def categorize(options, test=False):
from ckanext.dgu.lib.theme import categorize_package2, PRIMARY_THEME
stats = StatsList()
stats.report_value_limit = 1000
if options.dataset:
pkg = model.Package.get(options.dataset)
assert pkg
packages = [pkg]
else:
if test:
theme = True
else:
theme = False
packages = get_packages(publisher=options.publisher,
theme=theme,
uncategorized=options.uncategorized,
limit=options.limit)
themes_to_write = {} # pkg_name:themes
for pkg in packages:
print 'Dataset: %s' % pkg.name
themes = categorize_package2(pkg, stats)
if options.write and not pkg.extras.get(PRIMARY_THEME) and themes:
themes_to_write[pkg.name] = themes
print 'Categorize summary:'
print stats.report()
if options.write:
write_themes(themes_to_write)
def recategorize(options):
from ckanext.dgu.lib.theme import (categorize_package2, PRIMARY_THEME,
SECONDARY_THEMES, Themes)
stats = StatsList()
stats.report_value_limit = 1000
if options.dataset:
pkg = model.Package.get(options.dataset)
assert pkg
packages = [pkg]
else:
packages = get_packages(publisher=options.publisher,
theme=None,
uncategorized=options.uncategorized,
limit=options.limit)
# process the list of themes we are interested in setting on packages
themes = Themes.instance()
if options.theme:
theme_filter = set(options.theme.split(','))
for theme in theme_filter:
assert theme in themes.data, '"%s" not in %r' % (
theme, themes.data.keys())
else:
theme_filter = themes.data
themes_to_write = {} # pkg_name:themes
for pkg in packages:
print 'Dataset: %s' % pkg.name
themes = categorize_package2(pkg)
existing_theme = pkg.extras.get(PRIMARY_THEME)
pkg_identity = '%s (%s)' % (pkg.name, existing_theme)
if not themes:
print stats.add('Cannot decide theme', pkg_identity)
continue
if themes[0]['name'] not in theme_filter:
print stats.add('Not interested in theme', pkg_identity)
continue
if existing_theme == themes[0]['name']:
print stats.add('Theme unchanged %s' % themes[0]['name'],
pkg_identity)
continue
print stats.add('Recategorized to %s' % themes[0]['name'],
pkg_identity)
if options.write:
themes_to_write[pkg.name] = themes
print 'Recategorize summary:'
print stats.report()
if options.write:
write_themes(themes_to_write)
def suggest_themes(context, data_dict):
'''Suggests themes for a dataset or the component parts of a dataset
To be able to determine the primary and secondary theme, the description
tags and title are required for a Package. The categorize_package function
requires works with Package models and a dictionary, so both versions are
supported. If an id is passed, then the package will be retrieved and passed
to the categorisation, otherwise it will be formatted as per the required
dictionary.
'''
from ckanext.dgu.lib.theme import categorize_package2
themes = []
# TODO: Make this only available to logged in publishers
model = context['model']
id = data_dict.get('id')
if id:
pkg = model.Package.get(id)
themes = categorize_package2(pkg)
else:
pkg_dict = {
'name': data_dict.get('name'),
'title': data_dict.get('title'),
'notes': data_dict.get('notes'),
'tags': [t for t in data_dict.get('tags', '').split(',')],
'extras': [{
'key': '',
'value': ''
}]
}
themes = categorize_package2(pkg_dict)
results = {'primary-theme': {}, 'secondary-theme': []}
if len(themes) >= 1:
results['primary-theme'] = themes[0]
results['secondary-theme'] = themes[1:]
return results
def test_basic(self):
themes = categorize_package2(fish_pkg)
assert_equal(type(themes), list)
theme = themes[0]
assert_equal(type(theme), dict)
assert_equal(theme["name"], "Environment")
# be lenient as the algorithm may change
assert theme["score"] > 3, theme.get("score")
assert theme["reasons"], theme.get("reasons")
assert_equal(
[u'"fish" matched title', u'"river" matched title', u'"fish" matched description'], theme["reasons"]
)
def test_basic(self):
themes = categorize_package2(fish_pkg)
assert_equal(type(themes), list)
theme = themes[0]
assert_equal(type(theme), dict)
assert_equal(theme['name'], 'Environment')
# be lenient as the algorithm may change
assert theme['score'] > 3, theme.get('score')
assert theme['reasons'], theme.get('reasons')
assert_equal([u'"fish" matched title',
u'"river" matched title',
u'"fish" matched description'],
theme['reasons'])
def suggest_themes(context, data_dict):
'''Suggests themes for a dataset or the component parts of a dataset
To be able to determine the primary and secondary theme, the description
tags and title are required for a Package. The categorize_package function
requires works with Package models and a dictionary, so both versions are
supported. If an id is passed, then the package will be retrieved and passed
to the categorisation, otherwise it will be formatted as per the required
dictionary.
