maven.py

#
# Copyright (c) nexB Inc. and others. All rights reserved.
# ScanCode is a trademark of nexB Inc.
# SPDX-License-Identifier: Apache-2.0
# See http://www.apache.org/licenses/LICENSE-2.0 for the license text.
# See https://github.com/nexB/scancode-toolkit for support or download.
# See https://aboutcode.org for more information about nexB OSS projects.
#

import logging
import os.path
from pprint import pformat

import javaproperties
import lxml
from fnmatch import fnmatchcase
from packageurl import PackageURL
from pymaven import artifact
from pymaven import pom
from pymaven.pom import strip_namespace

from commoncode import fileutils
from packagedcode import models
from packagedcode.jar_manifest import parse_scm_connection
from packagedcode.jar_manifest import parse_manifest
from packagedcode.jar_manifest import get_normalized_java_manifest_data
from packagedcode.utils import normalize_vcs_url
from textcode import analysis
from typecode import contenttype

import saneyaml

TRACE = False

logger = logging.getLogger(__name__)

if TRACE:
    import sys
    logging.basicConfig(stream=sys.stdout)
    logger.setLevel(logging.DEBUG)

"""
Support for Maven POMs including resolution of variables using Maven properties
when possible.

We have seen Maven pom in three layout syles:

First case: a pom.xml inside a META-INF directory such as in
/META-INF/maven/log4j/log4j/pom.xml possibly with a pom.properties

Second case: a pom.xml at the root of a package codebase development tree
possibly with a pom.properties

Third case: a maven repo layout: the jars are side-by-side with a .pom and
there is no pom.properties check if there are side-by-side artifacts
"""


class MavenBasePackageHandler(models.DatafileHandler):

    @classmethod
    def assemble(cls, package_data, resource, codebase, package_adder=models.add_to_package):
        """
        Assembles from codebases where both a pom.xml and a MANIFEST.MF is present,
        otherwise uses the default assemble function and `assign_package_to_resources`
        function from respective DatafileHandlers.
        """
        if codebase.has_single_resource:
            yield from models.DatafileHandler.assemble(package_data, resource, codebase)
            return

        if not package_data.purl:
            return

        datafile_path = resource.path

        # This order is important as we want pom.xml to be used for package
        # creation and then to update from MANIFEST later
        manifest_path_pattern = '*/META-INF/MANIFEST.MF'
        nested_pom_xml_path_pattern = '*/META-INF/maven/**/pom.xml'
        datafile_name_patterns = (nested_pom_xml_path_pattern, manifest_path_pattern)

        is_manifest = fnmatchcase(datafile_path, manifest_path_pattern)
        is_pom_xml = fnmatchcase(datafile_path, nested_pom_xml_path_pattern)
        if is_manifest:
            meta_inf_resource = resource.parent(codebase)
        elif is_pom_xml:
            upward_segments = 4
            root = resource
            for _ in range(upward_segments):
                root = root.parent(codebase)
            meta_inf_resource = root
        else:
            yield from MavenPomXmlHandlerMixin.assemble(package_data, resource, codebase)
            return

        pom_xmls = []
        manifests = []
        for r in meta_inf_resource.walk(codebase):
            if fnmatchcase(r.path, manifest_path_pattern):
                manifests.append(r)
            if fnmatchcase(r.path, nested_pom_xml_path_pattern):
                pom_xmls.append(r)

        if len(pom_xmls) > 1:
            yield from MavenPomXmlHandlerMixin.assemble(package_data, resource, codebase)
            return

        if manifests and pom_xmls:
            # raise Exception(resource.path, meta_inf_resource, datafile_name_patterns, package_adder)
            parent_resource = meta_inf_resource.parent(codebase)
            if not parent_resource:
                parent_resource = meta_inf_resource
            yield from cls.assemble_from_many_datafiles_in_directory(
                    datafile_name_patterns=datafile_name_patterns,
                    directory=meta_inf_resource,
                    codebase=codebase,
                    package_adder=package_adder,
                    ignore_name_check=True,
                    parent_resource=parent_resource,
                )
        elif manifests and not pom_xmls:
            yield from JavaJarManifestHandlerMixin.assemble(package_data, resource, codebase, package_adder)
        elif pom_xmls and not manifests:
            yield from MavenPomXmlHandlerMixin.assemble(package_data, resource, codebase, package_adder)
        else:
            yield from models.DatafileHandler.assemble(package_data, resource, codebase, package_adder)


class JavaJarManifestHandler(MavenBasePackageHandler):
    datasource_id = 'java_jar_manifest'
    path_patterns = ('*/META-INF/MANIFEST.MF',)
    default_package_type = 'jar'
    default_primary_language = 'Java'
    description = 'Java JAR MANIFEST.MF'
    documentation_url = 'https://docs.oracle.com/javase/tutorial/deployment/jar/manifestindex.html'

    @classmethod
    def parse(cls, location, package_only=False):
        sections = parse_manifest(location)
        if sections:
            main_section = sections[0]
            manifest = get_normalized_java_manifest_data(main_section)
            if manifest:
                package_data = dict(**manifest,)
                yield models.PackageData.from_data(package_data, package_only)


class JavaJarManifestHandlerMixin(models.DatafileHandler):

