path: root/RepSys/log.py

#!/usr/bin/python
from RepSys import Error, config
from RepSys.svn import SVN
from RepSys.util import execcmd

from Cheetah.Template import Template

import sys
import os
import re
import time
import locale
import glob
import tempfile
import shutil


locale.setlocale(locale.LC_ALL, "C")

default_template = """
#for $rel in $releases_by_author
* $rel.date $rel.author_name <$rel.author_email> $rel.version-$rel.release
 ##
 #if not $rel.released
  (not released yet)
 #end if
 #for $rev in $rel.release_revisions
  #for $line in $rev.lines
  $line
  #end for
 #end for

 #for $author in $rel.authors
  + $author.name <$author.email>
  #for $rev in $author.revisions
    #for $line in $rev.lines
    $line
    #end for
  #end for

 #end for
#end for
"""

def getrelease(pkgdirurl, rev=None):
    """Tries to obtain the version-release of the package for a 
    yet-not-markrelease revision of the package.

    Is here where things should be changed if "automatic release increasing" 
    will be used.
    """
    svn = SVN(baseurl=pkgdirurl)
    tmpdir = tempfile.mktemp()
    try:
        pkgname = os.path.basename(pkgdirurl)
        pkgcurrenturl = os.path.join(pkgdirurl, "current")
        specurl = os.path.join(pkgcurrenturl, "SPECS")
        if svn.ls(specurl, noerror=1):
            svn.export(specurl, tmpdir, rev=rev)
            found = glob.glob(os.path.join(tmpdir, "*.spec"))
            if found:
                specpath = found[0]
                command = (("rpm -q --qf '%%{VERSION}-%%{RELEASE}\n' "
                           "--specfile %s") % specpath)
                status, output = execcmd(command)
                if status != 0:
                    raise Error, "Error in command %s: %s" % (command, output)
                releases = output.split()
                try:
                    version, release = releases[0].split("-", 1)
                except ValueError:
                    raise Error, "Invalid command output: %s: %s" % \
                            (command, output)
                return version, release
    finally:
        if os.path.isdir(tmpdir):
            shutil.rmtree(tmpdir)
            

class _Revision:
    lines = []
    date = None
    raw_date = None
    revision = None
    author_name = None
    author_email = None

    def __init__(self, **kwargs):
        self.__dict__.update(kwargs)


class _Release(_Revision):
    version = None
    release = None
    revisions = None

    def __init__(self, **kwargs):
        self.revisions = []
        _Revision.__init__(self, **kwargs)


def format_lines(lines):
    first = 1
    entrylines = []
    perexpr = re.compile(r"([^%])%([^%])")
    for line in lines:
        if line:
            line = perexpr.sub("\\1%%\\2", line)
            if first:
                first = 0
                line = line.lstrip()
                if line[0] != "-":
                    nextline = "- " + line
                else:
                    nextline = line
            elif line[0] != " " and line[0] != "-":
                nextline = "  " + line
            else:
                nextline = line
            if nextline not in entrylines:
                entrylines.append(nextline)
    return entrylines


class _Author:
    name = None
    email = None
    revisions = None


def group_releases_by_author(releases):
    allauthors = []
    for release in releases:
        authors = {}
        for revision in release.revisions:
            authors.setdefault(revision.author, []).append(revision)

        # all the mess below is to sort by author and by revision number
        decorated = []
        for authorname, revs in authors.iteritems():
            author = _Author()
            author.name = revs[0].author_name
            author.email = revs[0].author_email
            revdeco = [(r.revision, r) for r in revs]
            revdeco.sort(reverse=1)
            author.revisions = [t[1] for t in revdeco]
            decorated.append((max(revdeco)[0], author))

        decorated.sort(reverse=1)
        release.authors = [t[1] for t in decorated]
        # the difference between a released and a not released _Release is
        # the way the release numbers is obtained. So, when this is a
        # released, we already have it, but if we don't, we should get de
        # version/release string using getrelease and then get the first
        first, release.authors = release.authors[0], release.authors[1:]
        release.author_name = first.name
        release.author_email = first.email
        release.date = first.revisions[0].date
        release.raw_date = first.revisions[0].raw_date
        release.release_revisions = first.revisions
        release.revision = first.revisions[0].revision