'''
from ckanext.dgu.lib.theme import categorize_package2
themes = []
# TODO: Make this only available to logged in publishers
model = context['model']
id = data_dict.get('id')
if id:
pkg = model.Package.get(id)
themes = categorize_package2(pkg)
else:
pkg_dict = {'name': data_dict.get('name'),
'title': data_dict.get('title'),
'notes': data_dict.get('notes'),
'tags': [t for t in data_dict.get('tags', '').split(',')],
'extras': [{'key': '', 'value': ''}]
}
themes = categorize_package2(pkg_dict)
results = {'primary-theme': {}, 'secondary-theme': []}
if len(themes) >= 1:
results['primary-theme'] = themes[0]
results['secondary-theme'] = themes[1:]
return results
def test_with_secondary_theme(self):
themes = categorize_package2(fish_and_spend_pkg)
assert_equal(type(themes), list)
theme = themes[0]
assert_equal(type(theme), dict)
assert_equal(theme["name"], "Environment")
# be lenient as the algorithm may change
assert theme["score"] > 3, theme.get("score")
assert theme["reasons"], theme.get("reasons")
assert_equal(
[u'"fish" matched title', u'"river" matched title', u'"fish" matched description'], theme["reasons"]
)
theme = themes[1]
assert_equal(theme["name"], "Government Spending")
assert_equal(
[u'"spend" matched title', u'"transact" matched description', u'"spend" matched description'],
theme["reasons"],
)
def test_with_secondary_theme(self):
themes = categorize_package2(fish_and_spend_pkg)
assert_equal(type(themes), list)
theme = themes[0]
assert_equal(type(theme), dict)
assert_equal(theme['name'], 'Environment')
# be lenient as the algorithm may change
assert theme['score'] > 3, theme.get('score')
assert theme['reasons'], theme.get('reasons')
assert_equal([u'"fish" matched title',
u'"river" matched title',
u'"fish" matched description'],
theme['reasons'])
theme = themes[1]
assert_equal(theme['name'], 'Government Spending')
assert_equal([u'"spend" matched title',
u'"transact" matched description',
u'"spend" matched description'],
theme['reasons'])
def test_stem_exception(self):
themes = categorize_package2(employer_pkg)
theme_names = [theme["name"] for theme in themes]
assert_equal(set(("Business & Economy",)), set(theme_names))
def test_topic_in_two_categories(self):
themes = categorize_package2(death_pkg)
theme_names = [theme["name"] for theme in themes]
assert_equal(set(("Society", "Health")), set(theme_names))
def test_topic_in_two_categories(self):
themes = categorize_package2(death_pkg)
theme_names = [theme['name'] for theme in themes]
assert_equal(set(('Society', 'Health')), set(theme_names))
def test_stem_exception(self):
themes = categorize_package2(employer_pkg)
theme_names = [theme['name'] for theme in themes]
assert_equal(set(('Business & Economy',)), set(theme_names))
