def get(self, request):
"""
A simple starting implementation for search.
TODO:
- accept more parameters
- scoring on multiple factors
- decide on what id to use (MapItem, or [type]:[id], etc.)
- and much more...
"""
def _serialize(hit):
"""
TODO: Ideally elasticsearch will handle this itself
"""
return {
"id": hit.id,
"_meta": {
"doc_type": hit.meta.doc_type,
},
"locations": [
{
"latitude": loc['point']['lat'],
"longitude": loc['point']['lon'],
}
for loc in hit.locations
],
}
params = request.GET.dict()
esq = es_search()
q = params.pop("q", None)
if q:
esq = esq.query("simple_query_string", query=q, fields=["_all"])
# TODO: filter out unnecessary fields
# esq = esq.fields()
hits = [_serialize(m) for m in esq.execute().hits]
return self.success({
"results": hits
})
def get(self, request):
"""
A simple starting implementation for search.
TODO:
- accept more parameters
- scoring on multiple factors
- decide on what id to use (MapItem, or [type]:[id], etc.)
- and much more...
"""
def _serialize(hit):
"""
TODO: Ideally elasticsearch will handle this itself
"""
return {
"id":
hit.id,
"_meta": {
"doc_type": hit.meta.doc_type,
},
"locations": [{
"latitude": loc['point']['lat'],
"longitude": loc['point']['lon'],
} for loc in hit.locations],
}
params = request.GET.dict()
esq = es_search()
q = params.pop("q", None)
if q:
esq = esq.query("simple_query_string", query=q, fields=["_all"])
# TODO: filter out unnecessary fields
# esq = esq.fields()
hits = [_serialize(m) for m in esq.execute().hits]
return self.success({"results": hits})
def get(self, request):
"""
TODO:
- use intelligent ES query to facet across categories instead
of grouping with Python
"""
groups = defaultdict(list)
results = []
params = request.GET.dict()
esq = es_search()
q = params.pop("q", None)
nearest = params.pop("nearest", None)
radius = params.pop("radius", "100km")
if q:
esq = esq.query("simple_query_string", query=q, fields=["_all"])
if nearest:
lat, lon = map(float, nearest.split(','))
# more weighting is given to closer things
# based on https://www.elastic.co/guide/en/elasticsearch/guide/current/decay-functions.html
esq = esq.query('function_score', functions=[
{
"gauss": {
"locations.point": {
"origin": { "lat": lat, "lon": lon },
"offset": "0",
"scale": radius # something at this radius gets
# 50% score of something at the origin
}
},
"weight": 1
}
])
# collects all the remaining params as metadata filters
# TODO: not all metadata should be filterable (some may be private/admin)
for key, value in params.items():
esq = esq.filter("term", **{ "metadata.{}".format(key) : value })
print('p', key, value)
# faceting on doc type
esq.aggs.bucket("doc_type", "terms", field="_type")
# TODO: also facet on metadata values...
response = esq.execute()
hits = response.hits
for hit in hits:
d = hit.to_dict()
d["_meta"] = { k: hit.meta.to_dict()[k] for k in ('score', 'doc_type')}
groups[hit.meta['doc_type']].append(d)
for category, group in groups.items():
results.append({
"doc_type": category, # TODO: use category name/slug/id
"results": group,
"total": len(group) # TODO: get real total via faceting
})
return self.success({
"result_groups": results,
"aggregations": response.aggregations.to_dict()
# these are useful if you want to see the raw query/results
#"_query": esq.to_dict(),
#"_result": response.to_dict()
})
def get(self, request):
"""
TODO:
- use intelligent ES query to facet across categories instead
of grouping with Python
"""
groups = defaultdict(list)
results = []
params = request.GET.dict()
esq = es_search()
q = params.pop("q", None)
nearest = params.pop("nearest", None)
radius = params.pop("radius", "100km")
if q:
esq = esq.query("simple_query_string", query=q, fields=["_all"])
if nearest:
lat, lon = map(float, nearest.split(','))
# more weighting is given to closer things
# based on https://www.elastic.co/guide/en/elasticsearch/guide/current/decay-functions.html
esq = esq.query(
'function_score',
functions=[{
"gauss": {
"locations.point": {
"origin": {
"lat": lat,
"lon": lon
},
"offset": "0",
"scale": radius # something at this radius gets
# 50% score of something at the origin
}
},
"weight": 1
}])
# collects all the remaining params as metadata filters
# TODO: not all metadata should be filterable (some may be private/admin)
for key, value in params.items():
esq = esq.filter("term", **{"metadata.{}".format(key): value})
print('p', key, value)
# faceting on doc type
esq.aggs.bucket("doc_type", "terms", field="_type")
# TODO: also facet on metadata values...
response = esq.execute()
hits = response.hits
for hit in hits:
d = hit.to_dict()
d["_meta"] = {
k: hit.meta.to_dict()[k]
for k in ('score', 'doc_type')
}
groups[hit.meta['doc_type']].append(d)
for category, group in groups.items():
results.append({
"doc_type": category, # TODO: use category name/slug/id
"results": group,
"total": len(group) # TODO: get real total via faceting
})
return self.success({
"result_groups": results,
"aggregations": response.aggregations.to_dict()
# these are useful if you want to see the raw query/results
#"_query": esq.to_dict(),
#"_result": response.to_dict()
})
