameterDescriptor::optional)) {
// no parameters given but some of them are required
parseContext.addError(new WrongNumberOfArgs(ctx, function, 0));
}
}
FunctionExpression expr;
try {
expr = new FunctionExpression(
ctx.getStart(), new FunctionArgs(functionRegistry.resolveOrError(name), argsMap)
);
} catch (PrecomputeFailure precomputeFailure) {
parseContext.addError(new InvalidFunctionArgument(ctx, function, precomputeFailure));
expr = new FunctionExpression(ctx.getStart(), new FunctionArgs(Function.ERROR_FUNCTION, argsMap));
}
log.trace("FUNC: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitNamedArgs(RuleLangParser.NamedArgsContext ctx) {
final Map argMap = Maps.newHashMap();
for (RuleLangParser.PropAssignmentContext propAssignmentContext : ctx.propAssignment()) {
final String argName = unquote(propAssignmentContext.Identifier().getText(), '`');
final Expression argValue = exprs.get(propAssignmentContext.expression());
argMap.put(argName, argValue);
}
args.put(ctx, argMap);
}
@Override
public void exitPositionalArgs(RuleLangParser.PositionalArgsContext ctx) {
List expressions = Lists.newArrayListWithCapacity(ctx.expression().size());
expressions.addAll(ctx.expression().stream().map(exprs::get).collect(toList()));
argsList.put(ctx, expressions);
}
@Override
public void enterNested(RuleLangParser.NestedContext ctx) {
// nested field access is ok, these are not rule variables
isIdIsFieldAccess.push(true);
}
@Override
public void exitNested(RuleLangParser.NestedContext ctx) {
isIdIsFieldAccess.pop(); // reset for error checks
final Expression object = exprs.get(ctx.fieldSet);
final Expression field = exprs.get(ctx.field);
final FieldAccessExpression expr = new FieldAccessExpression(ctx.getStart(), object, field);
log.trace("FIELDACCESS: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitNot(RuleLangParser.NotContext ctx) {
final Expression expression = upgradeBoolFunctionExpression(ctx.expression());
final NotExpression expr = new NotExpression(ctx.getStart(), expression);
log.trace("NOT: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitAnd(RuleLangParser.AndContext ctx) {
// if the expressions are function calls but boolean valued, upgrade them,
// we allow testing boolean valued functions without explicit comparison operator
final Expression left = upgradeBoolFunctionExpression(ctx.left);
final Expression right = upgradeBoolFunctionExpression(ctx.right);
final AndExpression expr = new AndExpression(ctx.getStart(), left, right);
log.trace("AND: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
private Expression upgradeBoolFunctionExpression(RuleLangParser.ExpressionContext leftExprContext) {
Expression leftExpr = exprs.get(leftExprContext);
if (leftExpr instanceof FunctionExpression && leftExpr.getType().equals(Boolean.class)) {
leftExpr = new BooleanValuedFunctionWrapper(leftExprContext.getStart(), leftExpr);
}
return leftExpr;
}
@Override
public void exitOr(RuleLangParser.OrContext ctx) {
final Expression left = upgradeBoolFunctionExpression(ctx.left);
final Expression right = upgradeBoolFunctionExpression(ctx.right);
final OrExpression expr = new OrExpression(ctx.getStart(), left, right);
log.trace("OR: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitEquality(RuleLangParser.EqualityContext ctx) {
final Expression left = exprs.get(ctx.left);
final Expression right = exprs.get(ctx.right);
final boolean equals = ctx.equality.getText().equals("==");
final EqualityExpression expr = new EqualityExpression(ctx.getStart(), left, right, equals);
log.trace("EQUAL: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitComparison(RuleLangParser.ComparisonContext ctx) {
final Expression left = exprs.get(ctx.left);
final Expression right = exprs.get(ctx.right);
final String operator = ctx.comparison.getText();
final ComparisonExpression expr = new ComparisonExpression(ctx.getStart(), left, right, operator);
log.trace("COMPARE: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitInteger(RuleLangParser.IntegerContext ctx) {
// TODO handle different radix and length
final LongExpression expr = new LongExpression(ctx.getStart(), Long.parseLong(ctx.getText()));
log.trace("INT: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitFloat(RuleLangParser.FloatContext ctx) {
