All Downloads are FREE. Search and download functionalities are using the official Maven repository.

org.sonar.plugins.python.PythonScanner Maven / Gradle / Ivy

The newest version!
/*
 * SonarQube Python Plugin
 * Copyright (C) 2011-2024 SonarSource SA
 * mailto:info AT sonarsource DOT com
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 3 of the License, or (at your option) any later version.
 *
 * 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 GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public License
 * along with this program; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
 */
package org.sonar.plugins.python;

import com.sonar.sslr.api.AstNode;
import com.sonar.sslr.api.RecognitionException;
import java.io.File;
import java.io.IOException;
import java.util.ArrayDeque;
import java.util.ArrayList;
import java.util.Collections;
import java.util.Deque;
import java.util.HashMap;
import java.util.HashSet;
import java.util.List;
import java.util.Map;
import java.util.Optional;
import java.util.Set;
import javax.annotation.CheckForNull;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.sonar.api.SonarProduct;
import org.sonar.api.batch.fs.InputFile;
import org.sonar.api.batch.fs.TextRange;
import org.sonar.api.batch.sensor.SensorContext;
import org.sonar.api.batch.sensor.issue.NewIssue;
import org.sonar.api.batch.sensor.issue.NewIssueLocation;
import org.sonar.api.issue.NoSonarFilter;
import org.sonar.api.measures.CoreMetrics;
import org.sonar.api.measures.FileLinesContext;
import org.sonar.api.measures.FileLinesContextFactory;
import org.sonar.api.measures.Metric;
import org.sonar.api.rule.RuleKey;
import org.sonar.plugins.python.api.IssueLocation;
import org.sonar.plugins.python.api.PythonCheck;
import org.sonar.plugins.python.api.PythonCheck.PreciseIssue;
import org.sonar.plugins.python.api.PythonFile;
import org.sonar.plugins.python.api.PythonInputFileContext;
import org.sonar.plugins.python.api.PythonSubscriptionCheck;
import org.sonar.plugins.python.api.PythonVisitorContext;
import org.sonar.plugins.python.api.internal.EndOfAnalysis;
import org.sonar.plugins.python.api.quickfix.PythonQuickFix;
import org.sonar.plugins.python.api.quickfix.PythonTextEdit;
import org.sonar.plugins.python.api.tree.FileInput;
import org.sonar.plugins.python.cpd.PythonCpdAnalyzer;
import org.sonar.plugins.python.indexer.PythonIndexer;
import org.sonar.python.IPythonLocation;
import org.sonar.python.SubscriptionVisitor;
import org.sonar.python.metrics.FileLinesVisitor;
import org.sonar.python.metrics.FileMetrics;
import org.sonar.python.parser.PythonParser;
import org.sonar.python.tree.IPythonTreeMaker;
import org.sonar.python.tree.PythonTreeMaker;

public class PythonScanner extends Scanner {

  private static final Logger LOG = LoggerFactory.getLogger(PythonScanner.class);

  private final PythonParser parser;
  private final PythonChecks checks;
  private final FileLinesContextFactory fileLinesContextFactory;
  private final NoSonarFilter noSonarFilter;
  private final PythonCpdAnalyzer cpdAnalyzer;
  private final PythonIndexer indexer;
  private final Map> checksExecutedWithoutParsingByFiles = new HashMap<>();

  public PythonScanner(
    SensorContext context, PythonChecks checks,
    FileLinesContextFactory fileLinesContextFactory, NoSonarFilter noSonarFilter, PythonParser parser, PythonIndexer indexer) {
    super(context);
    this.checks = checks;
    this.fileLinesContextFactory = fileLinesContextFactory;
    this.noSonarFilter = noSonarFilter;
    this.cpdAnalyzer = new PythonCpdAnalyzer(context);
    this.parser = parser;
    this.indexer = indexer;
    this.indexer.buildOnce(context);
  }

