model_dictize.py

# encoding: utf-8

'''
These dictize functions generally take a domain object (such as Package) and
convert it to a dictionary, including related objects (e.g. for Package it
includes PackageTags, PackageExtras, PackageGroup etc).

The basic recipe is to call:

    dictized = ckan.lib.dictization.table_dictize(domain_object)

which builds the dictionary by iterating over the table columns.
'''
import datetime
import urlparse

from ckan.common import config
from sqlalchemy.sql import select

import ckan.logic as logic
import ckan.plugins as plugins
import ckan.lib.helpers as h
import ckan.lib.dictization as d
import ckan.authz as authz
import ckan.lib.search as search
import ckan.lib.munge as munge

## package save

def group_list_dictize(obj_list, context,
                       sort_key=lambda x: x['display_name'], reverse=False,
                       with_package_counts=True,
                       include_groups=False,
                       include_tags=False,
                       include_extras=False):

    group_dictize_context = dict(context.items()[:])
    # Set options to avoid any SOLR queries for each group, which would
    # slow things further.
    group_dictize_options = {
            'packages_field': 'dataset_count' if with_package_counts else None,
            # don't allow packages_field='datasets' as it is too slow
            'include_groups': include_groups,
            'include_tags': include_tags,
            'include_extras': include_extras,
            'include_users': False,  # too slow - don't allow
            }
    if with_package_counts and 'dataset_counts' not in group_dictize_context:
        # 'dataset_counts' will already be in the context in the case that
        # group_list_dictize recurses via group_dictize (groups in groups)
        group_dictize_context['dataset_counts'] = get_group_dataset_counts()
    if context.get('with_capacity'):
        group_list = [group_dictize(group, group_dictize_context,
                                    capacity=capacity, **group_dictize_options)
                      for group, capacity in obj_list]
    else:
        group_list = [group_dictize(group, group_dictize_context,
                                    **group_dictize_options)
                      for group in obj_list]

    return sorted(group_list, key=sort_key, reverse=reverse)

def resource_list_dictize(res_list, context):

    active = context.get('active', True)
    result_list = []
    for res in res_list:
        resource_dict = resource_dictize(res, context)
        if active and res.state != 'active':
            continue

        result_list.append(resource_dict)

    return sorted(result_list, key=lambda x: x["position"])

def extras_dict_dictize(extras_dict, context):
    result_list = []
    for name, extra in extras_dict.iteritems():
        dictized = d.table_dictize(extra, context)
        if not extra.state == 'active':
            continue
        value = dictized["value"]
        result_list.append(dictized)

    return sorted(result_list, key=lambda x: x["key"])

def extras_list_dictize(extras_list, context):
    result_list = []
    active = context.get('active', True)
    for extra in extras_list:
        dictized = d.table_dictize(extra, context)
        if active and extra.state != 'active':
            continue
        value = dictized["value"]
        result_list.append(dictized)

    return sorted(result_list, key=lambda x: x["key"])


def resource_dictize(res, context):
    model = context['model']
    resource = d.table_dictize(res, context)
    extras = resource.pop("extras", None)
    if extras:
        resource.update(extras)
    # some urls do not have the protocol this adds http:// to these
    url = resource['url']
    ## for_edit is only called at the times when the dataset is to be edited
    ## in the frontend. Without for_edit the whole qualified url is returned.
    if resource.get('url_type') == 'upload' and not context.get('for_edit'):
        url = url.rsplit('/')[-1]
        cleaned_name = munge.munge_filename(url)
        resource['url'] = h.url_for('resource.download',
                                    id=resource['package_id'],
                                    resource_id=res.id,
                                    filename=cleaned_name,
                                    qualified=True)
    elif resource['url'] and not urlparse.urlsplit(url).scheme and not context.get('for_edit'):
        resource['url'] = u'http://' + url.lstrip('/')
    return resource


def _execute(q, table, context):
    '''
    Takes an SqlAlchemy query (q) that is (at its base) a Select on an
    object table (table), and it returns the object.