    return releases


def group_revisions_by_author(currentlog):
    revisions = []
    last_author = None
    for entry in currentlog:
        revision = _Revision()
        revision.lines = format_lines(entry.lines)
        revision.raw_date = entry.date
        revision.date = parse_raw_date(entry.date)
        revision.revision = entry.revision
        if entry.author == last_author:
            revisions[-1].revisions.append(revision)
        else:
            author = _Author()
            author.name, author.email = get_author_name(entry.author)
            author.revisions = [revision]
            revisions.append(author)
        last_author = entry.author
    return revisions


emailpat = re.compile("(?P<name>.*?)\s*<(?P<email>.*?)>")

def get_author_name(author):
    found = emailpat.match(config.get("users", author, author))
    name = ((found and found.group("name")) or author)
    email = ((found and found.group("email")) or author)
    return name, email

def parse_raw_date(rawdate):
    return time.strftime("%a %b %d %Y", rawdate)

def make_release(author=None, revision=None, date=None, lines=None,
        entries=[], released=True, version=None, release=None):
    rel = _Release()
    rel.author = author
    found = emailpat.match(config.get("users", author, author or ""))
    rel.author_name = (found and found.group("name")) or author
    rel.author_email = (found and found.group("email")) or author
    rel.revision = revision
    rel.version = version
    rel.release = release
    rel.date = (date and parse_raw_date(date)) or None
    rel.lines = lines
    rel.released = released
    for entry in entries:
        revision = _Revision()
        revision.revision = entry.revision
        revision.lines = format_lines(entry.lines)
        revision.date = parse_raw_date(entry.date)
        revision.raw_date = entry.date
        revision.author = entry.author
        (revision.author_name, revision.author_email) = \
                get_author_name(entry.author)
        rel.revisions.append(revision)
    return rel


def dump_file(releases, currentlog=None, template=None):
    templpath = template or config.get("template", "path", None)
    params = {}
    if templpath is None or not os.path.exists(templpath):
        params["source"] = default_template
        sys.stderr.write("warning: %s not found. using built-in template.\n"%
                templpath)
    else:
        params["file"] = templpath
    releases_author = group_releases_by_author(releases)
    revisions_author = group_revisions_by_author(currentlog)
    params["searchList"] = [{"releases_by_author" : releases_author,
                             "releases" : releases,
                             "revisions_by_author": revisions_author}]
    t = Template(**params)
    return repr(t)


class InvalidEntryError(Exception):
    pass

def parse_repsys_entry(revlog):
    # parse entries in the format:
    # %repsys <operation>
    # key: value
    # ..
    # <newline>
    # <comments>
    #
    if len(revlog.lines) == 0 or not revlog.lines[0].startswith("%repsys"):
        raise InvalidEntryError
    try:       
        data = {"operation" : revlog.lines[0].split()[1]}
    except IndexError:
        raise InvalidEntryError
    for line in revlog.lines[1:]:
        if not line:
            break
        try:
            key, value = line.split(":", 1)
        except ValueError:
            raise InvalidEntryError
        data[key.strip().lower()] = value.strip() # ???
    return data
        

def get_revision_offset():
    try:
        revoffset = config.getint("log", "revision-offset", 0)
    except (ValueError, TypeError):
        raise Error, ("Invalid revision-offset number in configuration "
                      "file(s).")
    return revoffset or 0

oldmsgpat = re.compile(
        r"Copying release (?P<rel>[^\s]+) to (?P<dir>[^\s]+) directory\.")