  @Override
  protected String name() {
    return "rules execution";
  }

  @Override
  protected void scanFile(PythonInputFile inputFile) throws IOException {
    PythonFile pythonFile = SonarQubePythonFile.create(inputFile.wrappedFile());
    PythonVisitorContext visitorContext;
    InputFile.Type fileType = inputFile.wrappedFile().type();
    try {
      AstNode astNode = parser.parse(inputFile.contents());
      PythonTreeMaker treeMaker = getTreeMaker(inputFile);
      FileInput parse = treeMaker.fileInput(astNode);
      visitorContext = new PythonVisitorContext(parse,
        pythonFile,
        getWorkingDirectory(context),
        indexer.packageName(inputFile),
        indexer.projectLevelSymbolTable(),
        indexer.cacheContext(),
        context.runtime().getProduct());
      if (fileType == InputFile.Type.MAIN) {
        saveMeasures(inputFile, visitorContext);
      }
    } catch (RecognitionException e) {
      visitorContext = new PythonVisitorContext(pythonFile, e, context.runtime().getProduct());

      var line = (inputFile.kind() == PythonInputFile.Kind.IPYTHON) ? ((GeneratedIPythonFile) inputFile).locationMap().get(e.getLine()).line() : e.getLine();
      var newMessage = e.getMessage().replace("line " + e.getLine(), "line " + line);

      LOG.error("Unable to parse file: " + inputFile);
      LOG.error(newMessage);
      context.newAnalysisError()
        .onFile(inputFile.wrappedFile())
        .at(inputFile.wrappedFile().newPointer(line, 0))
        .message(newMessage)
        .save();
    }
    List checksBasedOnTree = new ArrayList<>();
    for (PythonCheck check : checks.all()) {
      if (!isCheckApplicable(check, fileType)
        || checksExecutedWithoutParsingByFiles.getOrDefault(inputFile, Collections.emptySet()).contains(check)) {
        continue;
      }
      if (check instanceof PythonSubscriptionCheck pythonSubscriptionCheck) {
        checksBasedOnTree.add(pythonSubscriptionCheck);
      } else {
        check.scanFile(visitorContext);
      }
    }
    SubscriptionVisitor.analyze(checksBasedOnTree, visitorContext);
    saveIssues(inputFile, visitorContext.getIssues());

    if (visitorContext.rootTree() != null && !isInSonarLint(context)) {
      new SymbolVisitor(context.newSymbolTable().onFile(inputFile.wrappedFile())).visitFileInput(visitorContext.rootTree());
      new PythonHighlighter(context, inputFile).scanFile(visitorContext);
    }
  }

  private static PythonTreeMaker getTreeMaker(PythonInputFile inputFile) {
    return Python.KEY.equals(inputFile.wrappedFile().language()) ? new PythonTreeMaker() : new IPythonTreeMaker(getOffsetLocations(inputFile));
  }

  private static Map getOffsetLocations(PythonInputFile inputFile) {
    if(inputFile.kind() == PythonInputFile.Kind.IPYTHON){
      return ((GeneratedIPythonFile) inputFile).locationMap();
    }
    return Map.of();
  }

  @Override
  public boolean scanFileWithoutParsing(PythonInputFile inputFile) {
    InputFile.Type fileType = inputFile.wrappedFile().type();
    boolean result = true;
    for (PythonCheck check : checks.all()) {
      if (!isCheckApplicable(check, fileType)) {
        continue;
      }
      if (checkRequiresParsingOfImpactedFile(inputFile, check)) {
        // For regular Python checks, only directly modified files need to be analyzed
        // For DBD and Security, transitively impacted files must be re-analyzed.
        result = false;
        continue;
      }
      PythonFile pythonFile = SonarQubePythonFile.create(inputFile.wrappedFile());
      PythonInputFileContext inputFileContext = new PythonInputFileContext(pythonFile, context.fileSystem().workDir(), indexer.cacheContext(), context.runtime().getProduct());
      if (check.scanWithoutParsing(inputFileContext)) {
        Set executedChecks = checksExecutedWithoutParsingByFiles.getOrDefault(inputFile, new HashSet<>());
        executedChecks.add(check);
        checksExecutedWithoutParsingByFiles.putIfAbsent(inputFile, executedChecks);
      } else {
        result = false;
      }
    }
    if (!result) {
      // If scan without parsing is not successful, measures will be pushed during regular scan.
      // We must avoid pushing measures twice due to the risk of duplicate cache key error.
      return false;
    }
    return restoreAndPushMeasuresIfApplicable(inputFile);
  }

  private boolean checkRequiresParsingOfImpactedFile(PythonInputFile inputFile, PythonCheck check) {
    return !indexer.canBeFullyScannedWithoutParsing(inputFile) && !check.getClass().getPackageName().startsWith("org.sonar.python.checks");
  }