final DoubleExpression expr = new DoubleExpression(ctx.getStart(), Double.parseDouble(ctx.getText()));
log.trace("FLOAT: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitChar(RuleLangParser.CharContext ctx) {
// TODO
super.exitChar(ctx);
}
@Override
public void exitString(RuleLangParser.StringContext ctx) {
final String text = unescape(unquote(ctx.getText(), '\"'));
final StringExpression expr = new StringExpression(ctx.getStart(), text);
log.trace("STRING: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitBoolean(RuleLangParser.BooleanContext ctx) {
final BooleanExpression expr = new BooleanExpression(ctx.getStart(), Boolean.valueOf(ctx.getText()));
log.trace("BOOL: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitLiteralPrimary(RuleLangParser.LiteralPrimaryContext ctx) {
// nothing to do, just propagate the ConstantExpression
exprs.put(ctx, exprs.get(ctx.literal()));
parseContext.addInnerNode(ctx);
}
@Override
public void exitArrayLiteralExpr(RuleLangParser.ArrayLiteralExprContext ctx) {
final List elements = ctx.expression().stream().map(exprs::get).collect(toList());
exprs.put(ctx, new ArrayLiteralExpression(ctx.getStart(), elements));
}
@Override
public void exitMapLiteralExpr(RuleLangParser.MapLiteralExprContext ctx) {
final HashMap map = Maps.newHashMap();
for (RuleLangParser.PropAssignmentContext propAssignmentContext : ctx.propAssignment()) {
final String key = unquote(propAssignmentContext.Identifier().getText(), '`');
final Expression value = exprs.get(propAssignmentContext.expression());
map.put(key, value);
}
exprs.put(ctx, new MapLiteralExpression(ctx.getStart(), map));
}
@Override
public void exitParenExpr(RuleLangParser.ParenExprContext ctx) {
// nothing to do, just propagate
exprs.put(ctx, exprs.get(ctx.expression()));
parseContext.addInnerNode(ctx);
}
@Override
public void exitSignedExpression(RuleLangParser.SignedExpressionContext ctx) {
final Expression right = exprs.get(ctx.expr);
final boolean isPlus = ctx.sign.getText().equals("+");
final SignedExpression expr = new SignedExpression(ctx.getStart(), right, isPlus);
log.trace("SIGN: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitAddition(RuleLangParser.AdditionContext ctx) {
final Expression left = exprs.get(ctx.left);
final Expression right = exprs.get(ctx.right);
final boolean isPlus = ctx.add.getText().equals("+");
final AdditionExpression expr = new AdditionExpression(ctx.getStart(), left, right, isPlus);
log.trace("ADD: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitMultiplication(RuleLangParser.MultiplicationContext ctx) {
final Expression left = exprs.get(ctx.left);
final Expression right = exprs.get(ctx.right);
final char operator = ctx.mult.getText().charAt(0);
final MultiplicationExpression expr = new MultiplicationExpression(ctx.getStart(), left, right, operator);
log.trace("MULT: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void enterMessageRef(RuleLangParser.MessageRefContext ctx) {
// nested field access is ok, these are not rule variables
isIdIsFieldAccess.push(true);
}
@Override
public void exitMessageRef(RuleLangParser.MessageRefContext ctx) {
isIdIsFieldAccess.pop(); // reset for error checks
final Expression fieldExpr = exprs.get(ctx.field);
final MessageRefExpression expr = new MessageRefExpression(ctx.getStart(), fieldExpr);
log.trace("$MSG: ctx {} => {}", ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitIdentifier(RuleLangParser.IdentifierContext ctx) {
// unquote identifier if necessary
final String identifierName = unquote(ctx.Identifier().getText(), '`');
if (!isIdIsFieldAccess.peek() && !definedVars.contains(identifierName)) {
parseContext.addError(new UndeclaredVariable(ctx));
}
final Expression expr;
String type;
// if the identifier is also a declared variable name prefer the variable
if (isIdIsFieldAccess.peek() && !definedVars.contains(identifierName)) {
expr = new FieldRefExpression(ctx.getStart(), identifierName, parseContext.getDefinedVar(identifierName));
type = "FIELDREF";
} else {
expr = new VarRefExpression(ctx.getStart(), identifierName, parseContext.getDefinedVar(identifierName));
type = "VARREF";
}
log.trace("{}: ctx {} => {}", type, ctx, expr);
exprs.put(ctx, expr);
}
@Override
public void exitFunc(RuleLangParser.FuncContext ctx) {
// nothing to do, just propagate