    Analogous with _execute_with_revision, so takes the same params, even
    though it doesn't need the table.
    '''
    model = context['model']
    session = model.Session
    return session.execute(q)


def _execute_with_revision(q, rev_table, context):
    '''
    Takes an SqlAlchemy query (q) that is (at its base) a Select on an object
    revision table (rev_table), and you provide revision_id or revision_date in
    the context and it will filter the object revision(s) to an earlier time.

    Raises NotFound if context['revision_id'] is provided, but the revision
    ID does not exist.

    Returns [] if there are no results.

    '''
    model = context['model']
    session = model.Session
    revision_id = context.get('revision_id')
    revision_date = context.get('revision_date')

    if revision_id:
        revision = session.query(context['model'].Revision).filter_by(
            id=revision_id).first()
        if not revision:
            raise logic.NotFound
        revision_date = revision.timestamp

    q = q.where(rev_table.c.revision_timestamp <= revision_date)
    q = q.where(rev_table.c.expired_timestamp > revision_date)

    return session.execute(q)


def package_dictize(pkg, context):
    '''
    Given a Package object, returns an equivalent dictionary.

    Normally this is the most recent version, but you can provide revision_id
    or revision_date in the context and it will filter to an earlier time.

    May raise NotFound if:
    * the specified revision_id doesn't exist
    * the specified revision_date was before the package was created
    '''
    model = context['model']
    is_latest_revision = not(context.get('revision_id') or
                             context.get('revision_date'))
    execute = _execute if is_latest_revision else _execute_with_revision
    #package
    if is_latest_revision:
        if isinstance(pkg, model.PackageRevision):
            pkg = model.Package.get(pkg.id)
        result = pkg
    else:
        package_rev = model.package_revision_table
        q = select([package_rev]).where(package_rev.c.id == pkg.id)
        result = execute(q, package_rev, context).first()
    if not result:
        raise logic.NotFound
    result_dict = d.table_dictize(result, context)
    #strip whitespace from title
    if result_dict.get('title'):
        result_dict['title'] = result_dict['title'].strip()

    #resources
    if is_latest_revision:
        res = model.resource_table
    else:
        res = model.resource_revision_table
    q = select([res]).where(res.c.package_id == pkg.id)
    result = execute(q, res, context)
    result_dict["resources"] = resource_list_dictize(result, context)
    result_dict['num_resources'] = len(result_dict.get('resources', []))

    #tags
    tag = model.tag_table
    if is_latest_revision:
        pkg_tag = model.package_tag_table
    else:
        pkg_tag = model.package_tag_revision_table
    q = select([tag, pkg_tag.c.state],
               from_obj=pkg_tag.join(tag, tag.c.id == pkg_tag.c.tag_id)
               ).where(pkg_tag.c.package_id == pkg.id)
    result = execute(q, pkg_tag, context)
    result_dict["tags"] = d.obj_list_dictize(result, context,
                                             lambda x: x["name"])
    result_dict['num_tags'] = len(result_dict.get('tags', []))

    # Add display_names to tags. At first a tag's display_name is just the
    # same as its name, but the display_name might get changed later (e.g.
    # translated into another language by the multilingual extension).
    for tag in result_dict['tags']:
        assert not 'display_name' in tag
        tag['display_name'] = tag['name']

    #extras - no longer revisioned, so always provide latest
    extra = model.package_extra_table
    q = select([extra]).where(extra.c.package_id == pkg.id)
    result = execute(q, extra, context)
    result_dict["extras"] = extras_list_dictize(result, context)

    #groups
    if is_latest_revision:
        member = model.member_table
    else:
        member = model.member_revision_table
    group = model.group_table
    q = select([group, member.c.capacity],
               from_obj=member.join(group, group.c.id == member.c.group_id)
               ).where(member.c.table_id == pkg.id)\
                .where(member.c.state == 'active') \
                .where(group.c.is_organization == False)
    result = execute(q, member, context)
    context['with_capacity'] = False
    ## no package counts as cannot fetch from search index at the same
    ## time as indexing to it.
    ## tags, extras and sub-groups are not included for speed
    result_dict["groups"] = group_list_dictize(result, context,
                                               with_package_counts=False)

