Add class SourceContentsScanner.

[dak.git] / daklib / contents.py

diff --git a/daklib/contents.py b/daklib/contents.py
index 6b3b81534d3862a296e58a00201abc431188654b..dffb425e3cc2e5f8c5245de9e209fdaaa7a1f454 100755 (executable)
--- a/daklib/contents.py
+++ b/daklib/contents.py
@@ -27,10 +27,11 @@ Helper code for contents generation.
 
 from daklib.dbconn import *
 from daklib.config import Config
 
 from daklib.dbconn import *
 from daklib.config import Config

-from daklib.threadpool import ThreadPool

 
 

-from sqlalchemy import desc, or_
-from subprocess import Popen, PIPE
+from multiprocessing import Pool
+from shutil import rmtree
+from subprocess import Popen, PIPE, check_call
+from tempfile import mkdtemp

 
 import os.path
 
 
 import os.path
 


@@ -44,25 +45,26 @@ class ContentsWriter(object):
         sure that the new ContentsWriter object can be executed in a different
         thread.
         '''
         sure that the new ContentsWriter object can be executed in a different
         thread.
         '''

-        self.suite = suite.clone()
-        self.session = self.suite.session()
-        self.architecture = architecture.clone(self.session)
-        self.overridetype = overridetype.clone(self.session)
-        if component is not None:
-            self.component = component.clone(self.session)
-        else:
-            self.component = None
+        self.suite = suite
+        self.architecture = architecture
+        self.overridetype = overridetype
+        self.component = component
+        self.session = suite.session()

 
     def query(self):
         '''
         Returns a query object that is doing most of the work.
         '''
 
     def query(self):
         '''
         Returns a query object that is doing most of the work.
         '''

+        overridesuite = self.suite
+        if self.suite.overridesuite is not None:
+            overridesuite = get_suite(self.suite.overridesuite, self.session)

         params = {
         params = {

-            'suite':    self.suite.suite_id,
-            'arch_all': get_architecture('all', self.session).arch_id,
-            'arch':     self.architecture.arch_id,
-            'type_id':  self.overridetype.overridetype_id,
-            'type':     self.overridetype.overridetype,
+            'suite':         self.suite.suite_id,
+            'overridesuite': overridesuite.suite_id,
+            'arch_all':      get_architecture('all', self.session).arch_id,
+            'arch':          self.architecture.arch_id,
+            'type_id':       self.overridetype.overridetype_id,
+            'type':          self.overridetype.overridetype,

         }
 
         if self.component is not None:
         }
 
         if self.component is not None:


@@ -86,13 +88,13 @@ with
 unique_override as
     (select o.package, s.section
         from override o, section s
 unique_override as
     (select o.package, s.section
         from override o, section s

-        where o.suite = :suite and o.type = :type_id and o.section = s.id and
+        where o.suite = :overridesuite and o.type = :type_id and o.section = s.id and

         o.component = :component)
 
         o.component = :component)
 

-select bc.file, substring(o.section from position('/' in o.section) + 1) || '/' || b.package as package
+select bc.file, string_agg(o.section || '/' || b.package, ',' order by b.package) as pkglist

     from newest_binaries b, bin_contents bc, unique_override o
     where b.id = bc.binary_id and o.package = b.package
     from newest_binaries b, bin_contents bc, unique_override o
     where b.id = bc.binary_id and o.package = b.package

-    order by bc.file, b.package'''
+    group by bc.file'''

 
         else:
             sql = '''
 
         else:
             sql = '''


@@ -114,39 +116,29 @@ with
 unique_override as
     (select distinct on (o.package, s.section) o.package, s.section
         from override o, section s
 unique_override as
     (select distinct on (o.package, s.section) o.package, s.section
         from override o, section s

-        where o.suite = :suite and o.type = :type_id and o.section = s.id
+        where o.suite = :overridesuite and o.type = :type_id and o.section = s.id

         order by o.package, s.section, o.modified desc)
 
         order by o.package, s.section, o.modified desc)
 

