Initial set of improvementes merged into master, including a class hierarchy redesign and performance improvements

Merge commit 'origin/improvements'

* commit 'origin/improvements': (38 commits)
  test_performance: module containing benchmarks to get an idea of the achieved throughput
  Removed plenty of mocked tree tests as they cannot work anymore with persistent commands that require stdin AND binary data - not even an adapter would help here. These tests will have to be replaced.
  tree: now reads tress directly by parsing the binary data, allowing it to safe possibly hundreds of command calls
  Refs are now truly dynamic - this costs a little bit of (persistent command) work, but assures refs behave as expected
  persistent command signature changed to also return the hexsha from a possible input ref - the objects pointed to by refs are now baked on demand - perhaps it should change to always be re-retrieved using a property as it is relatively fast - this way refs can always be cached
  test_blob: removed many redundant tests that would fail now as the mock cannot handle the complexity of the command backend
  Implemented git command facility to keep persistent commands for fast object information retrieval
  test: Added time-consuming test which could also be a benchmark in fact - currently it cause hundreds of command invocations which is slow
  cmd: added option to return the process directly, allowing to read the output directly from the output stream
  added Iterable interface to Ref type
  renamed find_all to list_all, changed commit to use iterable interface in preparation for command changes
  Added base for all iteratable objects
  unified name of utils module, recently it was named util and utils in different packages
  tree: renamed content_from_string to _from_string to make it private. Removed tests that were testing that method
  tree: now behaves like a list with string indexing functionality - using a dict as cache is a problem as the tree is ordered, added blobs, trees and traverse method
  test_base: Improved basic object creation as well as set hash tests
  repo.active_branch now returns a Head object, not a string
  IndexObjects are now checking their slots to raise a proper error message in case someone tries to access an unset path or mode - this information cannot be retrieved afterwards as IndexObject information is kept in the object that pointed at them. To find this information, one would have to search all objects which is not feasible
  refs now take repo as first argument and derive from LazyMixin to allow them to dynamically retrieve their objects
  renamed from_string and list_from_string to _from_string and _list_from_string to indicate their new status as private method, adjusted all callers respectively
  ...

1 files changed, 312 insertions, 195 deletions

diff --git a/lib/git/cmd.py b/lib/git/cmd.py
index aef53350..2965eb8b 100644
--- a/lib/git/cmd.py
+++ b/lib/git/cmd.py
@@ -6,7 +6,6 @@
 
 import os, sys
 import subprocess
-import re
 from utils import *
 from errors import GitCommandError
 
@@ -14,208 +13,326 @@ from errors import GitCommandError
 GIT_PYTHON_TRACE = os.environ.get("GIT_PYTHON_TRACE", False)
 
 execute_kwargs = ('istream', 'with_keep_cwd', 'with_extended_output',
-                  'with_exceptions', 'with_raw_output')
+				  'with_exceptions', 'with_raw_output', 'as_process')
 
 extra = {}
 if sys.platform == 'win32':
-    extra = {'shell': True}
+	extra = {'shell': True}
 
 class Git(object):
-    """
-    The Git class manages communication with the Git binary.
-    
+	"""
+	The Git class manages communication with the Git binary.
+	
 	It provides a convenient interface to calling the Git binary, such as in::
 	
 	 g = Git( git_dir )
-	 g.init()			        # calls 'git init' program
+	 g.init()					# calls 'git init' program
 	 rval = g.ls_files()		# calls 'git ls-files' program
 	
