1 # The Detector determines which Monoglot or Polyglot should be
2 # used to parse a source file.
4 # The Detector primarily uses filename extensions to identify languages.
6 # The hash EXTENSION_MAP maps a filename extension to the name of a parser.
8 # If a filename extension is not enough to determine the correct parser (for
9 # instance, the *.m extension can indicate either a Matlab or Objective-C file),
10 # then the EXTENSION_MAP hash will contain a symbol identifying a Ruby method
11 # which will be invoked. This Ruby method can examine the file
12 # contents and return the name of the correct parser.
14 # Many source files do not have an extension. The method +disambiguate_nil+
15 # is called in these cases. The +file+ command line tool is used to determine
16 # the type of file and select a parser.
18 # The Detector is covered by DetectorTest.
20 class Ohcount::Detector
23 # A performance hack -- once we've checked for the presence of *.m files, the result
24 # is stored here to avoid checking twice.
25 attr_accessor :contains_m
28 # The primary entry point for the detector.
29 # Given a file context containing the file name, content, and an array of
30 # other filenames in the source tree, attempts to detect which
31 # language family (Monoglot or Polyglot) is in use for this file.
33 # Returns nil if the language is not recognized or if the file does not
38 # # List all C/C++ files in the 'src' directory
39 # Dir.entries("src").each do |file|
40 # context = Ohcount::SimpleFileContext.new(file)
41 # polyglot = Ohcount::Detector.detect(context)
42 # puts "#{file}" if polyglot == 'cncpp'
45 def self.detect(file_context)
46 # start with extension
47 polyglot = EXTENSION_MAP[File.extname(file_context.filename).downcase]
51 return polyglot if polyglot.is_a?(String)
53 # extension is ambiguous - requires custom disambiguation
54 self.send(polyglot, file_context)
56 return disambiguate_nil(file_context)
58 raise RuntimeError.new("Unknown file detection type")
62 # Based solely on the filename, makes a judgment whether a file is a binary format.
63 def self.binary_filename?(filename)
91 ignore.include?(File.extname(filename))
94 # If an extension maps to a string, that string must be the name of a glot.
95 # If an extension maps to a Ruby symbol, that symbol must be the name of a
96 # Ruby method which will return the name of a glot.
100 '.asm' => "assembler",
102 '.bas' => "visualbasic",
111 '.el' => "emacslisp",
114 '.cs' => :disambiguate_cs,
117 '.frx' => "visualbasic",
118 '.groovy'=> "groovy",
119 '.h' => :disambiguate_h_header,
127 '.inc' => :disambiguate_inc,
129 '.js' => "javascript",
134 '.m' => :matlab_or_objective_c,
135 '.mm' => "objective_c",
160 '.vb' => "visualbasic",
161 '.vba' => "visualbasic",
162 '.vbs' => "visualbasic",
170 # Returns a count of lines in the buffer matching the given regular expression.
171 def self.lines_matching(buffer, re)
172 buffer.inject(0) { |total, line| line =~ re ? total+1 : total }
175 # For *.m files, differentiates Matlab from Objective-C.
177 # This is done with a weighted heuristic that
178 # scans the *.m file contents for keywords,
179 # and also checks for the presence of matching *.h files.
180 def self.matlab_or_objective_c(file_context)
181 buffer = file_context.contents
183 # if there are .h files in same directory, this probably isn't matlab
185 h_headers = -0.5 if file_context.filenames.select { |a| a =~ /\.h$/ }.any?
187 # if the contents contain 'function (' on a single line - very likely to be matlab
188 # if the contents contain lines starting with '%', its probably matlab comments
189 matlab_signatures = /(^\s*function\s*)|(^\s*%)/
190 matlab_sig_score = 0.1 * lines_matching(buffer, matlab_signatures)
192 # if the contents contains '//' or '/*', likely objective_c
193 objective_c_signatures = /(^\s*\/\/\s*)|(^\s*\/\*)|(^[+-])/
194 obj_c_sig_score = -0.1 * lines_matching(buffer, objective_c_signatures)
196 matlab = h_headers + matlab_sig_score + obj_c_sig_score
198 matlab > 0 ? 'matlab' : 'objective_c'
201 # For *.h files, differentiates C/C++ from Objective-C.
203 # This is done with a weighted heuristic that
204 # scans the *.h file contents for Objective-C keywords,
205 # and also checks for the presence of matching *.m files.
206 def self.disambiguate_h_header(file_context)
207 buffer = file_context.contents
211 # could it be realistically be objective_c ? are there any .m files at all?
212 # Speed hack - remember our findings in case we get the same filenames over and over
213 unless defined?(file_context.filenames.contains_m)
214 file_context.filenames.extend(ContainsM)
215 file_context.filenames.contains_m = file_context.filenames.select { |a| a =~ /\.m$/ }.any?
217 return 'cncpp' unless file_context.filenames.contains_m
219 # if the dir contains a matching *.m file, likely objective_c
220 if file_context.filename =~ /\.h$/
221 m_counterpart = file_context.filename.gsub(/\.h$/, ".m")
222 return 'objective_c' if file_context.filenames.include?(m_counterpart)
225 # ok - it just might be objective_c, let's check contents for objective_c signatures
226 objective_c_signatures = /(^\s*@interface)|(^\s*@end)/
227 objective_c += lines_matching(buffer, objective_c_signatures)
229 return objective_c > 1 ? 'objective_c' : 'cncpp'
232 # Tests whether the provided buffer contains binary or text content.
233 # This is not fool-proof -- we basically just check for zero values
234 # in the early bytes of the buffer. If we find a zero, we know it
235 # is not (ascii) text.
236 def self.binary_buffer?(buffer)
238 return true if buffer[i] == 0
243 # True if the provided buffer includes a '?php' directive
244 def self.php_instruction?(buffer)
248 # For *.inc files, checks for a PHP class.
249 def self.disambiguate_inc(file_context)
250 buffer = file_context.contents
251 return nil if binary_buffer?(buffer)
252 return 'php' if php_instruction?(buffer)
256 # For files with extention *.cs, differentiates C# from Clearsilver.
257 def self.disambiguate_cs(file_context)
258 buffer = file_context.contents
259 return 'clearsilver_template' if lines_matching(file_context.contents, /\<\?cs/) > 0
263 # Attempts to determine the Polyglot for files that do not have a
264 # filename extension.
266 # Relies on the bash +file+ command line tool as its primary method.
268 # There must be a file at <tt>file_context.file_location</tt> for +file+
271 def self.disambiguate_nil(file_context)
272 file_location = file_context.file_location
273 output = `file -b #{ file_location }`
275 when /([\w\/]+) script text/, /script text executable for ([\w\/]+)/
277 if script =~ /\/(\w*)$/
280 known_languages = EXTENSION_MAP.values
281 return script.downcase if known_languages.include?(script.downcase)
282 when /([\w\-]*) shell script text/