exprs.put(ctx, exprs.get(ctx.functionCall()));
parseContext.addInnerNode(ctx);
}
@Override
public void exitIndexedAccess(RuleLangParser.IndexedAccessContext ctx) {
final Expression array = exprs.get(ctx.array);
final Expression index = exprs.get(ctx.index);
final IndexedAccessExpression expr = new IndexedAccessExpression(ctx.getStart(), array, index);
exprs.put(ctx, expr);
log.trace("IDXACCESS: ctx {} => {}", ctx, expr);
}
}
private static class RuleTypeAnnotator extends RuleLangBaseListener {
private final ParseContext parseContext;
public RuleTypeAnnotator(ParseContext parseContext) {
this.parseContext = parseContext;
}
@Override
public void exitIdentifier(RuleLangParser.IdentifierContext ctx) {
final Expression expr = parseContext.expressions().get(ctx);
if (expr instanceof VarRefExpression) {
final VarRefExpression varRefExpression = (VarRefExpression) expr;
final String name = varRefExpression.varName();
final Expression expression = parseContext.getDefinedVar(name);
if (expression == null) {
if (parseContext.isSilent()) {
log.debug("Unable to retrieve expression for variable {}, this is a bug", name);
} else {
log.error("Unable to retrieve expression for variable {}, this is a bug", name);
}
return;
}
log.trace("Inferred type of variable {} to {}", name, expression.getType().getSimpleName());
varRefExpression.setType(expression.getType());
}
}
@Override
public void exitAddition(RuleLangParser.AdditionContext ctx) {
final AdditionExpression expr = (AdditionExpression) parseContext.expressions().get(ctx);
final Class leftType = expr.left().getType();
final Class rightType = expr.right().getType();
if (leftType.equals(rightType)) {
// propagate left type
expr.setType(leftType);
} else if (DateTime.class.equals(leftType) && DateTime.class.equals(rightType)) {
// fine to subtract two dates from each other, this results in a Duration
expr.setType(Duration.class);
} else if (DateTime.class.equals(leftType) && Period.class.equals(rightType) || Period.class.equals(leftType) && DateTime.class.equals(rightType)) {
expr.setType(DateTime.class);
} else {
// this will be detected as an error later
expr.setType(Void.class);
}
}
@Override
public void exitMultiplication(RuleLangParser.MultiplicationContext ctx) {
final MultiplicationExpression expr = (MultiplicationExpression) parseContext.expressions().get(ctx);
final Class leftType = expr.left().getType();
final Class rightType = expr.right().getType();
if (leftType.equals(rightType)) {
// propagate left type
expr.setType(leftType);
} else {
// this will be detected as an error later
expr.setType(Void.class);
}
}
}
private static class RuleTypeChecker extends RuleLangBaseListener {
private final ParseContext parseContext;
@SuppressWarnings("JdkObsolete")
StringBuffer sb = new StringBuffer();
public RuleTypeChecker(ParseContext parseContext) {
this.parseContext = parseContext;
}
@Override
public void exitRuleDeclaration(RuleLangParser.RuleDeclarationContext ctx) {
log.trace("Type tree {}", sb.toString());
}
@Override
public void exitAnd(RuleLangParser.AndContext ctx) {
checkBinaryExpression(ctx);
}
@Override
public void exitOr(RuleLangParser.OrContext ctx) {
checkBinaryExpression(ctx);
}
@Override
public void exitNot(RuleLangParser.NotContext ctx) {
final Expression expression = parseContext.expressions().get(ctx.expression());
Class type = expression.getType();
if (!Boolean.class.isAssignableFrom(type)) {
parseContext.addError(new IncompatibleType(ctx, Boolean.class, type));
}
}
@Override
public void exitComparison(RuleLangParser.ComparisonContext ctx) {
checkBinaryExpression(ctx);
}
@Override
public void exitAddition(RuleLangParser.AdditionContext ctx) {
final AdditionExpression addExpression = (AdditionExpression) parseContext.expressions().get(ctx);
final Class leftType = addExpression.left().getType();
final Class rightType = addExpression.right().getType();
// special case for DateTime/Period, which are all compatible
final boolean leftDate = DateTime.class.equals(leftType);
final boolean rightDate = DateTime.class.equals(rightType);
final boolean leftPeriod = Period.class.equals(leftType);
final boolean rightPeriod = Period.class.equals(rightType);
final boolean leftString = String.class.equals(leftType);
final boolean rightString = String.class.equals(rightType);