 	``Debugging``
-	    Set the GIT_PYTHON_TRACE environment variable print each invocation 
-	    of the command to stdout.
-	    Set its value to 'full' to see details about the returned values.
-    """
-    def __init__(self, git_dir=None):
-        """
-        Initialize this instance with:
-        
-        ``git_dir``
-           Git directory we should work in. If None, we always work in the current 
-           directory as returned by os.getcwd()
-        """
-        super(Git, self).__init__()
-        self.git_dir = git_dir
-
-    def __getattr__(self, name):
-        """
-        A convenience method as it allows to call the command as if it was 
-        an object.
-        Returns
-            Callable object that will execute call _call_process with your arguments.
-        """
-        if name[:1] == '_':
-            raise AttributeError(name)
-        return lambda *args, **kwargs: self._call_process(name, *args, **kwargs)
-
-    @property
-    def get_dir(self):
-        """
-        Returns
-            Git directory we are working on
-        """
-        return self.git_dir
-
-    def execute(self, command,
-                istream=None,
-                with_keep_cwd=False,
-                with_extended_output=False,
-                with_exceptions=True,
-                with_raw_output=False,
-                ):
-        """
-        Handles executing the command on the shell and consumes and returns
-        the returned information (stdout)
-
-        ``command``
-            The command argument list to execute.
-            It should be a string, or a sequence of program arguments. The
-            program to execute is the first item in the args sequence or string.
-
-        ``istream``
-            Standard input filehandle passed to subprocess.Popen.
-
-        ``with_keep_cwd``
-            Whether to use the current working directory from os.getcwd().
-            GitPython uses get_work_tree() as its working directory by
-            default and get_git_dir() for bare repositories.
-
-        ``with_extended_output``
-            Whether to return a (status, stdout, stderr) tuple.
-
-        ``with_exceptions``
-            Whether to raise an exception when git returns a non-zero status.
-
-        ``with_raw_output``
-            Whether to avoid stripping off trailing whitespace.
-
-        Returns::
-        
-         str(output)                                  # extended_output = False (Default)
-         tuple(int(status), str(stdout), str(stderr)) # extended_output = True
-        
-        Raise
-        	GitCommandError
-        
-        NOTE
-           If you add additional keyword arguments to the signature of this method, 
-           you must update the execute_kwargs tuple housed in this module.
-        """
-        if GIT_PYTHON_TRACE and not GIT_PYTHON_TRACE == 'full':
-            print ' '.join(command)
-
-        # Allow the user to have the command executed in their working dir.
-        if with_keep_cwd or self.git_dir is None:
-          cwd = os.getcwd()
-        else:
-          cwd=self.git_dir
-
-        # Start the process
-        proc = subprocess.Popen(command,
-                                cwd=cwd,
-                                stdin=istream,
-                                stderr=subprocess.PIPE,
-                                stdout=subprocess.PIPE,
-                                **extra
-                                )
-
-        # Wait for the process to return
-        try:
-            stdout_value = proc.stdout.read()
-            stderr_value = proc.stderr.read()
-            status = proc.wait()
-        finally:
-            proc.stdout.close()
-            proc.stderr.close()
-
-        # Strip off trailing whitespace by default
-        if not with_raw_output:
-            stdout_value = stdout_value.rstrip()
-            stderr_value = stderr_value.rstrip()
-
-        if with_exceptions and status != 0:
-            raise GitCommandError(command, status, stderr_value)
-
-        if GIT_PYTHON_TRACE == 'full':
-            if stderr_value:
-              print "%s -> %d: '%s' !! '%s'" % (command, status, stdout_value, stderr_value)
-            elif stdout_value:
-              print "%s -> %d: '%s'" % (command, status, stdout_value)
-            else:
-              print "%s -> %d" % (command, status)
-
-        # Allow access to the command's status code
-        if with_extended_output:
-            return (status, stdout_value, stderr_value)
-        else:
-            return stdout_value
-
-    def transform_kwargs(self, **kwargs):
-        """
-        Transforms Python style kwargs into git command line options.
-        """
-        args = []
-        for k, v in kwargs.items():
-            if len(k) == 1:
-                if v is True:
-                    args.append("-%s" % k)
-                elif type(v) is not bool:
-                    args.append("-%s%s" % (k, v))
-            else:
-                if v is True:
-                    args.append("--%s" % dashify(k))
-                elif type(v) is not bool:
