def generate_index_entry(args):
index = {}
# Generate title
country_names = None
if args[2] is not None:
country_names = [c.name_en for c in args[2]]
title = get_title(
api_name=args[0],
app_name=args[1],
country_names=country_names,
trade_flow=args[3],
years=args[4],
product_name=args[5].name_en if args[5] is not None else None
)
index["title"] = title
# Generate url
country_codes = None
if args[2] is not None:
country_codes = [c.name_3char.lower() for c in args[2]]
url = params_to_url(
api_name=args[0],
app_name=args[1][0],
country_codes=country_codes,
trade_flow=args[3],
years=args[4],
product_code=args[5].code if args[5] is not None else None
)
index["url"] = url
regions = None
if args[2] is not None:
regions = [c.region.name for c in args[2]]
# Add in params into elasticsearch in case we need them later
kwargs = dict(
api_name=args[0],
app_name=args[1],
country_names=country_names,
country_codes=country_codes,
regions=regions,
trade_flow=args[3],
years=args[4],
product_name=args[5].name_en if args[5] is not None else None,
product_code=args[5].code if args[5] is not None else None,
product_community=args[5].community.name if args[5] is not None else None,
)
kwargs = {k: v for k, v in kwargs.iteritems() if v is not None}
index.update(kwargs)
return index
def generate_index_entry(args):
index = {}
# Generate title
country_names = None
if args[2] is not None:
country_names = [c.name_en for c in args[2]]
title = get_title(
api_name=args[0],
app_name=args[1],
country_names=country_names,
trade_flow=args[3],
years=args[4],
product_name=args[5].name_en if args[5] is not None else None)
index["title"] = title
# Generate url
country_codes = None
if args[2] is not None:
country_codes = [c.name_3char.lower() for c in args[2]]
url = params_to_url(
api_name=args[0],
app_name=args[1][0],
country_codes=country_codes,
trade_flow=args[3],
years=args[4],
product_code=args[5].code if args[5] is not None else None)
index["url"] = url
regions = None
if args[2] is not None:
regions = [c.region.name for c in args[2]]
# Add in params into elasticsearch in case we need them later
kwargs = dict(
api_name=args[0],
app_name=args[1],
country_names=country_names,
country_codes=country_codes,
regions=regions,
trade_flow=args[3],
years=args[4],
product_name=args[5].name_en if args[5] is not None else None,
product_code=args[5].code if args[5] is not None else None,
product_community=args[5].community.name
if args[5] is not None else None,
)
kwargs = {k: v for k, v in kwargs.iteritems() if v is not None}
index.update(kwargs)
return index
def api_search(request):
query = request.GET.get("term", None)
if query is None:
return HttpResponse("[]")
# For user experiment, run search version 1 or 2, 2 being more feature
# rich and having parsed filters. See atlas-data#32
search_version = int(request.GET.get("search_var", 0))
# Parse search query
query, query_type, kwargs = parse_search(
query, strip_keywords=(search_version != 1))
# Resolve any synonyms. feasibility -> pie_scatter etc.
if "app_name" in kwargs:
given_app_name = kwargs["app_name"][0]
kwargs["app_name"] = [
APP_NAME_SYNONYMS.get(given_app_name, given_app_name)
]
# Viz params are not an elasticsearch filter so pop that off
viz_params = kwargs.pop("viz_params", None)
# Prepare elasticsearch filters
if search_version == 2 or search_version == 0:
filters = prepare_filters(kwargs)
else:
filters = {}
es_query = {"query": {"filtered": {}}, "size": 8}
# Add filters to the query if they were given. Filters are ANDed.
if len(filters) > 0:
es_filters = [{
"terms": {
k: [x.lower() for x in v]
}
} for k, v in filters.iteritems()]
es_filters = {"bool": {"must": es_filters}}
es_query["query"]["filtered"]["filter"] = es_filters
# Add fuzzy search for query string if any non-filter query string remains
# after taking out the filters
if query.strip() != "":
es_query["query"]["filtered"]["query"] = {
"fuzzy_like_this": {
"like_text": query,
"fields": ["title"],
"max_query_terms": 15,
"prefix_length": 3
}
}
# Do the query
es = Elasticsearch()
result = es.search(index="questions", body=es_query)
# Format the results in a way that complies with the OpenSearch standard's
# suggestion extension
labels = []
urls = []
for x in result['hits']['hits']:
data = x['_source']