    @classmethod
    def assign_package_to_resources(cls, package, resource, codebase, package_adder):
        # we want to root of the jar, two levels up
        parent = resource.parent(codebase)
        if parent:
            parent = parent.parent(codebase)
        if parent:
            models.DatafileHandler.assign_package_to_resources(
                package,
                resource=parent,
                codebase=codebase,
                package_adder=package_adder,
            )


class JavaOSGiManifestHandler(JavaJarManifestHandler):
    datasource_id = 'java_osgi_manifest'
    # This is an empty tuple to avoid getting back two packages
    # essentially this is the same as JavaJarManifestHandler
    path_patterns = ()
    default_primary_language = 'Java'
    description = 'Java OSGi MANIFEST.MF'
    default_package_type = 'osgi'


class MavenPomXmlHandler(MavenBasePackageHandler):
    datasource_id = 'maven_pom'
    # NOTE: Maven 1.x used project.xml
    path_patterns = ('*.pom', '*pom.xml',)
    default_package_type = 'maven'
    default_primary_language = 'Java'
    description = 'Apache Maven pom'
    documentation_url = 'https://maven.apache.org/pom.html'

    @classmethod
    def is_datafile(cls, location, filetypes=tuple()):
        """
        Return True if the file at location is highly likely to be a POM.
        """
        if super().is_datafile(location, filetypes=filetypes):
            return True

        T = contenttype.get_type(location)
        if not T.is_text:
            return

        maven_declarations = (
            b'http://maven.apache.org/POM/4.0.0',
            b'http://maven.apache.org/xsd/maven-4.0.0.xsd',
            b'<modelVersion>',
            # somehow we can still parse version 3 poms too
            b'<pomVersion>',
        )

        # check the POM version in the first 150 lines
        with open(location, 'rb') as pom:
            for n, line in enumerate(pom):
                if n > 150:
                    break
                if any(x in line for x in maven_declarations):
                    return True

    @classmethod
    def parse(cls, location, package_only=False, base_url='https://repo1.maven.org/maven2'):
        package_data = parse(
            location=location,
            datasource_id=cls.datasource_id,
            package_type=cls.default_package_type,
            primary_language=cls.default_primary_language,
            base_url=base_url,
            package_only=package_only,
        )
        if package_data:
            yield package_data

    @classmethod
    def get_top_level_resources(cls, manifest_resource, codebase):
        """
        Yield Resources that are top-level based on a JAR's directory structure
        """
        if 'META-INF' in manifest_resource.path:
            path_segments = manifest_resource.path.split('META-INF')
            leading_segment = path_segments[0].strip()
            meta_inf_dir_path = f'{leading_segment}/META-INF'
            meta_inf_resource = codebase.get_resource(meta_inf_dir_path)
            if meta_inf_resource:
                yield meta_inf_resource
                yield from meta_inf_resource.walk(codebase)


class MavenPomXmlHandlerMixin(models.DatafileHandler):

    @classmethod
    def assign_package_to_resources(cls, package, resource, codebase, package_adder):
        """
        Set the "for_packages" attributes to ``package``  for the whole
        resource tree of a ``resource`` object in the ``codebase``.

        handle the various cases where we can find a POM.
        """

        if resource.path.endswith('.pom'):
            # we only treat the parent as the root
            return models.DatafileHandler.assign_package_to_parent_tree(
                package,
                resource,
                codebase,
                package_adder
            )

        # the root is either the parent or further up for poms stored under
        # a META-INF dir

        # assert resource.name.endswith('pom.xml')

        root = None
        if resource.path.endswith(f'META-INF/maven/{package.namespace}/{package.name}/pom.xml'):
            # First case: a pom.xml inside a META-INF directory such as in
            # /META-INF/maven/log4j/log4j/pom.xml: This requires going up 5 times
            upward_segments = 4
            root = resource
            for _ in range(upward_segments):
                root = root.parent(codebase)

            number_poms = 0
            for child in root.walk(codebase):
                if 'pom.xml' in child.path:
                    number_poms += 1

            if number_poms > 1:
                root = resource
            else:
                root = root.parent(codebase)

        else:
            # Second case: a pom.xml at the root of codebase tree
            # FIXME: handle the cases opf parent POMs and nested POMs
            root = resource.parent(codebase)

        # get a root in all cases
        if not root:
            root = resource.parent(codebase)

        return models.DatafileHandler.assign_package_to_resources(
            package,
            resource=root,
            codebase=codebase,
            package_adder=package_adder
        )


# TODO: assemble with its pom!!
class MavenPomPropertiesHandler(models.NonAssemblableDatafileHandler):
    datasource_id = 'maven_pom_properties'
    path_patterns = ('*/pom.properties',)
    default_package_type = 'maven'
    default_primary_language = 'Java'
    description = 'Apache Maven pom properties file'
    documentation_url = 'https://maven.apache.org/pom.html'

    @classmethod
    def parse(cls, location, package_only=False):
        """
        Yield PackageData from a pom.properties file (which is typically side-
        by-side with its pom file.)
        """
        with open(location) as props:
            properties = javaproperties.load(props) or {}
            if TRACE:
                logger.debug(f'MavenPomPropertiesHandler.parse: properties: {properties!r}')
            if properties:
                yield from cls.parse_pom_properties(properties=properties, package_only=package_only)