    #owning organization
    if is_latest_revision:
        group = model.group_table
    else:
        group = model.group_revision_table
    q = select([group]
               ).where(group.c.id == pkg.owner_org) \
                .where(group.c.state == 'active')
    result = execute(q, group, context)
    organizations = d.obj_list_dictize(result, context)
    if organizations:
        result_dict["organization"] = organizations[0]
    else:
        result_dict["organization"] = None

    #relations
    if is_latest_revision:
        rel = model.package_relationship_table
    else:
        rel = model.package_relationship_revision_table
    q = select([rel]).where(rel.c.subject_package_id == pkg.id)
    result = execute(q, rel, context)
    result_dict["relationships_as_subject"] = \
        d.obj_list_dictize(result, context)
    q = select([rel]).where(rel.c.object_package_id == pkg.id)
    result = execute(q, rel, context)
    result_dict["relationships_as_object"] = \
        d.obj_list_dictize(result, context)

    # Extra properties from the domain object
    # We need an actual Package object for this, not a PackageRevision
    if isinstance(pkg, model.PackageRevision):
        pkg = model.Package.get(pkg.id)

    # isopen
    result_dict['isopen'] = pkg.isopen if isinstance(pkg.isopen, bool) \
                            else pkg.isopen()

    # type
    # if null assign the default value to make searching easier
    result_dict['type'] = pkg.type or u'dataset'

    # license
    if pkg.license and pkg.license.url:
        result_dict['license_url'] = pkg.license.url
        result_dict['license_title'] = pkg.license.title.split('::')[-1]
    elif pkg.license:
        result_dict['license_title'] = pkg.license.title
    else:
        result_dict['license_title'] = pkg.license_id

    # creation and modification date
    result_dict['metadata_modified'] = pkg.metadata_modified.isoformat()
    result_dict['metadata_created'] = pkg.metadata_created.isoformat() \
        if pkg.metadata_created else None

    return result_dict


def _get_members(context, group, member_type):

    model = context['model']
    Entity = getattr(model, member_type[:-1].capitalize())
    q = model.Session.query(Entity, model.Member.capacity).\
               join(model.Member, model.Member.table_id == Entity.id).\
               filter(model.Member.group_id == group.id).\
               filter(model.Member.state == 'active').\
               filter(model.Member.table_name == member_type[:-1])
    if member_type == 'packages':
        q = q.filter(Entity.private==False)
    if 'limits' in context and member_type in context['limits']:
        return q[:context['limits'][member_type]]
    return q.all()


def get_group_dataset_counts():
    '''For all public groups, return their dataset counts, as a SOLR facet'''
    query = search.PackageSearchQuery()
    q = {'q': '+capacity:public',
         'fl': 'groups', 'facet.field': ['groups', 'owner_org'],
         'facet.limit': -1, 'rows': 1}
    query.run(q)
    return query.facets


def group_dictize(group, context,
                  include_groups=True,
                  include_tags=True,
                  include_users=True,
                  include_extras=True,
                  packages_field='datasets',
                  **kw):
    '''
    Turns a Group object and related into a dictionary. The related objects
    like tags are included unless you specify it in the params.

    :param packages_field: determines the format of the `packages` field - can
    be `datasets`, `dataset_count` or None.
    '''
    assert packages_field in ('datasets', 'dataset_count', None)
    if packages_field == 'dataset_count':
        dataset_counts = context.get('dataset_counts', None)

    result_dict = d.table_dictize(group, context)
    result_dict.update(kw)

    result_dict['display_name'] = group.title or group.name

    if include_extras:
        result_dict['extras'] = extras_dict_dictize(
            group._extras, context)

    context['with_capacity'] = True

    if packages_field:
        def get_packages_for_this_group(group_, just_the_count=False):
            # Ask SOLR for the list of packages for this org/group
            q = {
                'facet': 'false',
                'rows': 0,
            }

            if group_.is_organization:
                q['fq'] = 'owner_org:"{0}"'.format(group_.id)
            else:
                q['fq'] = 'groups:"{0}"'.format(group_.name)