def parse_markrelease_log(relentry):
    if not ((relentry.lines and oldmsgpat.match(relentry.lines[0]) \
            or parse_repsys_entry(relentry))):
        raise InvalidEntryError
    from_rev = None
    path = None
    for changed in relentry.changed:
        if changed["action"] == "A" and changed["from_rev"]:
            from_rev = changed["from_rev"]
            path = changed["path"]
            break
    else:
        raise InvalidEntryError
    # get the version and release from the names in the path, do not relay
    # on log messages
    version, release = path.rsplit(os.path.sep, 3)[-2:]
    return version, release, from_rev


def svn2rpm(pkgdirurl, rev=None, size=None, submit=False, template=None):
    concat = config.get("log", "concat", "").split()
    revoffset = get_revision_offset()
    svn = SVN(baseurl=pkgdirurl)
    pkgreleasesurl = os.path.join(pkgdirurl, "releases")
    pkgcurrenturl = os.path.join(pkgdirurl, "current")
    releaseslog = svn.log(pkgreleasesurl, noerror=1)
    currentlog = svn.log(pkgcurrenturl, limit=size, start=rev,
            end=revoffset)

    # sort releases by copyfrom-revision, so that markreleases for same
    # revisions won't be look empty
    releasesdata = []
    for relentry in releaseslog[::-1]:
        try:
            (version, release, relrevision) = \
                    parse_markrelease_log(relentry)
        except InvalidEntryError:
            continue
        releasesdata.append((relrevision, -relentry.revision, relentry, version, release))
    releasesdata.sort()

    # collect valid releases using the versions provided by the changes and
    # the packages
    prevrevision = 0
    releases = []
    for (relrevision, dummy, relentry, version, release) in releasesdata:
        if prevrevision == relrevision: 
            # ignore older markrelease of the same revision, since they
            # will have history
            continue
        entries = [entry for entry in currentlog
                    if relrevision >= entry.revision and
                      (prevrevision < entry.revision)]
        if not entries:
            #XXX probably a forced release, without commits in current/,
            # check if this is the right behavior
            sys.stderr.write("warning: skipping (possible) release "
                    "%s-%s@%s, no commits since previous markrelease (r%r)\n" %
                    (version, release, relrevision, prevrevision))
            continue

        release = make_release(author=relentry.author,
                revision=relentry.revision, date=relentry.date,
                lines=relentry.lines, entries=entries,
                version=version, release=release)
        releases.append(release)
        prevrevision = relrevision
            
    # look for commits that have been not submited (released) yet
    # this is done by getting all log entries newer (revision larger)
    # than releaseslog[0] (in the case it exists)
    if releaseslog:
        latest_revision = releaseslog[0].revision
    else:
        latest_revision = 0
    notsubmitted = [entry for entry in currentlog 
                    if entry.revision > latest_revision]
    if notsubmitted:
        # if they are not submitted yet, what we have to do is to add
        # a release/version number from getrelease()
        version, release = getrelease(pkgdirurl)
        toprelease = make_release(entries=notsubmitted, released=False,
                        version=version, release=release)
        releases.append(toprelease)

    data = dump_file(releases[::-1], currentlog=currentlog, template=template)
    return data



def specfile_svn2rpm(pkgdirurl, specfile, rev=None, size=None,
        submit=False, template=None):
    newlines = []
    found = 0
    
    # Strip old changelogs
    for line in open(specfile):
        if line.startswith("%changelog"):
            found = 1
        elif not found:
            newlines.append(line)
        elif line.startswith("%"):
            found = 0
            newlines.append(line)

    # Create new changelog
    newlines.append("\n\n%changelog\n")
    newlines.append(svn2rpm(pkgdirurl, rev=rev, size=size, submit=submit,
        template=template))

    # Merge old changelog, if available
    oldurl = config.get("log", "oldurl")
    if oldurl:
        svn = SVN(baseurl=pkgdirurl)
        tmpdir = tempfile.mktemp()
        try:
            pkgname = os.path.basename(pkgdirurl)
            pkgoldurl = os.path.join(oldurl, pkgname)
            if svn.ls(pkgoldurl, noerror=1):
                svn.export(pkgoldurl, tmpdir, rev=rev)
                logfile = os.path.join(tmpdir, "log")
                if os.path.isfile(logfile):
                    file = open(logfile)
                    newlines.append("\n")
                    newlines.append(file.read())
                    file.close()
        finally:
            if os.path.isdir(tmpdir):
                shutil.rmtree(tmpdir)

    # Write new specfile
    file = open(specfile, "w")
    file.write("".join(newlines))
    file.close()


if __name__ == "__main__":
    l = svn2rpm(sys.argv[1])
    print l

# vim:et:ts=4:sw=4