-                    args.append("--%s=%s" % (dashify(k), v))
-        return args
-
-    def _call_process(self, method, *args, **kwargs):
-        """
-        Run the given git command with the specified arguments and return
-        the result as a String
-
-        ``method``
-            is the command. Contained "_" characters will be converted to dashes,
-            such as in 'ls_files' to call 'ls-files'.
-
-        ``args``
-            is the list of arguments
-
-        ``kwargs``
-            is a dict of keyword arguments.
-            This function accepts the same optional keyword arguments
-            as execute().
-
-        Examples::
-            git.rev_list('master', max_count=10, header=True)
-
-        Returns
-            Same as execute()
-        """
-
-        # Handle optional arguments prior to calling transform_kwargs
-        # otherwise these'll end up in args, which is bad.
-        _kwargs = {}
-        for kwarg in execute_kwargs:
-            try:
-                _kwargs[kwarg] = kwargs.pop(kwarg)
-            except KeyError:
-                pass
-
-        # Prepare the argument list
-        opt_args = self.transform_kwargs(**kwargs)
-        ext_args = map(str, args)
-        args = opt_args + ext_args
-
-        call = ["git", dashify(method)]
-        call.extend(args)
-
-        return self.execute(call, **_kwargs)
+		Set the GIT_PYTHON_TRACE environment variable print each invocation 
+		of the command to stdout.
+		Set its value to 'full' to see details about the returned values.
+	"""
+	class AutoInterrupt(object):
+		"""
+		Kill/Interrupt the stored process instance once this instance goes out of scope. It is 
+		used to prevent processes piling up in case iterators stop reading.
+		Besides all attributes are wired through to the contained process object
+		"""
+		__slots__= "proc"
+		
+		def __init__(self, proc ):
+			self.proc = proc
+			
+		def __del__(self):
+			# did the process finish already so we have a return code ?
+			if self.proc.poll() is not None:
+				return 
+			
+			# try to kill it
+			try:
+				os.kill(self.proc.pid, 2)	# interrupt signal
+			except AttributeError:
+				# try windows 
+				subprocess.call(("TASKKILL", "/T", "/PID", self.proc.pid))
+			# END exception handling 
+			
+		def __getattr__(self, attr):
+			return getattr(self.proc, attr)
+	
+	
+	def __init__(self, git_dir=None):
+		"""
+		Initialize this instance with:
+		
+		``git_dir``
+		   Git directory we should work in. If None, we always work in the current 
+		   directory as returned by os.getcwd()
+		"""
+		super(Git, self).__init__()
+		self.git_dir = git_dir
+		
+		# cached command slots
+		self.cat_file_header = None
+		self.cat_file_all = None
+
+	def __getattr__(self, name):
+		"""
+		A convenience method as it allows to call the command as if it was 
+		an object.
+		Returns
+			Callable object that will execute call _call_process with your arguments.
+		"""
+		if name[:1] == '_':
+			raise AttributeError(name)
+		return lambda *args, **kwargs: self._call_process(name, *args, **kwargs)
+
+	@property
+	def get_dir(self):
+		"""
+		Returns
+			Git directory we are working on
+		"""
+		return self.git_dir
+
+	def execute(self, command,
+				istream=None,
+				with_keep_cwd=False,
+				with_extended_output=False,
+				with_exceptions=True,
+				with_raw_output=False,
+				as_process=False
+				):
+		"""
+		Handles executing the command on the shell and consumes and returns
+		the returned information (stdout)
+
+		``command``
+			The command argument list to execute.
+			It should be a string, or a sequence of program arguments. The
+			program to execute is the first item in the args sequence or string.
+
+		``istream``
+			Standard input filehandle passed to subprocess.Popen.
+
+		``with_keep_cwd``
+			Whether to use the current working directory from os.getcwd().
+			GitPython uses get_work_tree() as its working directory by
+			default and get_git_dir() for bare repositories.
+
+		``with_extended_output``
+			Whether to return a (status, stdout, stderr) tuple.
+
+		``with_exceptions``
+			Whether to raise an exception when git returns a non-zero status.
+
+		``with_raw_output``
+			Whether to avoid stripping off trailing whitespace.
+			
+		 ``as_process``
+		 	Whether to return the created process instance directly from which 
+		 	streams can be read on demand. This will render with_extended_output, 
+		 	with_exceptions and with_raw_output ineffective - the caller will have 
+		 	to deal with the details himself.
+		 	It is important to note that the process will be placed into an AutoInterrupt
+		 	wrapper that will interrupt the process once it goes out of scope. If you 
+		 	use the command in iterators, you should pass the whole process instance 
+		 	instead of a single stream.
+
+		Returns::
+		
+		 str(output)								  # extended_output = False (Default)
+		 tuple(int(status), str(stdout), str(stderr)) # extended_output = True
+		
+		Raise
+			GitCommandError
+		
+		NOTE
+		   If you add additional keyword arguments to the signature of this method, 
