#! /usr/bin/python

# (Be in -*- python -*- mode.)
#
# ====================================================================
# Copyright (c) 2006-2008 CollabNet.  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 software consists of voluntary contributions made by many
# individuals.  For exact contribution history, see the revision
# history and logs, available at http://cvs2svn.tigris.org/.
# ====================================================================

"""Usage: destroy_repository.py OPTION... PATH...

Strip the text content out of RCS-format files.

*** This script irretrievably destroys any RCS files that it is applied to!

This script attempts to strip the file text, log messages, and author
names out of RCS files, in addition to renaming RCS files and directories.
(This is useful to make test cases smaller and to remove much of the
proprietary information that is stored in a repository.)  Note that this
script does NOT obliterate other information that might also be considered
proprietary, such as 'CVSROOT' directories and their contents, commit dates,
etc.  In fact, it's not guaranteed even to obliterate all of the file text,
or to do anything else for that matter.

The following OPTIONs are recognized:
  --all       destroy all data (this is the default if no options are given)
  --data      destroy revision data (file contents) only
  --metadata  destroy revision metadata (author, log message, description) only
  --symbols   destroy symbol names (branch/tag names) only
  --filenames destroy the filenames of RCS files
  --basenames destroy basenames only (keep filename extensions, such as '.txt')
              (--filenames overrides --basenames)
  --dirnames  destroy directory names within given PATH. PATH itself (if a
              directory) is not destroyed.
  --cvsroot   delete files within 'CVSROOT' directories, instead of leaving
              them untouched. The 'CVSROOT' directory itself is preserved.
  --no-<X>    where <X> is one of the above options negates the meaning of that
              option.

Each PATH that is a *,v file will be stripped.

Each PATH that is a directory will be traversed and all of its *,v
files stripped.

Other PATHs will be ignored.


Examples of usage:
  destroy_repository.py PATH
        destroys all data in PATH

  destroy_repository.py --all PATH
        same as above

  destroy_repository.py --data PATH
        destroys only revision data

  destroy_repository.py --no-data PATH
        destroys everything but revision data

  destroy_repository.py --data --metadata PATH
        destroys revision data and metadata only

---->8----

The *,v files must be writable by the user running the script.
Typically CVS repositories are read-only, so you might have to run
something like

    $ chmod -R ug+w my/repo/path

before running this script.

Most cvs2svn behavior is completely independent of the text contained
in an RCS file.  (The text is not even looked at until OutputPass.)

The idea is to use this script when preparing test cases for problems
that you experience with cvs2svn.  Instead of sending us your whole
CVS repository, you should:

1. Make a copy of the original repository

2. Run this script on the copy (NEVER ON THE ORIGINAL!!!)

3. Verify that the problem still exists when you use cvs2svn to
   convert the 'destroyed' copy

4. Send us the 'destroyed' copy along with the exact cvs2svn version
   that you used, the exact command line that you used to start the
   conversion, and the options file if you used one.

Please also consider using shrink_test_case.py to localize the problem
even further.

"""

import sys
import os
import shutil
import re

sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))

from cvs2svn_lib.key_generator import KeyGenerator
from cvs2svn_lib.rcsparser import parse
from rcs_file_filter import WriteRCSFileSink
from rcs_file_filter import FilterSink


# Which components to be destroyed. Default to all.
destroy = {
    'data': True,
    'metadata': True,
    'symbols': True,
    'filenames': True,
    'basenames': True,
    'dirnames': True,
    'cvsroot': True,
    }

tmpdir = 'destroy_repository-tmp'

file_key_generator = KeyGenerator(1)

def get_tmp_filename():
    return os.path.join(tmpdir, 'f%07d.tmp' % file_key_generator.gen_id())

# Mapping from "real" symbol name to rewritten symbol name
symbol_map = {}

def rewrite_symbol(name):
    if name not in symbol_map:
        symbol_map[name] = "symbol%05d" % (len(symbol_map))
    return symbol_map[name]

# Mapping from "real" filename to rewritten filename
filename_map = {
    # Empty filename should always map to empty filename. This is useful when
    # preserving the last component of filenames with only one component.
    '': '',
    }

# Set the following to true if we should not destroy the last filename
# component (aka. filename extension)
keep_last_filename_component = False

def rewrite_filename(pathname):
    if not destroy['filenames']:
        return pathname
    (dirname, filename) = os.path.split(pathname)
    extra = ''

    # Strip trailing ',v' now, and re-append it to the rewritten filename
    if filename.endswith(',v'):
        extra += ',v'
        filename = filename[:-2]

    if keep_last_filename_component:
        (filename, extension) = os.path.splitext(filename)
        if not extension:
            # filename has no extension.  Do not rewrite this filename
            # at all.
            return pathname
        extra = extension + extra

