svn_tree.py

Boost provides free peer-reviewed portable C++ source libraries. We emphasize libraries that work

#!/usr/bin/env python
#
#  tree.py: tools for comparing directory trees
#
#  Subversion is a tool for revision control.
#  See http://subversion.tigris.org for more information.
#
# ====================================================================
# Copyright (c) 2001 Sam Tobin-Hochstadt.  All rights reserved.
#
# This software is licensed as described in the file COPYING, which
# you should have received as part of this distribution.  The terms
# are also available at http://subversion.tigris.org/license-1.html.
# If newer versions of this license are posted there, you may use a
# newer version instead, at your option.
#
######################################################################

# This file was modified by Vladimir Prus to store modification times in tree
# nodes.

import re
import string
import os.path
import os
import stat


#========================================================================

# ===>  Overview of our Datastructures  <===

# The general idea here is that many, many things can be represented by
# a tree structure:

#   - a working copy's structure and contents
#   - the output of 'svn status'
#   - the output of 'svn checkout/update'
#   - the output of 'svn commit'

# The idea is that a test function creates a "expected" tree of some
# kind, and is then able to compare it to an "actual" tree that comes
# from running the Subversion client.  This is what makes a test
# automated; if an actual and expected tree match exactly, then the test
# has passed.  (See compare_trees() below.)

# The SVNTreeNode class is the fundamental data type used to build tree
# structures.  The class contains a method for "dropping" a new node
# into an ever-growing tree structure. (See also create_from_path()).

# We have four parsers in this file for the four use cases listed above:
# each parser examines some kind of input and returns a tree of
# SVNTreeNode objects.  (See build_tree_from_checkout(),
# build_tree_from_commit(), build_tree_from_status(), and
# build_tree_from_wc()).  These trees are the "actual" trees that result
# from running the Subversion client.

# Also necessary, of course, is a convenient way for a test to create an
# "expected" tree.  The test *could* manually construct and link a bunch
# of SVNTreeNodes, certainly.  But instead, all the tests are using the
# build_generic_tree() routine instead.

# build_generic_tree() takes a specially-formatted list of lists as
# input, and returns a tree of SVNTreeNodes.  The list of lists has this
# structure:

#   [ ['/full/path/to/item', 'text contents', {prop-hash}, {att-hash}],
#     [...],
#     [...],
#     ...   ]

# You can see that each item in the list essentially defines an
# SVNTreeNode.  build_generic_tree() instantiates a SVNTreeNode for each
# item, and then drops it into a tree by parsing each item's full path.

# So a typical test routine spends most of its time preparing lists of
# this format and sending them to build_generic_tree(), rather than
# building the "expected" trees directly.

#   ### Note: in the future, we'd like to remove this extra layer of
#   ### abstraction.  We'd like the SVNTreeNode class to be more
#   ### directly programmer-friendly, providing a number of accessor
#   ### routines, so that tests can construct trees directly.

# The first three fields of each list-item are self-explanatory.  It's
# the fourth field, the "attribute" hash, that needs some explanation.
# The att-hash is used to place extra information about the node itself,
# depending on the parsing context:

#   - in the 'svn co/up' use-case, each line of output starts with two
#     characters from the set of (A, D, G, U, C, _).  This status code
#     is stored in a attribute named 'status'.

#   - in the 'svn ci/im' use-case, each line of output starts with one
#      of the words (Adding, Deleting, Sending).  This verb is stored in
#      an attribute named 'verb'.

#   - in the 'svn status' use-case (which is always run with the -v
#     (--verbose) flag), each line of output contains a working revision
#     number and a two-letter status code similar to the 'svn co/up'
#     case.  The repository revision is also printed.  All of this
#     information is stored in attributes named 'wc_rev', 'status', and
#     'repos_rev', respectively.

#   - in the working-copy use-case, the att-hash is ignored.


# Finally, one last explanation: the file 'actions.py' contain a number
# of helper routines named 'run_and_verify_FOO'.  These routines take
# one or more "expected" trees as input, then run some svn subcommand,
# then push the output through an appropriate parser to derive an
# "actual" tree.  Then it runs compare_trees() and returns the result.
# This is why most tests typically end with a call to
# run_and_verify_FOO().




# A node in a tree.
#
# If CHILDREN is None, then the node is a file.  Otherwise, CHILDREN
# is a list of the nodes making up that directory's children.
#
# NAME is simply the name of the file or directory.  CONTENTS is a
# string that contains the file's contents (if a file), PROPS are
# properties attached to files or dirs, and ATTS is a dictionary of
# other metadata attached to the node.

class SVNTreeNode:

  def __init__(self, name, children=None, contents=None, props={}, atts={}):
    self.name = name
    self.mtime = 0
    self.children = children
    self.contents = contents
    self.props = props
    self.atts = atts
    self.path = name