  @Override
  public void endOfAnalysis() {
    checks.all().stream()
      .filter(EndOfAnalysis.class::isInstance)
      .map(EndOfAnalysis.class::cast)
      .forEach(c -> c.endOfAnalysis(indexer.cacheContext()));
  }

  boolean isCheckApplicable(PythonCheck pythonCheck, InputFile.Type fileType) {
    PythonCheck.CheckScope checkScope = pythonCheck.scope();
    if (checkScope == PythonCheck.CheckScope.ALL) {
      return true;
    }
    return fileType == InputFile.Type.MAIN;
  }

  // visible for testing
  static File getWorkingDirectory(SensorContext context) {
    return isInSonarLint(context) ? null : context.fileSystem().workDir();
  }

  private static boolean isInSonarLint(SensorContext context) {
    return context.runtime().getProduct().equals(SonarProduct.SONARLINT);
  }

  @Override
  protected void processException(Exception e, PythonInputFile file) {
    LOG.warn("Unable to analyze file: " + file, e);
  }

  @Override
  public boolean canBeScannedWithoutParsing(PythonInputFile inputFile) {
    return this.indexer.canBePartiallyScannedWithoutParsing(inputFile);
  }

  @Override
  protected void reportStatistics(int numSkippedFiles, int numTotalFiles) {
    LOG.info("The Python analyzer was able to leverage cached data from previous analyses for {} out of {} files. These files were not parsed.",
      numSkippedFiles, numTotalFiles);
  }

  private void saveIssues(PythonInputFile inputFile, List issues) {
    for (PreciseIssue preciseIssue : issues) {
      RuleKey ruleKey = checks.ruleKey(preciseIssue.check());
      NewIssue newIssue = context
        .newIssue()
        .forRule(ruleKey);

      Integer cost = preciseIssue.cost();
      if (cost != null) {
        newIssue.gap(cost.doubleValue());
      }

      NewIssueLocation primaryLocation = newLocation(inputFile, newIssue, preciseIssue.primaryLocation());
      newIssue.at(primaryLocation);

      Deque secondaryLocationsFlow = new ArrayDeque<>();

      for (IssueLocation secondaryLocation : preciseIssue.secondaryLocations()) {
        String fileId = secondaryLocation.fileId();
        if (fileId != null) {
          InputFile issueLocationFile = component(fileId, context);
          if (issueLocationFile != null) {
            secondaryLocationsFlow.addFirst(newLocation(new PythonInputFileImpl(issueLocationFile), newIssue, secondaryLocation));
          }
        } else {
          newIssue.addLocation(newLocation(inputFile, newIssue, secondaryLocation));
        }
      }

      // secondary locations on multiple files are only supported using flows
      if (!secondaryLocationsFlow.isEmpty()) {
        secondaryLocationsFlow.addFirst(primaryLocation);
        newIssue.addFlow(secondaryLocationsFlow);
      }

      handleQuickFixes(inputFile.wrappedFile(), ruleKey, newIssue, preciseIssue);

      newIssue.save();
    }
  }

  @CheckForNull
  private InputFile component(String fileId, SensorContext sensorContext) {
    InputFile inputFile = Optional.ofNullable(sensorContext.fileSystem().inputFile(sensorContext.fileSystem().predicates().is(new File(fileId))))
      .orElseGet(() -> indexer.getFileWithId(fileId));
    if (inputFile == null) {
      LOG.debug("Failed to find InputFile for {}", fileId);
    }
    return inputFile;
  }

  private static NewIssueLocation newLocation(PythonInputFile inputFile, NewIssue issue, IssueLocation location) {
    NewIssueLocation newLocation = issue.newLocation()
      .on(inputFile.wrappedFile());
    if (location.startLine() != IssueLocation.UNDEFINED_LINE) {
      TextRange range;
      if (location.startLineOffset() == IssueLocation.UNDEFINED_OFFSET) {
        range = inputFile.wrappedFile().selectLine(location.startLine());
      } else {
        range = inputFile.wrappedFile().newRange(location.startLine(), location.startLineOffset(), location.endLine(), location.endLineOffset());
      }
      newLocation.at(range);
    }

    String message = location.message();
    if (message != null) {
      newLocation.message(message);
    }
    return newLocation;
  }

  private void saveMeasures(PythonInputFile inputFile, PythonVisitorContext visitorContext) {
    FileMetrics fileMetrics = new FileMetrics(visitorContext, isNotebook(inputFile));
    FileLinesVisitor fileLinesVisitor = fileMetrics.fileLinesVisitor();

    noSonarFilter.noSonarInFile(inputFile.wrappedFile(), fileLinesVisitor.getLinesWithNoSonar());

    if (!isInSonarLint(context)) {
      if (inputFile.kind() == PythonInputFile.Kind.PYTHON) {
        cpdAnalyzer.pushCpdTokens(inputFile.wrappedFile(), visitorContext);
      }