if (leftDate && rightDate) {
if (addExpression.isPlus()) {
parseContext.addError(new InvalidOperation(ctx, addExpression, "Unable to add two dates"));
}
return;
} else if (leftDate && rightPeriod || leftPeriod && rightDate || leftPeriod && rightPeriod) {
return;
} else if (leftString && rightString) {
if (!addExpression.isPlus()) {
parseContext.addError(new InvalidOperation(ctx, addExpression, "Unable to subtract two strings"));
}
return;
}
// otherwise check generic binary expression
checkBinaryExpression(ctx);
}
@Override
public void exitMultiplication(RuleLangParser.MultiplicationContext ctx) {
checkBinaryExpression(ctx);
}
@Override
public void exitEquality(RuleLangParser.EqualityContext ctx) {
// TODO equality allows arbitrary types, in the future optimize by using specialized operators
}
private void checkBinaryExpression(RuleLangParser.ExpressionContext ctx) {
final BinaryExpression binaryExpr = (BinaryExpression) parseContext.expressions().get(ctx);
final Class leftType = binaryExpr.left().getType();
final Class rightType = binaryExpr.right().getType();
if (!leftType.equals(rightType) || Void.class.equals(leftType) || Void.class.equals(rightType)) {
parseContext.addError(new IncompatibleTypes(ctx, binaryExpr));
}
}
@Override
public void exitFunctionCall(RuleLangParser.FunctionCallContext ctx) {
final FunctionExpression expr = (FunctionExpression) parseContext.expressions().get(ctx);
final FunctionDescriptor descriptor = expr.getFunction().descriptor();
final FunctionArgs args = expr.getArgs();
for (ParameterDescriptor p : descriptor.params()) {
final Expression argExpr = args.expression(p.name());
if (argExpr != null && !p.type().isAssignableFrom(argExpr.getType())) {
parseContext.addError(new IncompatibleArgumentType(ctx, expr, p, argExpr));
}
}
}
@Override
public void exitMessageRef(RuleLangParser.MessageRefContext ctx) {
final MessageRefExpression expr = (MessageRefExpression) parseContext.expressions().get(ctx);
final Class type = expr.getFieldExpr().getType();
if (!type.equals(String.class) && !type.equals(Message.class)) {
parseContext.addError(new IncompatibleType(ctx, String.class, type));
}
}
@Override
public void enterEveryRule(ParserRuleContext ctx) {
final Expression expression = parseContext.expressions().get(ctx);
if (expression != null && !parseContext.isInnerNode(ctx)) {
sb.append(" ( ");
sb.append(expression.getClass().getSimpleName());
sb.append(":").append(ctx.getClass().getSimpleName()).append(" ");
sb.append(" <").append(expression.getType().getSimpleName()).append("> ");
sb.append(ctx.getText());
}
}
@Override
public void exitEveryRule(ParserRuleContext ctx) {
final Expression expression = parseContext.expressions().get(ctx);
if (expression != null && !parseContext.isInnerNode(ctx)) {
sb.append(" ) ");
}
}
@Override
public void exitIndexedAccess(RuleLangParser.IndexedAccessContext ctx) {
final IndexedAccessExpression idxExpr = (IndexedAccessExpression) parseContext.expressions().get(
ctx);
final Class indexableType = idxExpr.getIndexableObject().getType();
final Class indexType = idxExpr.getIndex().getType();
final boolean isMap = Map.class.isAssignableFrom(indexableType);
if (indexableType.isArray()
|| List.class.isAssignableFrom(indexableType)
|| Iterable.class.isAssignableFrom(indexableType)
|| isMap) {
// then check if the index type is compatible, must be long for array-like and string for map-like types
if (isMap) {
if (!String.class.equals(indexType)) {
// add type error
parseContext.addError(new IncompatibleIndexType(ctx, String.class, indexType));
}
} else {
if (!Long.class.equals(indexType)) {
parseContext.addError(new IncompatibleIndexType(ctx, Long.class, indexType));
}
}
} else {
// not an indexable type
parseContext.addError(new NonIndexableType(ctx, indexableType));
}
}
}
/**
* Contains meta data about the parse tree, such as AST nodes, link to the function registry etc.
*
* Being used by tree walkers or visitors to perform AST construction, type checking and so on.
*/
private static class ParseContext {
private final ParseTreeProperty exprs = new ParseTreeProperty<>();
private final ParseTreeProperty> args = new ParseTreeProperty<>();
/**
* Should the parser be more silent about its error logging, useful for interactive parsing in the UI.
