# Copyright (c) 2012 The Chromium Authors. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. """Checks Java files for illegal imports.""" import codecs import os import re import results from rules import Rule class JavaChecker(object): """Import checker for Java files. The CheckFile method uses real filesystem paths, but Java imports work in terms of package names. To deal with this, we have an extra "prescan" pass that reads all the .java files and builds a mapping of class name -> filepath. In CheckFile, we convert each import statement into a real filepath, and check that against the rules in the DEPS files. Note that in Java you can always use classes in the same directory without an explicit import statement, so these imports can't be blocked with DEPS files. But that shouldn't be a problem, because same-package imports are pretty much always correct by definition. (If we find a case where this is *not* correct, it probably means the package is too big and needs to be split up.) Properties: _classmap: dict of fully-qualified Java class name -> filepath """ EXTENSIONS = ['.java'] def __init__(self, base_directory, verbose): self._base_directory = base_directory self._verbose = verbose self._classmap = {} self._PrescanFiles() def _PrescanFiles(self): for root, dirs, files in os.walk(self._base_directory): # Skip unwanted subdirectories. TODO(husky): it would be better to do # this via the skip_child_includes flag in DEPS files. Maybe hoist this # prescan logic into checkdeps.py itself? for d in dirs: # Skip hidden directories. if d.startswith('.'): dirs.remove(d) # Skip the "out" directory, as dealing with generated files is awkward. # We don't want paths like "out/Release/lib.java" in our DEPS files. # TODO(husky): We need some way of determining the "real" path to # a generated file -- i.e., where it would be in source control if # it weren't generated. if d == 'out': dirs.remove(d) # Skip third-party directories. if d == 'third_party': dirs.remove(d) for f in files: if f.endswith('.java'): self._PrescanFile(os.path.join(root, f)) def _PrescanFile(self, filepath): if self._verbose: print 'Prescanning: ' + filepath with codecs.open(filepath, encoding='utf-8') as f: short_class_name, _ = os.path.splitext(os.path.basename(filepath)) for line in f: for package in re.findall('^package ([\w\.]+);', line): full_class_name = package + '.' + short_class_name if full_class_name in self._classmap: print 'WARNING: multiple definitions of %s:' % full_class_name print ' ' + filepath print ' ' + self._classmap[full_class_name] print else: self._classmap[full_class_name] = filepath return print 'WARNING: no package definition found in %s' % filepath def CheckFile(self, rules, filepath): if self._verbose: print 'Checking: ' + filepath dependee_status = results.DependeeStatus(filepath) with codecs.open(filepath, encoding='utf-8') as f: for line in f: for clazz in re.findall('^import\s+(?:static\s+)?([\w\.]+)\s*;', line): if clazz not in self._classmap: # Importing a class from outside the Chromium tree. That's fine -- # it's probably a Java or Android system class. continue include_path = os.path.relpath( self._classmap[clazz], self._base_directory) # Convert Windows paths to Unix style, as used in DEPS files. include_path = include_path.replace(os.path.sep, '/') rule = rules.RuleApplyingTo(include_path, filepath) if rule.allow == Rule.DISALLOW: dependee_status.AddViolation( results.DependencyViolation(include_path, rule, rules)) if '{' in line: # This is code, so we're finished reading imports for this file. break return dependee_status