    @classmethod
    def parse_pom_properties(cls, properties, package_only=False):
        namespace = properties.pop("groupId", None)
        name = properties.pop("artifactId", None)
        version = properties.pop("version", None)
        if properties:
            extra_data = dict(pom_properties=properties)
        else:
            extra_data = {}

        package_data = dict(
            datasource_id=cls.datasource_id,
            type=cls.default_package_type,
            primary_language=cls.default_primary_language,
            name=name,
            namespace=namespace,
            version=version,
            extra_data=extra_data,
        )
        yield models.PackageData.from_data(package_data, package_only)


def build_url(
    group_id,
    artifact_id,
    version,
    filename=None,
    base_url='https://repo1.maven.org/maven2',
):
    """
    Return a download URL for a Maven artifact built from its POM "coordinates".
    """
    filename = filename or ''
    if group_id:
        group_id = group_id.replace('.', '/')
        path = f'{group_id}/{artifact_id}/{version}'
    else:
        path = f'{artifact_id}/{version}'

    return f'{base_url}/{path}/{filename}'


def build_filename(artifact_id, version, extension, classifier=None):
    """
    Return a filename for a Maven artifact built from its coordinates.
    """
    extension = extension or ''
    classifier = classifier or ''
    if classifier:
        classifier = f'-{classifier}'
    return f'{artifact_id}-{version}{classifier}.{extension}'


class ParentPom(artifact.Artifact):
    """
    A minimal Artifact subclass used to store parent poms when no POM
    file is available for these.
    """

    def __init__(self, coordinate):
        super().__init__(coordinate)

        # add empty, pom.Pom-class-like empty attributes
        self.client = None
        self.dependencies = {}
        self.dependency_management = {}
        self.parent = None
        self.properties = {}
        # TODO: ????
        # self.pom_data/self.pom_data = None

    def resolve(self, **properties):
        """
        Resolve possible parent POM properties using the provided
        properties.
        """
        if not properties:
            return
        self.group_id = MavenPom._replace_props(self.group_id, properties)
        self.artifact_id = MavenPom._replace_props(self.artifact_id, properties)
        self.version = MavenPom._replace_props(self.version, properties)

    def to_dict(self):
        """
        Return a mapping representing this POM
        """
        return {
            'group_id': self.group_id,
            'artifact_id': self.artifact_id,
            'version': str(self.version) if self.version else None,
            'classifier': self.classifier,
            'type': self.type,
        }


class MavenPom(pom.Pom):

    def __init__(self, location=None, text=None):
        """
        Build a POM from a location or unicode text.
        """
        assert (location or text) and (not (location and text))

        if location:
            try:
                with open(location) as fh:
                    xml_text = fh.read()
            except UnicodeDecodeError as _a:
                xml_text = analysis.unicode_text(location)
        else:
            xml_text = text
        xml_text = strip_namespace(xml_text)
        xml_text = xml_text.encode('utf-8')
        if TRACE:
            logger.debug('MavenPom.__init__: xml_text: {}'.format(xml_text))

        self._pom_data = lxml.etree.fromstring(xml_text, parser=pom.POM_PARSER)  # NOQA

        # collect and then remove XML comments from the XML elements tree
        self.comments = self._get_comments()
        lxml.etree.strip_tags(self._pom_data, lxml.etree.Comment)  # NOQA

        # FIXME: we do not use a client for now.
        # There are pending issues at pymaven to address this
        self._client = None

        self.model_version = self._get_attribute('modelVersion')
        if not self.model_version:
            # for older POM version 3
            self.model_version = self._get_attribute('pomVersion')
        self.group_id = self._get_attribute('groupId')
        self.artifact_id = self._get_attribute('artifactId')
        if TRACE:
            logger.debug('MavenPom.__init__: self.artifact_id: {}'.format(self.artifact_id))

        self.version = self._get_attribute('version')
        self.classifier = self._get_attribute('classifier')
        self.packaging = self._get_attribute('packaging') or 'jar'
        self.name = self._get_attribute('name')
        self.description = self._get_attribute('description')
        self.inception_year = self._get_attribute('inceptionYear')
        self.url = self._get_attribute('url')
        self.organization_name = self._get_attribute('organization/name')
        self.organization_url = self._get_attribute('organization/url')
        self.licenses = list(self._find_licenses())
        self.developers = list(self._find_parties('developers/developer'))
        self.contributors = list(self._find_parties('contributors/contributor'))
        self.mailing_lists = list(self._find_mailing_lists())
        self.scm = self._find_scm()
        self.issue_management = self._find_issue_management()
        self.ci_management = self._find_ci_management()
        self.distribution_management = self._find_distribution_management()
        self.repositories = list(self._find_repositories('repositories/repository'))
        self.plugin_repositories = list(self._find_repositories('pluginRepositories/pluginRepository'))
        self.modules = self._get_attributes_list('modules/module')

        # FIXME: this attribute should be collected with the parent but
        # is not retrieved yet by pymaven it points to the relative path
        # where to find the full parent POM
        self.parent_relative_path = self._get_attribute('relativePath')  # or '../pom.xml_text'

        # FIXME: Other types that are not collected for now (or
        # indirectly through dependencies management) include: build,
        # reporting, profiles, etc

        # dynamic attributes
        self._parent = None
        self._dep_mgmt = None
        self._dependencies = None
        self._properties = None

    def _extra_properties(self):
        """
        Return a mapping of extra properties
        """
        properties = {}
        for name in build_property_variants('classifier'):
            properties[name] = self.classifier

        for name in build_property_variants('packaging'):
            properties[name] = self.packaging

        for name in build_property_variants('organization.name'):
            properties[name] = self.organization_name

        for name in build_property_variants('organization.url'):
            properties[name] = self.organization_url