-select bc.file, substring(o.section from position('/' in o.section) + 1) || '/' || b.package as package
+select bc.file, string_agg(o.section || '/' || b.package, ',' order by b.package) as pkglist

     from newest_binaries b, bin_contents bc, unique_override o
     where b.id = bc.binary_id and o.package = b.package
     from newest_binaries b, bin_contents bc, unique_override o
     where b.id = bc.binary_id and o.package = b.package

-    order by bc.file, b.package'''
+    group by bc.file'''

 
 

-        return self.session.query("file", "package").from_statement(sql). \
+        return self.session.query("file", "pkglist").from_statement(sql). \

             params(params)
 
     def formatline(self, filename, package_list):
         '''
         Returns a formatted string for the filename argument.
         '''
             params(params)
 
     def formatline(self, filename, package_list):
         '''
         Returns a formatted string for the filename argument.
         '''

-        package_list = ','.join(package_list)

         return "%-55s %s\n" % (filename, package_list)
 
     def fetch(self):
         '''
         Yields a new line of the Contents-$arch.gz file in filename order.
         '''
         return "%-55s %s\n" % (filename, package_list)
 
     def fetch(self):
         '''
         Yields a new line of the Contents-$arch.gz file in filename order.
         '''

-        last_filename = None
-        package_list = []
-        for filename, package in self.query().yield_per(100):
-            if filename != last_filename:
-                if last_filename is not None:
-                    yield self.formatline(last_filename, package_list)
-                last_filename = filename
-                package_list = []
-            package_list.append(package)
-        if last_filename is not None:
-            yield self.formatline(last_filename, package_list)
+        for filename, package_list in self.query().yield_per(100):
+            yield self.formatline(filename, package_list)

         # end transaction to return connection to pool
         self.session.rollback()
 
         # end transaction to return connection to pool
         self.session.rollback()
 


@@ -183,63 +175,100 @@ select bc.file, substring(o.section from position('/' in o.section) + 1) || '/'
             if header_file:
                 header_file.close()
 
             if header_file:
                 header_file.close()
 

-    def write_file(self, dummy_arg = None):
+    def write_file(self):

         '''
         '''

-        Write the output file. The argument dummy_arg is ignored but needed by
-        our threadpool implementation.
+        Write the output file.

         '''
         command = ['gzip', '--rsyncable']
         '''
         command = ['gzip', '--rsyncable']

-        output_file = open(self.output_filename(), 'w')
-        pipe = Popen(command, stdin = PIPE, stdout = output_file).stdin
-        pipe.write(self.get_header())
+        final_filename = self.output_filename()
+        temp_filename = final_filename + '.new'
+        output_file = open(temp_filename, 'w')
+        gzip = Popen(command, stdin = PIPE, stdout = output_file)
+        gzip.stdin.write(self.get_header())

         for item in self.fetch():
         for item in self.fetch():

-            pipe.write(item)
-        pipe.close()
+            gzip.stdin.write(item)
+        gzip.stdin.close()

         output_file.close()
         output_file.close()

+        gzip.wait()
+        try:
+            os.remove(final_filename)
+        except:
+            pass
+        os.rename(temp_filename, final_filename)
+        os.chmod(final_filename, 0664)
+
+    @classmethod
+    def log_result(class_, result):
+        '''
+        Writes a result message to the logfile.
+        '''
+        class_.logger.log(result)

 
     @classmethod
 
     @classmethod

-    def write_all(class_, suite_names = [], force = False):
+    def write_all(class_, logger, suite_names = [], force = False):

         '''
         Writes all Contents files for suites in list suite_names which defaults
         to all 'touchable' suites if not specified explicitely. Untouchable
         suites will be included if the force argument is set to True.
         '''
         '''
         Writes all Contents files for suites in list suite_names which defaults
         to all 'touchable' suites if not specified explicitely. Untouchable
         suites will be included if the force argument is set to True.
         '''

+        class_.logger = logger

         session = DBConn().session()
         suite_query = session.query(Suite)
         if len(suite_names) > 0:
             suite_query = suite_query.filter(Suite.suite_name.in_(suite_names))
         if not force:
             suite_query = suite_query.filter_by(untouchable = False)
         session = DBConn().session()
         suite_query = session.query(Suite)
         if len(suite_names) > 0:
             suite_query = suite_query.filter(Suite.suite_name.in_(suite_names))
         if not force:
             suite_query = suite_query.filter_by(untouchable = False)