# TODO: Check to make sure contents and children are mutually exclusive

  def add_child(self, newchild):
    if self.children is None:  # if you're a file,
      self.children = []     # become an empty dir.
    child_already_exists = 0
    for a in self.children:
      if a.name == newchild.name:
        child_already_exists = 1
        break
    if child_already_exists == 0:
      self.children.append(newchild)
      newchild.path = os.path.join (self.path, newchild.name)      

    # If you already have the node,
    else:      
      if newchild.children is None:
        # this is the 'end' of the chain, so copy any content here.
        a.contents = newchild.contents
        a.props = newchild.props
        a.atts = newchild.atts
        a.path = os.path.join (self.path, newchild.name)
      else:
        # try to add dangling children to your matching node
        for i in newchild.children:
          a.add_child(i)


  def pprint(self):
    print " * Node name:  ", self.name
    print "    Path:      ", self.path
    print "    Contents:  ", self.contents
    print "    Properties:", self.props
    print "    Attributes:", self.atts
    ### FIXME: I'd like to be able to tell the difference between
    ### self.children is None (file) and self.children == [] (empty
    ### diretory), but it seems that most places that construct
    ### SVNTreeNode objects don't even try to do that.  --xbc
    if self.children is not None:
      print "    Children:  ", len(self.children)
    else:
      print "    Children: is a file."

# reserved name of the root of the tree

root_node_name = "__SVN_ROOT_NODE"

# Exception raised if you screw up in this module.

class SVNTreeError(Exception): pass

# Exception raised if two trees are unequal

class SVNTreeUnequal(Exception): pass

# Exception raised if one node is file and other is dir

class SVNTypeMismatch(Exception): pass

# Exception raised if get_child is passed a file.

class SVNTreeIsNotDirectory(Exception): pass


# Some attributes 'stack' on each other if the same node is added
# twice to a tree.  Place all such special cases in here.
def attribute_merge(orighash, newhash):
  "Merge the attributes in NEWHASH into ORIGHASH."

  if orighash.has_key('verb') and newhash.has_key('verb'):
    # Special case: if a commit reports a node as "deleted", then
    # "added", it's a replacment.
    if orighash['verb'] == "Deleting":
      if newhash['verb'] == "Adding":
        orighash['verb'] = "Replacing"

  # Add future stackable attributes here...

  return orighash


# helper func
def add_elements_as_path(top_node, element_list):
  """Add the elements in ELEMENT_LIST as if they were a single path
  below TOP_NODE."""

  # The idea of this function is to take a list like so:
  # ['A', 'B', 'C'] and a top node, say 'Z', and generate a tree
  # like this:
  #
  #             Z -> A -> B -> C
  #
  # where 1 -> 2 means 2 is a child of 1.
  #

  prev_node = top_node
  for i in element_list:
    new_node = SVNTreeNode(i, None)
    prev_node.add_child(new_node)
    prev_node = new_node


# Sorting function -- sort 2 nodes by their names.
def node_is_greater(a, b):
  "Sort the names of two nodes."
  # Interal use only
  if a.name == b.name:
    return 0
  if a.name > b.name:
    return 1
  else:
    return -1


# Helper for compare_trees
def compare_file_nodes(a, b):
  """Compare two nodes' names, contents, and properties, ignoring
  children.  Return 0 if the same, 1 otherwise."""
  if a.name != b.name:
    return 1
  if a.contents != b.contents:
    return 1
  if a.props != b.props:
    return 1
  if a.atts != b.atts:
    return 1


# Internal utility used by most build_tree_from_foo() routines.
#
# (Take the output and .add_child() it to a root node.)

def create_from_path(path, contents=None, props={}, atts={}):
  """Create and return a linked list of treenodes, given a PATH
  representing a single entry into that tree.  CONTENTS and PROPS are
  optional arguments that will be deposited in the tail node."""

  # get a list of all the names in the path
  # each of these will be a child of the former
  elements = path.split("/")
  if len(elements) == 0:
    raise SVNTreeError

  root_node = SVNTreeNode(elements[0], None)

  add_elements_as_path(root_node, elements[1:])

  # deposit contents in the very last node.
  node = root_node
  while 1:
    if node.children is None:
      node.contents = contents
      node.props = props
      node.atts = atts
      break
    node = node.children[0]

  return root_node


# helper for handle_dir(), which is a helper for build_tree_from_wc()
def get_props(path):
  "Return a hash of props for PATH, using the svn client."

  # It's not kosher to look inside SVN/ and try to read the internal
  # property storage format.  Instead, we use 'svn proplist'.  After
  # all, this is the only way the user can retrieve them, so we're
  # respecting the black-box paradigm.

  props = {}
  output, errput = main.run_svn(1, "proplist", path, "--verbose")

  for line in output:
    name, value = line.split(' : ')
    name = string.strip(name)
    value = string.strip(value)
    props[name] = value

  return props


# helper for handle_dir(), which helps build_tree_from_wc()
def get_text(path):
  "Return a string with the textual contents of a file at PATH."

  # sanity check
  if not os.path.isfile(path):
    return None

  fp = open(path, 'r')
  contents = fp.read()
  fp.close()
  return contents


# main recursive helper for build_tree_from_wc()