        # TODO: collect props defined in a properties file side-by-side the pom file
        # see https://maven.apache.org/shared/maven-archiver/#class_archive
        # this applies to POMs stored inside a JAR or in a plain directory

        return properties

    @classmethod
    def _replace_props(cls, text, properties):
        if not text:
            return text

        def subfunc(matchobj):
            """Return the replacement value for a matched property key."""
            key = matchobj.group(1)
            # does this key contain a substring?
            real_key, start_end = _get_substring_expression(key)
            if not start_end:
                value = properties.get(key)

                return value
            # apply the substring transform
            value = properties.get(real_key)
            if not value:
                return value
            start, end = start_end
            return substring(value, start, end)

        result = pom.PROPERTY_RE.sub(subfunc, text)
        # we loop a maximum of 5 times to avoid cases of infinite recursion
        cycles = 5
        while cycles > 0 and result and pom.PROPERTY_RE.match(result):
            result = pom.PROPERTY_RE.sub(subfunc, result)
            cycles -= 1

        if not result:
            result = text
        return result.strip()

    def _replace_properties(self, text, properties=None):
        # copied from pymavem.pom.Pom
        if not text:
            return text
        if properties is None:
            properties = self.properties
        return MavenPom._replace_props(text, properties)

    def resolve(self, **extra_properties):
        """
        Resolve POM Maven "properties" in attribute values and inherit
        from parent. Update the POM attributes in place. Use the extra
        keywords to override and supplement existing properties.
        """
        # inherit first to get essential parent properties
        # FIXME: the parent needs to be resolved first!? but we have a chicken and egg problem then
        if self.parent:
            pass

        self._inherit_from_parent()

        # then collect properties + extra
        properties = dict(self.properties)
        properties.update(self._extra_properties())

        # "extra" properties are things that would be loaded from a
        # properties files and provided as is they overwrite any
        # existing property
        properties.update(extra_properties)

        if TRACE:
            logger.debug('MavenPom.resolve: properties before self-resolution:\n{}'.format(pformat(properties)))

        # FIXME: we could remove any property that itself contains
        # ${property} as we do not know how to resolve these
        # recursively. Or we could do a single pass on properties to
        # resolve values against themselves
        for key, value in list(properties.items()):
            properties[key] = MavenPom._replace_props(value, properties)

        if TRACE:
            logger.debug('MavenPom.resolve: used properties:\n{}'.format(pformat(properties)))

        # these attributes are plain strings
        plain_attributes = [
            'group_id',
            'version',
            'classifier',
            'packaging',
            'name',
            'description',
            'inception_year',
            'url',
            'organization_name',
            'organization_url',
        ]
        for attr in plain_attributes:
            attr_val = getattr(self, attr, None)
            if not attr_val:
                continue
            resolved = self._replace_properties(attr_val, properties)
            setattr(self, attr, resolved)

        # these attributes are mappings
        mapping_attributes = [
            'scm',
            'issue_management',
            'ci_management',
        ]
        for map_attr in mapping_attributes:
            mapping = getattr(self, map_attr, {})
            if not mapping:
                continue
            for key, value in mapping.items():
                if not value:
                    continue
                mapping[key] = self._replace_properties(value, properties)

        # these attributes are lists of mappings
        mappings_list_attributes = [
            'repositories',
            'plugin_repositories',
        ]
        for lmap_attr in mappings_list_attributes:
            lmapping = getattr(self, lmap_attr, [])
            if not lmapping:
                continue
            for mapping in lmapping:
                for key, value in mapping.items():
                    if not value:
                        continue
                    mapping[key] = self._replace_properties(value, properties)

        # these attributes below are complex nested dicts and/or lists mappings

        for scope, dependencies in self.dependencies.items():
            resolved_deps = []
            # FIXME: this is missing the packaging/type and classifier
            for (group, artifact, version,), required in dependencies:
                group = self._replace_properties(group, properties)
                artifact = self._replace_properties(artifact, properties)
                version = self._replace_properties(version, properties)
                # skip weird damaged POMs such as
                # http://repo1.maven.org/maven2/net/sourceforge/findbugs/coreplugin/1.0.0/coreplugin-1.0.0.pom
                if not group or not artifact:
                    continue
                resolved_deps.append(((group, artifact, version,), required))
            self._dependencies[scope] = resolved_deps

        for (group, artifact), (version, scope, optional) in self.dependency_management.items():
            group = self._replace_properties(group, properties)
            artifact = self._replace_properties(artifact, properties)
            version = self._replace_properties(version, properties)
            required = not optional
            resolved_dep = (group, artifact, version,), required
            # we craft a scope prefix for these
            scope = f'dependencyManagement'
            if scope in self.dependencies:
                scope_deps = self._dependencies[scope]
                # TODO: Check we don't insert duplicates
                scope_deps.append(resolved_dep)
            else:
                self._dependencies[scope] = [resolved_dep]

        if TRACE:
            logger.debug('MavenPom.resolve: artifactId after resolve: {}'.format(self.artifact_id))

        # TODO: add:
        # nest dict: 'distribution_management',
        # nest list: 'mailing_lists',

    def _inherit_from_parent(self):
        """
        Update attributes using inheritance from parent attributes. For
        instance, the parent group_id is used if group_id is not defined.
        """
        # TODO: there are more attributes (all) that can be inherited
        if not self.parent:
            return
        if self.group_id is None and self.parent.group_id:
            self.group_id = self.parent.group_id
            if TRACE: logger.debug('_inherit_from_parent: group_id: {}'.format(self.parent.group_id))
        if self.version is None and self.parent.version:
            self.version = str(self.parent.version)
            if TRACE: logger.debug('_inherit_from_parent: version: {}'.format(self.parent.version))
        if not self.classifier is None and self.parent.classifier:
            self.classifier = self.parent.classifier
            if TRACE: logger.debug('_inherit_from_parent: classifier: {}'.format(self.parent.classifier))