-        main = get_component('main', session)
-        non_free = get_component('non-free', session)
-        deb = get_override_type('deb', session)
-        udeb = get_override_type('udeb', session)
-        threadpool = ThreadPool()
+        deb_id = get_override_type('deb', session).overridetype_id
+        udeb_id = get_override_type('udeb', session).overridetype_id
+        main_id = get_component('main', session).component_id
+        non_free_id = get_component('non-free', session).component_id
+        pool = Pool()

         for suite in suite_query:
         for suite in suite_query:

+            suite_id = suite.suite_id

             for architecture in suite.get_architectures(skipsrc = True, skipall = True):
             for architecture in suite.get_architectures(skipsrc = True, skipall = True):

+                arch_id = architecture.arch_id

                 # handle 'deb' packages
                 # handle 'deb' packages

-                writer = ContentsWriter(suite, architecture, deb)
-                threadpool.queueTask(writer.write_file)
+                pool.apply_async(generate_helper, (suite_id, arch_id, deb_id), \
+                    callback = class_.log_result)

                 # handle 'udeb' packages for 'main' and 'non-free'
                 # handle 'udeb' packages for 'main' and 'non-free'

-                writer = ContentsWriter(suite, architecture, udeb, component = main)
-                threadpool.queueTask(writer.write_file)
-                writer = ContentsWriter(suite, architecture, udeb, component = non_free)
-                threadpool.queueTask(writer.write_file)
-        threadpool.joinAll()
+                pool.apply_async(generate_helper, (suite_id, arch_id, udeb_id, main_id), \
+                    callback = class_.log_result)
+                pool.apply_async(generate_helper, (suite_id, arch_id, udeb_id, non_free_id), \
+                    callback = class_.log_result)
+        pool.close()
+        pool.join()

         session.close()
 
         session.close()
 

-
-class ContentsScanner(object):
+def generate_helper(suite_id, arch_id, overridetype_id, component_id = None):
+    '''
+    This function is called in a new subprocess.

     '''
     '''

-    ContentsScanner provides a threadsafe method scan() to scan the contents of
-    a DBBinary object.
+    session = DBConn().session()
+    suite = Suite.get(suite_id, session)
+    architecture = Architecture.get(arch_id, session)
+    overridetype = OverrideType.get(overridetype_id, session)
+    log_message = [suite.suite_name, architecture.arch_string, overridetype.overridetype]
+    if component_id is None:
+        component = None
+    else:
+        component = Component.get(component_id, session)
+        log_message.append(component.component_name)
+    contents_writer = ContentsWriter(suite, architecture, overridetype, component)
+    contents_writer.write_file()
+    return log_message
+
+
+class BinaryContentsScanner(object):

     '''
     '''

-    def __init__(self, binary):
+    BinaryContentsScanner provides a threadsafe method scan() to scan the
+    contents of a DBBinary object.
+    '''
+    def __init__(self, binary_id):

         '''
         '''

-        The argument binary is the actual DBBinary object that should be
-        scanned.
+        The argument binary_id is the id of the DBBinary object that
+        should be scanned.

         '''
         '''

-        self.binary_id = binary.binary_id
+        self.binary_id = binary_id

 
     def scan(self, dummy_arg = None):
         '''
 
     def scan(self, dummy_arg = None):
         '''


@@ -249,7 +278,10 @@ class ContentsScanner(object):
         '''
         session = DBConn().session()
         binary = session.query(DBBinary).get(self.binary_id)
         '''
         session = DBConn().session()
         binary = session.query(DBBinary).get(self.binary_id)

-        for filename in binary.scan_contents():
+        fileset = set(binary.scan_contents())
+        if len(fileset) == 0:
+            fileset.add('EMPTY_PACKAGE')
+        for filename in fileset:

             binary.contents.append(BinContents(file = filename))
         session.commit()
         session.close()
             binary.contents.append(BinContents(file = filename))
         session.commit()
         session.close()


@@ -268,10 +300,136 @@ class ContentsScanner(object):
         if limit is not None:
             query = query.limit(limit)
         processed = query.count()
         if limit is not None:
             query = query.limit(limit)
         processed = query.count()