            # Allow members of organizations to see private datasets.
            if group_.is_organization:
                is_group_member = (context.get('user') and
                    authz.has_user_permission_for_group_or_org(
                        group_.id, context.get('user'), 'read'))
                if is_group_member:
                    q['include_private'] = True

            if not just_the_count:
                # package_search limits 'rows' anyway, so this is only if you
                # want even fewer
                try:
                    packages_limit = context['limits']['packages']
                except KeyError:
                    del q['rows']  # leave it to package_search to limit it
                else:
                    q['rows'] = packages_limit

            search_context = dict((k, v) for (k, v) in context.items()
                                  if k != 'schema')
            search_results = logic.get_action('package_search')(search_context,
                                                                q)
            return search_results['count'], search_results['results']

        if packages_field == 'datasets':
            package_count, packages = get_packages_for_this_group(group)
            result_dict['packages'] = packages
        else:
            if dataset_counts is None:
                package_count, packages = get_packages_for_this_group(
                    group, just_the_count=True)
            else:
                # Use the pre-calculated package_counts passed in.
                facets = dataset_counts
                if group.is_organization:
                    package_count = facets['owner_org'].get(group.id, 0)
                else:
                    package_count = facets['groups'].get(group.name, 0)

        result_dict['package_count'] = package_count

    if include_tags:
        # group tags are not creatable via the API yet, but that was(/is) a
        # future intention (see kindly's commit 5c8df894 on 2011/12/23)
        result_dict['tags'] = tag_list_dictize(
            _get_members(context, group, 'tags'),
            context)

    if include_groups:
        # these sub-groups won't have tags or extras for speed
        result_dict['groups'] = group_list_dictize(
            _get_members(context, group, 'groups'),
            context, include_groups=True)

    if include_users:
        result_dict['users'] = user_list_dictize(
            _get_members(context, group, 'users'),
            context)

    context['with_capacity'] = False

    if context.get('for_view'):
        if result_dict['is_organization']:
            plugin = plugins.IOrganizationController
        else:
            plugin = plugins.IGroupController
        for item in plugins.PluginImplementations(plugin):
            result_dict = item.before_view(result_dict)

    image_url = result_dict.get('image_url')
    result_dict['image_display_url'] = image_url
    if image_url and not image_url.startswith('http'):
        #munge here should not have an effect only doing it incase
        #of potential vulnerability of dodgy api input
        image_url = munge.munge_filename_legacy(image_url)
        result_dict['image_display_url'] = h.url_for_static(
            'uploads/group/%s' % result_dict.get('image_url'),
            qualified=True
        )
    return result_dict

def tag_list_dictize(tag_list, context):

    result_list = []
    for tag in tag_list:
        if context.get('with_capacity'):
            tag, capacity = tag
            dictized = d.table_dictize(tag, context, capacity=capacity)
        else:
            dictized = d.table_dictize(tag, context)

        # Add display_names to tag dicts. At first a tag's display_name is just
        # the same as its name, but the display_name might get changed later
        # (e.g.  translated into another language by the multilingual
        # extension).
        assert not dictized.has_key('display_name')
        dictized['display_name'] = dictized['name']

        if context.get('for_view'):
            for item in plugins.PluginImplementations(
                    plugins.ITagController):
                dictized = item.before_view(dictized)

        result_list.append(dictized)

    return result_list

def tag_dictize(tag, context, include_datasets=True):
    tag_dict = d.table_dictize(tag, context)

    if include_datasets:
        query = search.PackageSearchQuery()

        tag_query = u'+capacity:public '
        vocab_id = tag_dict.get('vocabulary_id')

        if vocab_id:
            model = context['model']
            vocab = model.Vocabulary.get(vocab_id)
            tag_query += u'+vocab_{0}:"{1}"'.format(vocab.name, tag.name)
        else:
            tag_query += u'+tags:"{0}"'.format(tag.name)

        q = {'q': tag_query, 'fl': 'data_dict', 'wt': 'json', 'rows': 1000}

        package_dicts = [h.json.loads(result['data_dict'])
                         for result in query.run(q)['results']]