        # FIXME: the parent may need to be resolved too?
        # special handling for URLs: see
        # http://maven.apache.org/ref/3.5.0/maven-model-builder/index.html#Inheritance_Assembly
        # Notice that the 5 URLs from the model:
        # project.url,
        # project.scm.connection, project.scm.developerConnection, project.scm.url
        # project.distributionManagement.site.url)
        # ... have a special inheritance handling: if not configured in
        # current model, the inherited value is the parent's one with
        # current artifact id appended.
        if (self.url is None
            and hasattr(self.parent, 'url')
            and getattr(self.parent, 'url', None)
            and self.artifact_id):
            # FIXME: this is not the way to join URLs parts!
            self.url = self.parent.url + self.artifact_id

        # FIXME: this is not the way to join URLs parts!
        parent_scm = getattr(self.parent, 'scm', None)
        if self.scm and parent_scm and self.artifact_id:
            ps_url = parent_scm.get('url')
            if not self.scm.get('url') and ps_url:
                # FIXME: this is not the way to join URLs parts!
                self.scm['url'] = ps_url + self.artifact_id

            ps_connection = parent_scm.get('connection')
            if not self.scm.get('connection') and ps_connection:
                # FIXME: this is not the way to join URLs parts!
                self.scm['connection'] = ps_connection + self.artifact_id

            ps_devconnection = parent_scm.get('developer_connection')
            if not self.scm.get('developer_connection') and ps_devconnection:
                # FIXME: this is not the way to join URLs parts!
                self.scm['developer_connection'] = ps_devconnection + self.artifact_id

        # TODO: distribution_management.site.url

    def _pom_factory(self, group_id, artifact_id, version):
        return ParentPom('%s:%s:pom:%s' % (group_id, artifact_id, version))

    def _get_attribute(self, xpath, xml=None):
        """Return a single value text attribute for a given xpath or None."""
        if xml is None:
            xml = self.pom_data
        attr = xml.findtext(xpath)
        val = attr and attr.strip() or None
        if TRACE:
            if 'artifactId' in xpath:
                logger.debug('MavenPom._get_attribute: xpath: {}'.format(xpath))
                logger.debug('MavenPom._get_attribute: xml: {}'.format(xml))
        return val

    def _get_attributes_list(self, xpath, xml=None):
        """Return a list of text attributes for a given xpath or empty list."""
        if xml is None:
            xml = self.pom_data
        attrs = xml.findall(xpath)
        attrs = [attr.text for attr in attrs]
        return [attr.strip() for attr in attrs if attr and attr.strip()]

    def _get_comments(self, xml=None):
        """Return a list of comment texts or an empty list."""
        if xml is None:
            xml = self.pom_data
        comments = [c.text for c in xml.xpath('//comment()')]
        return [c.strip() for c in comments if c and c.strip()]

    def _find_licenses(self):
        """Return an iterable of license mappings."""
        for lic in self.pom_data.findall('licenses/license'):
            yield dict([
                ('name', self._get_attribute('name', lic)),
                ('url', self._get_attribute('url', lic)),
                ('comments', self._get_attribute('comments', lic)),
                # arcane and seldom used
                ('distribution', self._get_attribute('distribution', lic)),
            ])

    def _find_parties(self, key='developers/developer'):
        """Return an iterable of party mappings for a given xpath."""
        for party in self.pom_data.findall(key):
            yield dict([
                ('id', self._get_attribute('id', party)),
                ('name', self._get_attribute('name', party)),
                ('email', self._get_attribute('email', party)),
                ('url', self._get_attribute('url', party)),
                ('organization', self._get_attribute('organization', party)),
                ('organization_url', self._get_attribute('organizationUrl', party)),
                ('roles', [role.findtext('.') for role in party.findall('roles/role')]),
            ])

    def _find_mailing_lists(self):
        """Return an iterable of mailing lists mappings."""
        for ml in self.pom_data.findall('mailingLists/mailingList'):
            archive_url = self._get_attribute('archive', ml)
            # TODO: add 'otherArchives/otherArchive' as lists?
            yield dict([
                ('name', self._get_attribute('name', ml)),
                ('archive_url', archive_url),
            ])

    def _find_scm(self):
        """Return a version control/scm mapping."""
        scm = self.pom_data.find('scm')
        if scm is None:
            return {}
        return dict([
            ('connection', self._get_attribute('connection', scm)),
            ('developer_connection', self._get_attribute('developer_connection', scm)),
            ('url', self._get_attribute('url', scm)),
            ('tag', self._get_attribute('tag', scm)),
        ])

    def _find_issue_management(self):
        """Return an issue management mapping."""
        imgt = self.pom_data.find('issueManagement')
        if imgt is None:
            return {}
        return dict([
            ('system', self._get_attribute('system', imgt)),
            ('url', self._get_attribute('url', imgt)),
        ])

    def _find_ci_management(self):
        """Return a CI mapping."""
        cimgt = self.pom_data.find('ciManagement')
        if cimgt is None:
            return {}
        return dict([
            ('system', self._get_attribute('system', cimgt)),
            ('url', self._get_attribute('url', cimgt)),
        ])

    def _find_repository(self, xpath, xml=None):
        """Return a repository mapping for an xpath."""
        if xml is None:
            xml = self.pom_data
        repo = xml.find(xpath)
        if repo is None:
            return {}
        return dict([
            ('id', self._get_attribute('id', repo)),
            ('name', self._get_attribute('name', repo)),
            ('url', self._get_attribute('url', repo)),
        ])