    # Rewrite filename
    try:
        return os.path.join(dirname, filename_map[filename] + extra)
    except KeyError:
        # filename_map[filename] does not exist. Generate automatically:
        num = len(filename_map)
        while True:
            filename_map[filename] = "file%03d" % (num)
            retval = os.path.join(dirname, filename_map[filename] + extra)
            if not os.path.exists(retval):
                return retval
            num += 1

# List of directory names to be renamed. This list is filled while we walk
# the directory structure, and then processed afterwards, in order to not
# mess up the directory structure while it is being walked.
rename_dir_list = []

def rename_dirs():
    """Rename all directories occuring in rename_dir_list"""
    # Make sure we rename subdirs _before_ renaming their parents
    rename_dir_list.reverse()
    rename_map = {}
    num = 0
    for d in rename_dir_list:
        (parent, name) = os.path.split(d)
        # Skip rewriting 'Attic' directories
        if name == "Attic":
            continue
        if name not in rename_map:
            while True:
                num += 1
                rename_map[name] = "dir%03d" % (num)
                if not os.path.exists(os.path.join(parent, rename_map[name])):
                    break
        new_d = os.path.join(parent, rename_map[name])
        assert not os.path.exists(new_d)
        shutil.move(d, new_d)


class Substituter:
    def __init__(self, template):
        self.template = template
        self.key_generator = KeyGenerator(1)

        # A map from old values to new ones.
        self.substitutions = {}

    def get_substitution(self, s):
        r = self.substitutions.get(s)
        if r == None:
            r = self.template % self.key_generator.gen_id()
            self.substitutions[s] = r
        return r


class LogSubstituter(Substituter):
    # If a log messages matches any of these regular expressions, it
    # is passed through untouched.
    untouchable_log_res = [
        re.compile(r'^Initial revision\n$'),
        re.compile(r'^file (?P<filename>.+) was initially added'
                   r' on branch (?P<symbol>.+)\.\n$'),
        re.compile(r'^\*\*\* empty log message \*\*\*\n$'),
        re.compile(r'^initial checkin$'),
        ]

    def __init__(self):
        Substituter.__init__(self, 'log %d')

    def get_substitution(self, log):
        keep_log = ''
        for untouchable_log_re in self.untouchable_log_res:
            m = untouchable_log_re.search(log)
            if m:
                # We have matched one of the above regexps
                # Keep log message
                keep_log = log
                # Check if we matched a regexp with a named subgroup
                groups = m.groupdict()
                if 'symbol' in groups and destroy['symbols']:
                    # Need to rewrite symbol name
                    symbol = groups['symbol']
                    keep_log = keep_log.replace(symbol, rewrite_symbol(symbol))
                if 'filename' in groups and destroy['filenames']:
                    # Need to rewrite filename
                    filename = groups['filename']
                    keep_log = keep_log.replace(
                        filename, rewrite_filename(filename)
                        )
        if keep_log:
            return keep_log
        if destroy['metadata']:
            return Substituter.get_substitution(self, log)
        return log


class DestroyerFilterSink(FilterSink):
    def __init__(self, author_substituter, log_substituter, sink):
        FilterSink.__init__(self, sink)

        self.author_substituter = author_substituter
        self.log_substituter = log_substituter

    def set_head_revision(self, revision):
        self.head_revision = revision
        FilterSink.set_head_revision(self, revision)

    def define_tag(self, name, revision):
        if destroy['symbols']:
            name = rewrite_symbol(name)
        FilterSink.define_tag(self, name, revision)

    def define_revision(
        self, revision, timestamp, author, state, branches, next
        ):
        if destroy['metadata']:
            author = self.author_substituter.get_substitution(author)
        FilterSink.define_revision(
            self, revision, timestamp, author, state, branches, next
            )

    def set_description(self, description):
        if destroy['metadata']:
            description = ''
        FilterSink.set_description(self, description)