    # Add display_names to tags. At first a tag's display_name is just the
    # same as its name, but the display_name might get changed later (e.g.
    # translated into another language by the multilingual extension).
    assert 'display_name' not in tag_dict
    tag_dict['display_name'] = tag_dict['name']

    if context.get('for_view'):
        for item in plugins.PluginImplementations(plugins.ITagController):
            tag_dict = item.before_view(tag_dict)

        if include_datasets:
            tag_dict['packages'] = []
            for package_dict in package_dicts:
                for item in plugins.PluginImplementations(plugins.IPackageController):
                    package_dict = item.before_view(package_dict)
                tag_dict['packages'].append(package_dict)
    else:
        if include_datasets:
            tag_dict['packages'] = package_dicts

    return tag_dict

def user_list_dictize(obj_list, context,
                      sort_key=lambda x:x['name'], reverse=False):

    result_list = []

    for obj in obj_list:
        user_dict = user_dictize(obj, context)
        user_dict.pop('reset_key', None)
        user_dict.pop('apikey', None)
        user_dict.pop('email', None)
        result_list.append(user_dict)
    return sorted(result_list, key=sort_key, reverse=reverse)

def member_dictize(member, context):
    return d.table_dictize(member, context)

def user_dictize(user, context, include_password_hash=False):

    if context.get('with_capacity'):
        user, capacity = user
        result_dict = d.table_dictize(user, context, capacity=capacity)
    else:
        result_dict = d.table_dictize(user, context)

    password_hash = result_dict.pop('password')
    del result_dict['reset_key']

    result_dict['display_name'] = user.display_name
    result_dict['email_hash'] = user.email_hash
    result_dict['number_of_edits'] = user.number_of_edits()
    result_dict['number_created_packages'] = user.number_created_packages(
        include_private_and_draft=context.get(
            'count_private_and_draft_datasets', False))

    requester = context.get('user')

    reset_key = result_dict.pop('reset_key', None)
    apikey = result_dict.pop('apikey', None)
    email = result_dict.pop('email', None)

    if context.get('keep_email', False):
        result_dict['email'] = email

    if context.get('keep_apikey', False):
        result_dict['apikey'] = apikey

    if requester == user.name:
        result_dict['apikey'] = apikey
        result_dict['email'] = email

    if authz.is_sysadmin(requester):
        result_dict['apikey'] = apikey
        result_dict['email'] = email

        if include_password_hash:
            result_dict['password_hash'] = password_hash

    model = context['model']
    session = model.Session

    return result_dict

def task_status_dictize(task_status, context):
    return d.table_dictize(task_status, context)

## conversion to api

def group_to_api(group, context):
    api_version = context.get('api_version')
    assert api_version, 'No api_version supplied in context'
    dictized = group_dictize(group, context)
    dictized["extras"] = dict((extra["key"], extra["value"])
                              for extra in dictized["extras"])
    if api_version == 1:
        dictized["packages"] = sorted([pkg["name"] for pkg in dictized["packages"]])
    else:
        dictized["packages"] = sorted([pkg["id"] for pkg in dictized["packages"]])
    return dictized

def tag_to_api(tag, context):
    api_version = context.get('api_version')
    assert api_version, 'No api_version supplied in context'
    dictized = tag_dictize(tag, context)
    if api_version == 1:
        return sorted([package["name"] for package in dictized["packages"]])
    else:
        return sorted([package["id"] for package in dictized["packages"]])


def resource_dict_to_api(res_dict, package_id, context):
    res_dict.pop("revision_id")
    res_dict.pop("state")
    res_dict["package_id"] = package_id


def package_to_api(pkg, context):
    api_version = context.get('api_version')
    assert api_version, 'No api_version supplied in context'
    dictized = package_dictize(pkg, context)

    dictized["tags"] = [tag["name"] for tag in dictized["tags"] \
                        if not tag.get('vocabulary_id')]
    dictized["extras"] = dict((extra["key"], extra["value"])
                              for extra in dictized["extras"])
    dictized['license'] = pkg.license.title if pkg.license else None
    dictized['ratings_average'] = pkg.get_average_rating()
    dictized['ratings_count'] = len(pkg.ratings)
    dictized['notes_rendered'] = h.render_markdown(pkg.notes)

    site_url = config.get('ckan.site_url', None)
    if site_url:
        dictized['ckan_url'] = '%s/dataset/%s' % (site_url, pkg.name)