    def _find_distribution_management(self):
        """Return a distribution management mapping."""
        dmgt = self.pom_data.find('distributionManagement')
        if dmgt is None:
            return {}
        return dict([
            ('download_url', self._get_attribute('distributionManagement/downloadUrl')),
            ('site', self._find_repository('distributionManagement/site')),
            ('repository', self._find_repository('distributionManagement/repository')),
            ('snapshot_repository', self._find_repository('distributionManagement/snapshotRepository'))
        ])

    def _find_repositories(self, key='repositories/repository'):
        """Return an iterable or repository mappings for an xpath."""
        for repo in self.pom_data.findall(key):
            rep = self._find_repository('.', repo)
            if rep:
                yield rep

    def to_dict(self):
        """
        Return a mapping representing this POM.
        """
        dependencies = {}
        for scope, deps in self.dependencies.items():
            dependencies[scope] = [
                dict([
                    ('group_id', gid),
                    ('artifact_id', aid),
                    ('version', version),
                    ('required', required),
                ])
            for ((gid, aid, version), required) in deps]

        return dict([
            ('model_version', self.model_version),
            ('group_id', self.group_id),
            ('artifact_id', self.artifact_id),
            ('version', self.version),
            ('classifier', self.classifier),
            ('packaging', self.packaging),

            ('parent', self.parent.to_dict() if self.parent else {}),

            ('name', self.name),
            ('description', self.description),
            ('inception_year', self.inception_year),
            ('url', self.url),
            ('organization_name', self.organization_name),
            ('organization_url', self.organization_url),

            ('licenses', self.licenses or []),

            ('developers', self.developers or []),
            ('contributors', self.contributors or []),

            ('modules', self.modules or []),
            ('mailing_lists', self.mailing_lists),
            ('scm', self.scm),
            ('issue_management', self.issue_management),
            ('ci_management', self.ci_management),
            ('distribution_management', self.distribution_management),
            ('repositories', self.repositories),
            ('plugin_repositories', self.plugin_repositories),
            ('dependencies', dependencies or {}),
        ])


def build_property_variants(name):
    """
    Return an iterable of property variant names given a a property name.
    """
    yield name
    yield 'project.{}'.format(name)
    yield 'pom.{}'.format(name)


def _get_substring_expression(text):
    """
    Return a tuple of (text, start/end) such that:

    - if there is a substring() expression in text, the returned text
    has been stripped from it and start/end is a tuple representing
    slice indexes for the substring expression.

    - if there is no substring() expression in text, text is returned
    as-is and start/end is None.

    For example:
    >>> assert ('pom.artifactId', (8, None)) == _get_substring_expression('pom.artifactId.substring(8)')
    >>> assert ('pom.artifactId', None) == _get_substring_expression('pom.artifactId')
    """
    key, _, start_end = text.partition('.substring(')
    if not start_end:
        return text, None

    start_end = start_end.rstrip(')')
    start_end = [se.strip() for se in start_end.split(',')]

    # we cannot parse less than 1 and more than 2 slice indexes
    if len(start_end) not in (1, 2):
        return text, None

    # we cannot parse slice indexes that are not numbers
    if not all(se.isdigit() for se in start_end):
        return text, None
    start_end = [int(se) for se in start_end]

    if len(start_end) == 1:
        start = start_end[0]
        end = None
    else:
        start, end = start_end

    return key, (start, end)


def substring(s, start, end):
    """
    Return a slice of s based on start and end indexes (that can be None).
    """
    startless = start is None
    endless = end is None
    if startless and endless:
        return s
    if endless:
        return s[start:]
    if startless:
        return s[:end]
    return s[start:end]


def has_basic_pom_attributes(pom):
    """
    Return True if a POM object has basic attributes needed to make this
    a POM.
    """
    basics = pom.model_version and pom.group_id and pom.artifact_id
    if TRACE and not basics:
        logger.debug(
            'has_basic_pom_attributes: not a POM, incomplete GAV: '
            '"{}":"{}":"{}"'.format(pom.model_version, pom.group_id, pom.artifact_id))
    return basics


def get_maven_pom(location=None, text=None):
    """
    Return a MavenPom object from a POM file at `location` or provided as a
    `text` string.
    """
    pom = MavenPom(location=location, text=text)

    extra_properties = {}

    # do we have a pom.properties file side-by-side?
    # FIXME: we should treat pom.properties as a datafile
    if location and os.path.exists(location):
        parent = fileutils.parent_directory(location)
        pom_properties = os.path.join(parent, 'pom.properties')
        if os.path.exists(pom_properties):
            with open(pom_properties) as props:
                properties = javaproperties.load(props) or {}
                if TRACE:
                    logger.debug(f'get_maven_pom: properties: {properties!r}')
            extra_properties.update(properties)
    pom.resolve(**extra_properties)
    # TODO: we cannot do much without these??
    hbpa = has_basic_pom_attributes(pom)

    if not hbpa:
        if TRACE:
            logger.debug(f'get_maven_pom: has_basic_pom_attributes: {hbpa}')
        return
    return pom