# Regenerate title and url so we can add stuff into it dynamically,
# like the year being searched for, or forcing an app.
years = kwargs.get('years', None)
# Possible apps this title could be visualized as
app_names = data['app_name']
# If the app the user requested is possible, use that. Otherwise, use
# the first one as default. App names in the elasticsearch index are
# sorted in a certain way for this to make sense so check out the
# indexer script
requested_app_name = filters.get("app_name", [None])[0]
if requested_app_name in app_names:
app_name = requested_app_name
else:
app_name = app_names[0]
if years and len(years) == 2:
if app_name in ["map", "tree_map"]:
# If multiple years are specified and we can do a stacked
# graph, do a stacked graph instead of a treemap or map
app_name = "stacked"
elif app_name in ["product_space", "pie_scatter"]:
# Some apps can never have multiple years so just use the first
# one specified
years = [years[0]]
# If no years specified, use default years
if years is None:
if app_name == "stacked":
years = [settings.YEAR_MIN_HS4, settings.YEAR_MAX_HS4]
else:
years = [settings.YEAR_MAX_HS4]
# You can't show a product space based on imports so ignore those
if app_name == "product_space" and data["trade_flow"] == "import":
continue
title = get_title(api_name=data['api_name'],
app_name=app_name,
country_names=data.get('country_names', None),
trade_flow=data['trade_flow'],
years=years,
product_name=data.get('product_name', None))
url = params_to_url(api_name=data['api_name'],
app_name=app_name,
country_codes=data.get('country_codes', None),
trade_flow=data['trade_flow'],
years=years,
product_code=data.get('product_code', None))
if viz_params:
if app_name == "pie_scatter":
url += "?queryActivated=True"
url += "&yaxis=%s" % viz_params[0]
labels.append(title)
urls.append(settings.HTTP_HOST + url)
return HttpResponse(json.dumps([query, labels, [], urls]))
def api_search(request):
query = request.GET.get("term", None)
if query is None:
return HttpResponse("[]")
# For user experiment, run search version 1 or 2, 2 being more feature
# rich and having parsed filters. See atlas-data#32
search_version = int(request.GET.get("search_var", 0))
# Parse search query
query, query_type, kwargs = parse_search(
query,
strip_keywords=(search_version != 1))
# Resolve any synonyms. feasibility -> pie_scatter etc.
if "app_name" in kwargs:
given_app_name = kwargs["app_name"][0]
kwargs["app_name"] = [APP_NAME_SYNONYMS.get(given_app_name,
given_app_name)]
# Viz params are not an elasticsearch filter so pop that off
viz_params = kwargs.pop("viz_params", None)
# Prepare elasticsearch filters
if search_version == 2 or search_version == 0:
filters = prepare_filters(kwargs)
else:
filters = {}
es_query = {
"query": {
"filtered": {}
},
"size": 8
}
# Add filters to the query if they were given. Filters are ANDed.
if len(filters) > 0:
es_filters = [{"terms": {k: [x.lower() for x in v]}}
for k, v in filters.iteritems()]
es_filters = {"bool": {"must": es_filters}}
es_query["query"]["filtered"]["filter"] = es_filters
# Add fuzzy search for query string if any non-filter query string remains
# after taking out the filters
if query.strip() != "":
es_query["query"]["filtered"]["query"] = {
"fuzzy_like_this": {
"like_text": query,
"fields": ["title"],
"max_query_terms": 15,
"prefix_length": 3
}
}
# Do the query
es = Elasticsearch()
result = es.search(index="questions", body=es_query)
# Format the results in a way that complies with the OpenSearch standard's
# suggestion extension
labels = []
urls = []
for x in result['hits']['hits']:
data = x['_source']
# Regenerate title and url so we can add stuff into it dynamically,
# like the year being searched for, or forcing an app.
years = kwargs.get('years', None)
# Possible apps this title could be visualized as
app_names = data['app_name']
# If the app the user requested is possible, use that. Otherwise, use
# the first one as default. App names in the elasticsearch index are
# sorted in a certain way for this to make sense so check out the
# indexer script
requested_app_name = filters.get("app_name", [None])[0]
if requested_app_name in app_names:
app_name = requested_app_name
else:
app_name = app_names[0]
if years and len(years) == 2:
if app_name in ["map", "tree_map"]:
# If multiple years are specified and we can do a stacked
# graph, do a stacked graph instead of a treemap or map
app_name = "stacked"
elif app_name in ["product_space", "pie_scatter"]:
# Some apps can never have multiple years so just use the first
# one specified
years = [years[0]]
# If no years specified, use default years
if years is None:
if app_name == "stacked":
years = [settings.YEAR_MIN_HS4, settings.YEAR_MAX_HS4]
else:
years = [settings.YEAR_MAX_HS4]
# You can't show a product space based on imports so ignore those
if app_name == "product_space" and data["trade_flow"] == "import":
continue
title = get_title(
api_name=data['api_name'],
app_name=app_name,
country_names=data.get('country_names', None),
trade_flow=data['trade_flow'],
years=years,
product_name=data.get('product_name', None)
)
url = params_to_url(
api_name=data['api_name'],
app_name=app_name,
country_codes=data.get('country_codes', None),
trade_flow=data['trade_flow'],
years=years,
product_code=data.get('product_code', None)
)
if viz_params:
if app_name == "pie_scatter":
url += "?queryActivated=True"
url += "&yaxis=%s" % viz_params[0]
labels.append(title)
urls.append(settings.HTTP_HOST + url)
return HttpResponse(json.dumps([
query,
labels,
[],
urls
]))