      Set linesOfCode = fileLinesVisitor.getLinesOfCode();
      saveMetricOnFile(inputFile, CoreMetrics.NCLOC, linesOfCode.size());
      saveMetricOnFile(inputFile, CoreMetrics.STATEMENTS, fileMetrics.numberOfStatements());
      saveMetricOnFile(inputFile, CoreMetrics.FUNCTIONS, fileMetrics.numberOfFunctions());
      saveMetricOnFile(inputFile, CoreMetrics.CLASSES, fileMetrics.numberOfClasses());
      saveMetricOnFile(inputFile, CoreMetrics.COMPLEXITY, fileMetrics.complexity());
      saveMetricOnFile(inputFile, CoreMetrics.COGNITIVE_COMPLEXITY, fileMetrics.cognitiveComplexity());
      saveMetricOnFile(inputFile, CoreMetrics.COMMENT_LINES, fileLinesVisitor.getCommentLineCount());

      FileLinesContext fileLinesContext = fileLinesContextFactory.createFor(inputFile.wrappedFile());
      if (inputFile.kind() == PythonInputFile.Kind.PYTHON) {
        for (int line : linesOfCode) {
          fileLinesContext.setIntValue(CoreMetrics.NCLOC_DATA_KEY, line, 1);
        }
      }
      for (int line : fileLinesVisitor.getExecutableLines()) {
        fileLinesContext.setIntValue(CoreMetrics.EXECUTABLE_LINES_DATA_KEY, line, 1);
      }
      fileLinesContext.save();
    }
  }

  static boolean isNotebook(PythonInputFile inputFile) {
    return inputFile.kind() == PythonInputFile.Kind.IPYTHON;
  }

  private boolean restoreAndPushMeasuresIfApplicable(PythonInputFile inputFile) {
    if (inputFile.wrappedFile().type() == InputFile.Type.TEST) {
      return true;
    }

    return cpdAnalyzer.pushCachedCpdTokens(inputFile.wrappedFile(), indexer.cacheContext());
  }

  private void saveMetricOnFile(PythonInputFile inputFile, Metric metric, Integer value) {
    context.newMeasure()
      .withValue(value)
      .forMetric(metric)
      .on(inputFile.wrappedFile())
      .save();
  }

  private void handleQuickFixes(InputFile inputFile, RuleKey ruleKey, NewIssue newIssue, PreciseIssue preciseIssue) {
    if (isInSonarLint(context)) {
      List quickFixes = preciseIssue.quickFixes();
      addQuickFixes(inputFile, ruleKey, quickFixes, newIssue);
    }
  }

  private static void addQuickFixes(InputFile inputFile, RuleKey ruleKey, Iterable quickFixes, NewIssue sonarLintIssue) {
    try {
      for (PythonQuickFix quickFix : quickFixes) {
        var newQuickFix = sonarLintIssue.newQuickFix()
          .message(quickFix.getDescription());

        var edit = newQuickFix.newInputFileEdit().on(inputFile);

        quickFix.getTextEdits().stream()
          .map(pythonTextEdit -> edit.newTextEdit().at(rangeFromTextSpan(inputFile, pythonTextEdit))
            .withNewText(pythonTextEdit.replacementText()))
          .forEach(edit::addTextEdit);
        newQuickFix.addInputFileEdit(edit);
        sonarLintIssue.addQuickFix(newQuickFix);
      }
      // TODO : is this try/catch still necessary ?
    } catch (RuntimeException e) {
      // We still want to report the issue if we did not manage to create a quick fix.
      LOG.warn(String.format("Could not report quick fixes for rule: %s. %s: %s", ruleKey, e.getClass().getName(), e.getMessage()));
    }
  }

  private static TextRange rangeFromTextSpan(InputFile file, PythonTextEdit pythonTextEdit) {
    return file.newRange(pythonTextEdit.startLine(), pythonTextEdit.startLineOffset(), pythonTextEdit.endLine(), pythonTextEdit.endLineOffset());
  }
}




© 2015 - 2024 Weber Informatics LLC | Privacy Policy