SUPPORTED_PACKAGING = set([
    u'aar',
    u'apk',
    u'gem',
    u'jar',
    u'nar',
    u'pom',
    u'so',
    u'swc',
    u'tar',
    u'tar.gz',
    u'war',
    u'xar',
    u'zip',
])


def get_dependencies(pom):
    """
    Return a list of Dependent package objects found in a MavenPom `pom` object.
    """
    dependencies = []
    for scope, deps in pom.dependencies.items():
        if TRACE:
            logger.debug('parse: dependencies.deps: {}'.format(deps))
        if scope:
            scope = scope.strip().lower()
        if not scope:
            # maven default
            scope = 'compile'

        for (dgroup_id, dartifact_id, dversion), drequired in deps:
            if TRACE:
                logger.debug('parse: dependencies.deps: {}, {}, {}, {}'.
                             format(dgroup_id, dartifact_id, dversion, drequired))
            # pymaven whart
            if dversion == 'latest.release':
                dversion = None

            is_pinned = bool(dversion and not any(c in dversion for c in '$[,]'))

            dqualifiers = {}
            # FIXME: this is missing from the original Pom parser
            # classifier = dep.get('classifier')
            # if classifier:
            #     qualifiers['classifier'] = classifier
            #
            # packaging = dep.get('type')
            # if packaging and packaging != 'jar':
            #     qualifiers['packaging'] = packaging

            if is_pinned:
                dpurl = models.PackageURL(
                    type='maven',
                    namespace=dgroup_id,
                    name=dartifact_id,
                    version=dversion,
                    qualifiers=dqualifiers or None,
                )
            else:
                dpurl = models.PackageURL(
                    type='maven',
                    namespace=dgroup_id,
                    name=dartifact_id,
                    qualifiers=dqualifiers or None,
                )

            # TODO: handle dependency management and pom type
            is_runtime = scope in ('runtime', 'system', 'provided')
            is_optional = bool(scope in ('test', 'compile',) or not drequired)

            dep_pack = models.DependentPackage(
                purl=str(dpurl),
                extracted_requirement=dversion,
                scope=scope,
                is_runtime=is_runtime,
                is_optional=is_optional,
                is_pinned=is_pinned,
            )
            dependencies.append(dep_pack)

    return dependencies


def get_parties(pom):
    """
    Return a list of Party object found in a MavenPom `pom` object.
    """
    parties = []
    for dev in pom.developers:
        parties.append(
            models.Party(
                type=models.party_person,
                name=dev['name'],
                role='developer',
                email=dev['email'],
                url=dev['url']))

    for cont in pom.contributors:
        parties.append(
            models.Party(
                type=models.party_person,
                name=cont['name'],
                role='contributor',
                email=cont['email'],
                url=cont['url']))

    # FIXME: we are skipping most other organization related fields, roles and the id
    party_name = pom.organization_name
    party_url = pom.organization_url
    if party_name or party_url:
        parties.append(
            models.Party(
                type=models.party_org,
                name=party_name,
                role='owner',
                url=party_url))

    return parties


def get_urls(namespace, name, version, qualifiers, base_url='https://repo1.maven.org/maven2', **kwargs):
    """
    Return a mapping of URLs.
    """
    if not namespace or not name:
        return {}
    repository_homepage_url = build_url(
        group_id=namespace,
        artifact_id=name,
        version=version,
        base_url=base_url,
    )

    qualifiers = qualifiers or {}
    filename = build_filename(
        artifact_id=name,
        version=version,
        extension=qualifiers.get('type') or 'jar',
        classifier=qualifiers.get('classifier'),
    )

    repository_download_url = build_url(
        group_id=namespace,
        artifact_id=name,
        version=version,
        filename=filename,
        base_url=base_url,
    )

    # treat the POM as "API"
    pom_filename = build_filename(
        artifact_id=name,
        version=version,
        extension='pom',
    )

    api_data_url = build_url(
        group_id=namespace,
        artifact_id=name,
        version=version,
        filename=pom_filename,
        base_url=base_url,
    )

    return dict(
        repository_homepage_url=repository_homepage_url,
        repository_download_url=repository_download_url,
        api_data_url=api_data_url,
    )


def parse(
    location,
    datasource_id,
    package_type,
    primary_language,
    base_url='https://repo1.maven.org/maven2',
    package_only=False,
):
    """
    Return Packagedata objects from parsing a Maven pom file at `location` or
    using the provided `text` (one or the other but not both).
    """
    package = _parse(
        datasource_id=datasource_id,
        package_type=package_type,
        primary_language=primary_language,
        location=location,
        base_url=base_url,
        package_only=package_only,
    )
    if package:
        return package


def _parse(
    datasource_id,
    package_type,
    primary_language,
    location=None,
    text=None,
    base_url='https://repo1.maven.org/maven2',
    package_only=False,
):
    """
    Yield Packagedata objects from parsing a Maven pom file at `location` or
    using the provided `text` (one or the other but not both).
    """
    pom = get_maven_pom(location=location, text=text)

    if not pom:
        return

    if TRACE:
        ptd = pformat(pom.to_dict())
        logger.debug(f'PomXmlHandler.parse: pom:.to_dict()\n{ptd}')

    version = pom.version
    # pymaven whart
    if version == 'latest.release':
        version = None

    qualifiers = {}
    classifier = pom.classifier
    if classifier:
        qualifiers['classifier'] = classifier

    packaging = pom.packaging
    if packaging:
        extension = get_extension(packaging)
        if extension and extension not in ('jar', 'pom'):
            # we use type as in the PURL spec: this is a problematic field with
            # complex defeinition in Maven
            qualifiers['type'] = extension

    extracted_license_statement = pom.licenses

    group_id = pom.group_id
    artifact_id = pom.artifact_id

    # craft a source package purl for the main binary
    source_packages = []
    is_main_binary_jar = not classifier and all([group_id, artifact_id, version])
    if is_main_binary_jar:
        spurl = PackageURL(
            type=package_type,
            namespace=group_id,
            name=artifact_id,
            version=version,
            # we hardcode the source qualifier for now...
            qualifiers=dict(classifier='sources'))
        source_packages = [spurl.to_string()]

    pname = pom.name or ''
    pdesc = pom.description or ''
    if pname == pdesc:
        description = pname
    else:
        description = [d for d in (pname, pdesc) if d]
        description = '\n'.join(description)

    issue_mngt = pom.issue_management or {}
    bug_tracking_url = issue_mngt.get('url')

    scm = pom.scm or {}
    urls = build_vcs_and_code_view_urls(scm)
    urls.update(get_urls(
        namespace=group_id,
        name=artifact_id, version=version,
        qualifiers=qualifiers,
        base_url=base_url,
    ))