+		   you must update the execute_kwargs tuple housed in this module.
+		"""
+		if GIT_PYTHON_TRACE and not GIT_PYTHON_TRACE == 'full':
+			print ' '.join(command)
+
+		# Allow the user to have the command executed in their working dir.
+		if with_keep_cwd or self.git_dir is None:
+		  cwd = os.getcwd()
+		else:
+		  cwd=self.git_dir
+
+		# Start the process
+		proc = subprocess.Popen(command,
+								cwd=cwd,
+								stdin=istream,
+								stderr=subprocess.PIPE,
+								stdout=subprocess.PIPE,
+								**extra
+								)
+
+		if as_process:
+			return self.AutoInterrupt(proc)
+		
+		# Wait for the process to return
+		status = 0
+		try:
+			stdout_value = proc.stdout.read()
+			stderr_value = proc.stderr.read()
+			status = proc.wait()
+		finally:
+			proc.stdout.close()
+			proc.stderr.close()
+
+		# Strip off trailing whitespace by default
+		if not with_raw_output:
+			stdout_value = stdout_value.rstrip()
+			stderr_value = stderr_value.rstrip()
+
+		if with_exceptions and status != 0:
+			raise GitCommandError(command, status, stderr_value)
+
+		if GIT_PYTHON_TRACE == 'full':
+			if stderr_value:
+			  print "%s -> %d: '%s' !! '%s'" % (command, status, stdout_value, stderr_value)
+			elif stdout_value:
+			  print "%s -> %d: '%s'" % (command, status, stdout_value)
+			else:
+			  print "%s -> %d" % (command, status)
+
+		# Allow access to the command's status code
+		if with_extended_output:
+			return (status, stdout_value, stderr_value)
+		else:
+			return stdout_value
+
+	def transform_kwargs(self, **kwargs):
+		"""
+		Transforms Python style kwargs into git command line options.
+		"""
+		args = []
+		for k, v in kwargs.items():
+			if len(k) == 1:
+				if v is True:
+					args.append("-%s" % k)
+				elif type(v) is not bool:
+					args.append("-%s%s" % (k, v))
+			else:
+				if v is True:
+					args.append("--%s" % dashify(k))
+				elif type(v) is not bool:
+					args.append("--%s=%s" % (dashify(k), v))
+		return args
+
+	def _call_process(self, method, *args, **kwargs):
+		"""
+		Run the given git command with the specified arguments and return
+		the result as a String
+
+		``method``
+			is the command. Contained "_" characters will be converted to dashes,
+			such as in 'ls_files' to call 'ls-files'.
+
+		``args``
+			is the list of arguments
+
+		``kwargs``
+			is a dict of keyword arguments.
+			This function accepts the same optional keyword arguments
+			as execute().
+
+		Examples::
+			git.rev_list('master', max_count=10, header=True)
+
+		Returns
+			Same as execute()
+		"""
+
+		# Handle optional arguments prior to calling transform_kwargs
+		# otherwise these'll end up in args, which is bad.
+		_kwargs = {}
+		for kwarg in execute_kwargs:
+			try:
+				_kwargs[kwarg] = kwargs.pop(kwarg)
+			except KeyError:
+				pass
+
+		# Prepare the argument list
+		opt_args = self.transform_kwargs(**kwargs)
+		ext_args = map(str, args)
+		args = opt_args + ext_args
+
+		call = ["git", dashify(method)]
+		call.extend(args)
+
+		return self.execute(call, **_kwargs)
+		
+	def _parse_object_header(self, header_line):
+		"""
+		``header_line``
+			<hex_sha> type_string size_as_int
+			
+		Returns
+			(hex_sha, type_string, size_as_int)
+			
+		Raises
+			ValueError if the header contains indication for an error due to incorrect 
+			input sha
+		"""
+		tokens = header_line.split()
+		if len(tokens) != 3:
+			raise ValueError( "SHA named %s could not be resolved" % tokens[0] )
+			
+		return (tokens[0], tokens[1], int(tokens[2]))
+	
+	def __prepare_ref(self, ref):
+		# required for command to separate refs on stdin
+		refstr = str(ref)				# could be ref-object
+		if refstr.endswith("\n"):
+			return refstr
+		return refstr + "\n"
+	
+	def __get_persistent_cmd(self, attr_name, cmd_name, *args,**kwargs):
+		cur_val = getattr(self, attr_name)
+		if cur_val is not None:
+			return cur_val
+			
+		options = { "istream" : subprocess.PIPE, "as_process" : True }
+		options.update( kwargs )
+		
+		cmd = self._call_process( cmd_name, *args, **options )
+		setattr(self, attr_name, cmd )
+		return cmd
+	
+	def __get_object_header(self, cmd, ref):
+		cmd.stdin.write(self.__prepare_ref(ref))
+		cmd.stdin.flush()
+		return self._parse_object_header(cmd.stdout.readline())
+	
+	def get_object_header(self, ref):
+		"""
+		Use this method to quickly examine the type and size of the object behind 
+		the given ref. 
+		
+		NOTE
+			The method will only suffer from the costs of command invocation 
+			once and reuses the command in subsequent calls. 
+		
+		Return:
+			(hexsha, type_string, size_as_int)
+		"""
+		cmd = self.__get_persistent_cmd("cat_file_header", "cat_file", batch_check=True)
+		return self.__get_object_header(cmd, ref)
+		
+	def get_object_data(self, ref):
+		"""
+		As get_object_header, but returns object data as well
+		
+		Return:
+			(hexsha, type_string, size_as_int,data_string)
+		"""
+		cmd = self.__get_persistent_cmd("cat_file_all", "cat_file", batch=True)
+		hexsha, typename, size = self.__get_object_header(cmd, ref)
+		data = cmd.stdout.read(size)
+		cmd.stdout.read(1)		# finishing newlines
+		
+		return (hexsha, typename, size, data)