    for resource in dictized["resources"]:
        resource_dict_to_api(resource, pkg.id, context)

    def make_api_1(package_id):
        return pkg.get(package_id).name

    def make_api_2(package_id):
        return package_id

    if api_version == 1:
        api_fn = make_api_1
        dictized["groups"] = [group["name"] for group in dictized["groups"]]
        # FIXME why is this just for version 1?
        if pkg.resources:
            dictized['download_url'] = pkg.resources[0].url
    else:
        api_fn = make_api_2
        dictized["groups"] = [group["id"] for group in dictized["groups"]]

    subjects = dictized.pop("relationships_as_subject")
    objects = dictized.pop("relationships_as_object")

    relationships = []
    for rel in objects:
        model = context['model']
        swap_types = model.PackageRelationship.forward_to_reverse_type
        type = swap_types(rel['type'])
        relationships.append({'subject': api_fn(rel['object_package_id']),
                              'type': type,
                              'object': api_fn(rel['subject_package_id']),
                              'comment': rel["comment"]})
    for rel in subjects:
        relationships.append({'subject': api_fn(rel['subject_package_id']),
                              'type': rel['type'],
                              'object': api_fn(rel['object_package_id']),
                              'comment': rel["comment"]})

    dictized['relationships'] = relationships

    return dictized

def vocabulary_dictize(vocabulary, context, include_datasets=False):
    vocabulary_dict = d.table_dictize(vocabulary, context)
    assert not vocabulary_dict.has_key('tags')

    vocabulary_dict['tags'] = [tag_dictize(tag, context, include_datasets)
                               for tag in vocabulary.tags]
    return vocabulary_dict

def vocabulary_list_dictize(vocabulary_list, context):
    return [vocabulary_dictize(vocabulary, context)
            for vocabulary in vocabulary_list]

def activity_dictize(activity, context):
    activity_dict = d.table_dictize(activity, context)
    return activity_dict

def activity_list_dictize(activity_list, context):
    return [activity_dictize(activity, context) for activity in activity_list]

def activity_detail_dictize(activity_detail, context):
    return d.table_dictize(activity_detail, context)

def activity_detail_list_dictize(activity_detail_list, context):
    return [activity_detail_dictize(activity_detail, context)
            for activity_detail in activity_detail_list]


def package_to_api1(pkg, context):
    # DEPRICIATED set api_version in context and use package_to_api()
    context['api_version'] = 1
    return package_to_api(pkg, context)

def package_to_api2(pkg, context):
    # DEPRICIATED set api_version in context and use package_to_api()
    context['api_version'] = 2
    return package_to_api(pkg, context)

def group_to_api1(group, context):
    # DEPRICIATED set api_version in context and use group_to_api()
    context['api_version'] = 1
    return group_to_api(group, context)

def group_to_api2(group, context):
    # DEPRICIATED set api_version in context and use group_to_api()
    context['api_version'] = 2
    return group_to_api(group, context)

def tag_to_api1(tag, context):
    # DEPRICIATED set api_version in context and use tag_to_api()
    context['api_version'] = 1
    return tag_to_api(tag, context)

def tag_to_api2(tag, context):
    # DEPRICIATED set api_version in context and use tag_to_api()
    context['api_version'] = 2
    return tag_to_api(tag, context)

def user_following_user_dictize(follower, context):
    return d.table_dictize(follower, context)

def user_following_dataset_dictize(follower, context):
    return d.table_dictize(follower, context)

def user_following_group_dictize(follower, context):
    return d.table_dictize(follower, context)

def resource_view_dictize(resource_view, context):
    dictized = d.table_dictize(resource_view, context)
    dictized.pop('order')
    config = dictized.pop('config', {})
    dictized.update(config)
    resource = context['model'].Resource.get(resource_view.resource_id)
    package_id = resource.package_id
    dictized['package_id'] = package_id
    return dictized

def resource_view_list_dictize(resource_views, context):
    resource_view_dicts = []
    for view in resource_views:
        resource_view_dicts.append(resource_view_dictize(view, context))
    return resource_view_dicts