    # FIXME: there are still other data to map in a PackageData
    package_data = dict(
        datasource_id=datasource_id,
        type=package_type,
        primary_language=primary_language,
        namespace=group_id,
        name=artifact_id,
        version=version,
        qualifiers=qualifiers or None,
        description=description or None,
        homepage_url=pom.url or None,
        extracted_license_statement=extracted_license_statement or None,
        parties=get_parties(pom),
        dependencies=get_dependencies(pom),
        source_packages=source_packages,
        bug_tracking_url=bug_tracking_url,
        **urls,
    )
    return MavenPackageData.from_data(package_data, package_only)


class MavenPackageData(models.PackageData):

    datasource_id = 'maven_pom'

    @classmethod
    def get_license_detections_for_extracted_license_statement(
        cls,
        extracted_license,
        try_as_expression=True,
        approximate=True,
        expression_symbols=None,
    ):
        from packagedcode.licensing import get_normalized_license_detections
        from packagedcode.licensing import get_license_detections_for_extracted_license_statement

        if not cls.check_extracted_license_statement_structure(extracted_license):
            return get_normalized_license_detections(
                extracted_license=extracted_license,
                try_as_expression=try_as_expression,
                approximate=approximate,
                expression_symbols=expression_symbols,
            )

        new_extracted_license = extracted_license.copy()

        for license_entry in new_extracted_license:
            license_entry.pop("distribution")
            if not license_entry.get("name"):
                license_entry.pop("name")
            if not license_entry.get("url"):
                license_entry.pop("url")
            if not license_entry.get("comments"):
                license_entry.pop("comments")

        extracted_license_statement = saneyaml.dump(new_extracted_license)

        return get_license_detections_for_extracted_license_statement(
            extracted_license_statement=extracted_license_statement,
            try_as_expression=try_as_expression,
            approximate=approximate,
            expression_symbols=expression_symbols,
        )

    @classmethod
    def check_extracted_license_statement_structure(cls, extracted_license):

        is_list_of_mappings = False
        if not isinstance(extracted_license, list):
            return is_list_of_mappings
        else:
            is_list_of_mappings = True

        for extracted_license_item in extracted_license:
            if not isinstance(extracted_license_item, dict):
                is_list_of_mappings = False
                break

        return is_list_of_mappings


def build_vcs_and_code_view_urls(scm):
    """
    Return a mapping of vcs_url and code_view_url from a Maven `scm` mapping.
    For example:

    >>> scm = dict(
    ...     connection='scm:git:git@github.com:histogrammar/histogrammar-scala.git',
    ...     tag='HEAD',
    ...     url='https://github.com/histogrammar/histogrammar-scala')
    >>> expected = {
    ...     'vcs_url': 'git+https://github.com/histogrammar/histogrammar-scala.git',
    ...     'code_view_url': 'https://github.com/histogrammar/histogrammar-scala'}
    >>> assert build_vcs_and_code_view_urls(scm) == expected
    """

    vcs_url = scm.get('connection') or None
    code_view_url = scm.get('url') or None

    if code_view_url:
        cvu = normalize_vcs_url(code_view_url) or None
        if cvu:
            code_view_url = cvu

    if not vcs_url:
        if code_view_url:
            # we can craft a vcs_url in some cases
            vcs_url = code_view_url
        return dict(vcs_url=vcs_url, code_view_url=code_view_url,)

    vcs_url = parse_scm_connection(vcs_url)

    # TODO: handle tag
    # vcs_tag = scm.get('tag')

    return dict(vcs_url=vcs_url, code_view_url=code_view_url,)


def get_extension(packaging):
    """
    We only care for certain artifacts extension/packaging/classifier.

    Maven has some intricate interrelated values for these fields
        type, extension, packaging, classifier, language
    See http://maven.apache.org/ref/3.5.4/maven-core/artifact-handlers.html

    These are the defaults:

    type            extension   packaging    classifier   language
    --------------------------------------------------------------
    pom             = type      = type                    none
    jar             = type      = type                    java
    maven-plugin    jar         = type                    java
    ejb             jar         ejb = type                java
    ejb3            = type      ejb3 = type               java
    war             = type      = type                    java
    ear             = type      = type                    java
    rar             = type      = type                    java
    par             = type      = type                    java
    java-source     jar         = type        sources     java
    javadoc         jar         = type        javadoc     java
    ejb-client      jar         ejb           client      java
    test-jar        jar         jar           tests       java
    """

    extensions = set([
        'ejb3',
        'ear',
        'aar',
        'apk',
        'gem',
        'jar',
        'nar',
        'pom',
        'so',
        'swc',
        'tar',
        'tar.gz',
        'war',
        'xar',
        'zip'
    ])

    if packaging in extensions:
        return packaging
    else:
        return 'jar'