    def set_revision_info(self, revision, log, text):
        if destroy['data']:
            if revision == self.head_revision:
                # Set the HEAD text unconditionally.  (It could be
                # that revision HEAD-1 has an empty deltatext, in
                # which case the HEAD text was actually committed in
                # an earlier commit.)
                text = (
                    'This text was last seen in HEAD (revision %s)\n'
                    ) % (revision,)
            elif text == '':
                # This is a no-op revision; preserve that fact.  (It
                # might be relied on by cvs2svn).
                pass
            else:
                # Otherwise, replace the data.
                if revision.count('.') == 1:
                    # On trunk, it could be that revision N-1 has an
                    # empty deltatext, in which case text for revision
                    # N was actually committed in an earlier commit.
                    text = (
                        'd1 1\n'
                        'a1 1\n'
                        'This text was last seen in revision %s\n'
                        ) % (revision,)
                else:
                    # On a branch, we know that the text was changed
                    # in revision N (even though the same text might
                    # also be kept across later revisions N+1 etc.)
                    text = (
                        'd1 1\n'
                        'a1 1\n'
                        'This text was committed in revision %s\n'
                        ) % (revision,)
        if destroy['metadata'] or destroy['symbols'] or destroy['filenames']:
            log = self.log_substituter.get_substitution(log)
        FilterSink.set_revision_info(self, revision, log, text)


class FileDestroyer:
    def __init__(self):
        self.log_substituter = LogSubstituter()
        self.author_substituter = Substituter('author%d')

    def destroy_file(self, filename):
        tmp_filename = get_tmp_filename()
        f = open(tmp_filename, 'wb')
        new_filename = rewrite_filename(filename)
        parse(
            open(filename, 'rb'),
            DestroyerFilterSink(
                self.author_substituter,
                self.log_substituter,
                WriteRCSFileSink(f),
                )
            )
        f.close()

        # Replace the original file with the new one:
        assert filename == new_filename or not os.path.exists(new_filename)
        os.remove(filename)
        shutil.move(tmp_filename, new_filename)

    def visit(self, dirname, names):
        # Special handling of CVSROOT directories
        if "CVSROOT" in names:
            path = os.path.join(dirname, "CVSROOT")
            if destroy['cvsroot']:
                # Remove all contents within CVSROOT
                sys.stderr.write('Deleting %s contents...' % path)
                shutil.rmtree(path)
                os.mkdir(path)
            else:
                # Leave CVSROOT alone
                sys.stderr.write('Skipping %s...' % path)
            del names[names.index("CVSROOT")]
            sys.stderr.write('done.\n')
        for name in names:
            path = os.path.join(dirname, name)
            if os.path.isfile(path) and path.endswith(',v'):
                sys.stderr.write('Destroying %s...' % path)
                self.destroy_file(path)
                sys.stderr.write('done.\n')
            elif os.path.isdir(path):
                if destroy['dirnames']:
                    rename_dir_list.append(path)
                # Subdirectories are traversed automatically
                pass
            else:
                sys.stderr.write('File %s is being ignored.\n' % path)

    def destroy_dir(self, path):
        os.path.walk(path, FileDestroyer.visit, self)


def usage_abort(msg):
    if msg:
        print >>sys.stderr, "ERROR:", msg
        print >>sys.stderr
    # Use this file's docstring as a usage string, but only the first part
    print __doc__.split('\n---->8----', 1)[0]
    sys.exit(1)

if __name__ == '__main__':
    if not os.path.isdir(tmpdir):
        os.makedirs(tmpdir)

    # Paths to be destroyed
    paths = []

    # Command-line argument processing
    first_option = True
    for arg in sys.argv[1:]:
        if arg.startswith("--"):
            # Option processing
            option = arg[2:].lower()
            value = True
            if option.startswith("no-"):
                value = False
                option = option[3:]
            if first_option:
                # Use the first option on the command-line to determine the
                # default actions. If the first option is negated (i.e. --no-X)
                # the default action should be to destroy everything.
                # Otherwise, the default action should be to destroy nothing.
                # This makes both positive and negative options work
                # intuitively (e.g. "--data" will destroy only data, while
                # "--no-data" will destroy everything BUT data).
                for d in destroy.keys():
                    destroy[d] = not value
                first_option = False
            if option in destroy:
                destroy[option] = value
            elif option == "all":
                for d in destroy.keys():
                    destroy[d] = value
            else:
                usage_abort("Unknown OPTION '%s'" % arg)
        else:
            # Path argument
            paths.append(arg)

    # If --basenames if given (and not also --filenames), we shall destroy
    # filenames, up to, but not including the last component.
    if destroy['basenames'] and not destroy['filenames']:
        destroy['filenames'] = True
        keep_last_filename_component = True

    if not paths:
        usage_abort("No PATH given")

    # Destroy given PATHs
    file_destroyer = FileDestroyer()
    for path in paths:
        if os.path.isfile(path) and path.endswith(',v'):
            file_destroyer.destroy_file(path)
        elif os.path.isdir(path):
            file_destroyer.destroy_dir(path)
        else:
            sys.stderr.write('PATH %s is being ignored.\n' % path)

    if destroy['dirnames']:
        rename_dirs()