*/
private final boolean silent;
private ParseTreeProperty> argsLists = new ParseTreeProperty<>();
private Set errors = Sets.newHashSet();
// inner nodes in the parse tree will be ignored during type checker printing, they only transport type information
private Set innerNodes = new IdentityHashSet<>();
public List statements = Lists.newArrayList();
public List rules = Lists.newArrayList();
private Map varDecls = Maps.newHashMap();
public List pipelines = Lists.newArrayList();
public ParseContext(boolean silent) {
this.silent = silent;
}
public ParseTreeProperty expressions() {
return exprs;
}
public ParseTreeProperty> arguments() {
return args;
}
public List getRules() {
return rules;
}
public void setRules(List rules) {
this.rules = rules;
}
public void addRule(Rule rule) {
this.rules.add(rule);
}
public List getPipelines() {
return pipelines;
}
public Set getErrors() {
return errors;
}
public void addError(ParseError error) {
errors.add(error);
}
public void addInnerNode(RuleContext node) {
innerNodes.add(node);
}
public boolean isInnerNode(RuleContext node) {
return innerNodes.contains(node);
}
/**
* Links the declared var to its expression.
*
* @param name var name
* @param expr expression
* @return true if successful, false if previously declared
*/
public boolean defineVar(String name, Expression expr) {
return varDecls.put(name, expr) == null;
}
public Expression getDefinedVar(String name) {
return varDecls.get(name);
}
public ParseTreeProperty> argumentLists() {
return argsLists;
}
public boolean isSilent() {
return silent;
}
}
private static class PipelineAstBuilder extends RuleLangBaseListener {
private final ParseContext parseContext;
public PipelineAstBuilder(ParseContext parseContext) {
this.parseContext = parseContext;
}
@Override
public void exitPipelineDeclaration(RuleLangParser.PipelineDeclarationContext ctx) {
final Pipeline.Builder builder = Pipeline.builder();
builder.name(unquote(ctx.name.getText(), '"'));
final ImmutableSortedSet.Builder stages = orderedBy(comparingInt(Stage::stage));
for (RuleLangParser.StageDeclarationContext stage : ctx.stageDeclaration()) {
final Stage.Builder stageBuilder = Stage.builder();
final Token stageToken = stage.stage;
if (stageToken == null) {
parseContext.addError(new SyntaxError(null, 0, 0, "", null));
return;
}
final int stageNumber = Integer.parseInt(stageToken.getText());
stageBuilder.stage(stageNumber);
final Token modifier = stage.modifier;
if (modifier == null) {
parseContext.addError(new SyntaxError(null, stageToken.getLine(), stageToken.getCharPositionInLine(), "", null));
return;
}
try {
stageBuilder.match(Stage.Match.valueOf(modifier.getText().toUpperCase(Locale.ROOT)));
} catch (IllegalArgumentException e) {
parseContext.addError(new SyntaxError(null, stageToken.getLine(), stageToken.getCharPositionInLine(), "", null));
return;
}
final List ruleRefs = stage.ruleRef().stream()
.map(ruleRefContext -> {
final Token name = ruleRefContext.name;
if (name == null) {
final Token symbol = ruleRefContext.Rule().getSymbol();
parseContext.addError(new SyntaxError(symbol, symbol.getLine(), symbol.getCharPositionInLine(), "invalid rule reference", null));
return "__illegal_reference";
}
return unquote(name.getText(), '"');
})
.collect(toList());
stageBuilder.ruleReferences(ruleRefs);
stages.add(stageBuilder.build());
}
builder.stages(stages.build());
parseContext.pipelines.add(builder.build());
}
@Override
public void exitInteger(RuleLangParser.IntegerContext ctx) {
// TODO handle different radix and length
final LongExpression expr = new LongExpression(ctx.getStart(), Long.parseLong(ctx.getText()));
log.trace("INT: ctx {} => {}", ctx, expr);
parseContext.exprs.put(ctx, expr);
}
@Override
public void exitString(RuleLangParser.StringContext ctx) {
final String text = unescape(unquote(ctx.getText(), '\"'));
final StringExpression expr = new StringExpression(ctx.getStart(), text);
log.trace("STRING: ctx {} => {}", ctx, expr);
parseContext.exprs.put(ctx, expr);
}
}
}
© 2015 - 2024 Weber Informatics LLC | Privacy Policy