-        threadpool = ThreadPool()
+        pool = Pool()

         for binary in query.yield_per(100):
         for binary in query.yield_per(100):

-            threadpool.queueTask(ContentsScanner(binary).scan)
-        threadpool.joinAll()
+            pool.apply_async(binary_scan_helper, (binary.binary_id, ))
+        pool.close()
+        pool.join()
+        remaining = remaining()
+        session.close()
+        return { 'processed': processed, 'remaining': remaining }
+
+def binary_scan_helper(binary_id):
+    '''
+    This function runs in a subprocess.
+    '''
+    scanner = BinaryContentsScanner(binary_id)
+    scanner.scan()
+
+
+class UnpackedSource(object):
+    '''
+    UnpackedSource extracts a source package into a temporary location and
+    gives you some convinient function for accessing it.
+    '''
+    def __init__(self, dscfilename):
+        '''
+        The dscfilename is a name of a DSC file that will be extracted.
+        '''
+        self.root_directory = os.path.join(mkdtemp(), 'root')
+        command = ('dpkg-source', '--no-copy', '--no-check', '-x', dscfilename,
+            self.root_directory)
+        # dpkg-source does not have a --quiet option
+        devnull = open(os.devnull, 'w')
+        check_call(command, stdout = devnull, stderr = devnull)
+        devnull.close()
+
+    def get_root_directory(self):
+        '''
+        Returns the name of the package's root directory which is the directory
+        where the debian subdirectory is located.
+        '''
+        return self.root_directory
+
+    def get_changelog_file(self):
+        '''
+        Returns a file object for debian/changelog or None if no such file exists.
+        '''
+        changelog_name = os.path.join(self.root_directory, 'debian', 'changelog')
+        try:
+            return open(changelog_name)
+        except IOError:
+            return None
+
+    def get_all_filenames(self):
+        '''
+        Returns an iterator over all filenames. The filenames will be relative
+        to the root directory.
+        '''
+        skip = len(self.root_directory) + 1
+        for root, _, files in os.walk(self.root_directory):
+            for name in files:
+                yield os.path.join(root[skip:], name)
+
+    def cleanup(self):
+        '''
+        Removes all temporary files.
+        '''
+        if self.root_directory is None:
+            return
+        parent_directory = os.path.dirname(self.root_directory)
+        rmtree(parent_directory)
+        self.root_directory = None
+
+    def __del__(self):
+        '''
+        Enforce cleanup.
+        '''
+        self.cleanup()
+
+
+class SourceContentsScanner(object):
+    '''
+    SourceContentsScanner provides a method scan() to scan the contents of a
+    DBSource object.
+    '''
+    def __init__(self, source_id):
+        '''
+        The argument source_id is the id of the DBSource object that
+        should be scanned.
+        '''
+        self.source_id = source_id
+
+    def scan(self):
+        '''
+        This method does the actual scan and fills in the associated SrcContents
+        property. It commits any changes to the database.
+        '''
+        session = DBConn().session()
+        source = session.query(DBSource).get(self.source_id)
+        fileset = set(source.scan_contents())
+        for filename in fileset:
+            source.contents.append(SrcContents(file = filename))
+        session.commit()
+        session.close()
+
+    @classmethod
+    def scan_all(class_, limit = None):
+        '''
+        The class method scan_all() scans all source using multiple processes.
+        The number of sources to be scanned can be limited with the limit
+        argument. Returns the number of processed and remaining packages as a
+        dict.
+        '''
+        session = DBConn().session()
+        query = session.query(DBSource).filter(DBSource.contents == None)
+        remaining = query.count
+        if limit is not None:
+            query = query.limit(limit)
+        processed = query.count()
+        pool = Pool()
+        for source in query.yield_per(100):
+            pool.apply_async(source_scan_helper, (source.source_id, ))
+        pool.close()
+        pool.join()

         remaining = remaining()
         session.close()
         return { 'processed': processed, 'remaining': remaining }
         remaining = remaining()
         session.close()
         return { 'processed': processed, 'remaining': remaining }

+
+def source_scan_helper(binary_id):
+    '''
+    This function runs in a subprocess.
+    '''
+    scanner = SourceContentsScanner(source_id)
+    scanner.scan()
+