Recursively scans a directory for text files and finds those files not mentioned by name in any file within this tree.

refcol.py

import os
import re
import sys

class ReferenceCollector:
  def __init__(self, pattern, ignored_files):
                self.files = {}
                self.ignored_files = ignored_files
                self.PATTERN = pattern 
                
  def run_scan(self, dir):
    """Scan given directory for unreferenced files."""
    for root, dirs, files in os.walk(os.path.normpath(dir)):
            for filename in files:
                    self.register(root, filename)
                    self.scan_content_for_references(root, filename)
                
  def register(self, path, filename):
          """Initially register file if unknown."""
          if not filename in self.ignored_files and not filename in self.files:
                  self.files[filename] = { "references": 0, "path": path }
  
  def scan_content_for_references(self, path, filename):
          """Scan content of given file for references to other files."""
          content = self._read_file(path, filename)
          referenced_files = self._extract_file_references(content)
          
          for referenced_file in referenced_files:
                  if not referenced_file in self.files:
                          self.register(path, referenced_file)
                          
                  self.files[referenced_file]["references"] += 1

  def _extract_file_references(self, content):
          matches = re.findall(self.PATTERN, content)
          referenced_files = [groups[0] for groups in matches]
          return self._strip_path(referenced_files)
                  
  def _read_file(self, path, filename):
          with open(os.path.join(path, filename), "r") as file:
                  content = file.read()
          
          return content
                  
  def _strip_path(self, files):
          filenames = []
          for file in files:
                  filenames.append(os.path.basename(file))
                  
          return filenames
          
  def get_unreferenced_files(self):
          """Return unreferenced files with fully qualified path."""
          unreferenced = []
          for filename, data in self.files.items():
                  if data["references"] == 0:
                          unreferenced.append(os.path.join(data["path"], filename))
                          
          unreferenced.sort() 
          return unreferenced
          
  def get_number_of_scanned_files(self):
          return len(self.files)
                
if __name__ == "__main__":
  reference_collector = ReferenceCollector(pattern="([/\w-]+\.(xml|sql))", ignored_files=["ignore.xml"])

  reference_collector.run_scan(sys.argv[1])

  print("Number of scanned files: {}".format(reference_collector.get_number_of_scanned_files()))
  print("Number of unreferenced file(s): {}".format(len(reference_collector.get_unreferenced_files())))
  print("Unreferenced file(s):")

  for file in reference_collector.get_unreferenced_files():
        print(file)