def get_graph():
"""
get explicit params from url
接受url的Param作为root nodeId, 然后返回一张dag
dag似乎可以提前就计算好? 因为要有children node的相关度排序
计算好之后就直接提出来用? 在app载入的时候就加到内存里面去
考虑可拓展性? 如果图特别大呢, 那显然和设计的架构不一致,
因为构想是前端处理整个graph的裁剪, 后端提供整张图
如果图很大, 那么后端必须裁剪好前端直接用, 不能指望前端那点计算能力
因为不止一个点, <del>但是现在没做多关键词查询</del>, 如果做了那么就是一个列表
<del>params里面传入的应该是一个serialized json string, 然后解析为json然后得到数组</del>
使用;分隔, 就像search那样
"""
# 然而现在需要考虑同时查询两个
wids = request.args.get('wids', '').lower()
print('graph params', wids)
# params = GraphParams(request.args)
if not wids:
response = Response(ResponseType.PARAMETERS_ERR)
return response.get_json()
try:
# 由于概念格的特点, 子节点关键词集合一定包含父节点的关键词集合, 文档集合反之
# 找到含有extent的所有点, 然后把data里面无关的连线删掉, 找到fah==[]就是新root
wids = convert_int(wids.split(';'))
g = sub_graph(NODES, wids)
# add data in g
g = add_info(g, KEYWORDS_DICT)
if len(g) == 1:
# response = Response.get_custom_response(ResponseType.FAILURE, 'No Such sub graph')
response = Response(ResponseType.SUCCESS)
response.update_attr('graph', g)
else:
response = Response(ResponseType.SUCCESS)
response.update_attr('graph', g)
except Exception as e:
print(e)
response = Response(ResponseType.INTERNAL_ERR)
return response.get_json()
def get_courses_list():
# 返回图和返回文章的接口分离
# 因为文章还要翻页 查询文章直接在mongodb中查询
# 因为不需要做查询相关度排序, 提供排序为, 年份, 默认(字典序), 被引量
try:
course_params = CourseListParams(request.args)
# validate params passed in
if not course_params.validate():
raise ValidationError
# return courses
node_id = course_params.node_id.data
page = course_params.page.data
page_size = course_params.page_size.data
# order = course_params.order.data
node = DB['nodes'].find_one({'nid': node_id})
if not node:
raise ValueError('no such node id exists')
extents = node['extent']
courses = DB['courses'].find({'cid': {"$in": extents}}) \
.limit(page_size) \
.skip((page-1)*page_size)
# if order == 'time':
# courses = DB['courses'].find({'cid': {"$in": extents}}) \
# .sort([('date',-1)]).limit(page_size) \
# .skip((page-1)*page_size)
# elif order == 'citation':
# courses = DB['courses'].find({'cid': {"$in": extents}}) \
# .sort([('citation',-1)]).limit(page_size) \
# .skip((page-1)*page_size)
courses = [after_pop(d, '_id') for d in courses]
response = Response(ResponseType.SUCCESS)
response.update_attr('courses', courses)
response.update_attr('page', page)
response.update_attr('pageSize', page_size)
response.update_attr('totalSize', len(extents))
except ValidationError as ve:
print(ve)
response = Response(ResponseType.PARAMETERS_ERR)
except ValueError as ve:
response = Response.get_custom_response(ResponseType.FAILURE, str(ve))
except Exception as e:
print(e)
response = Response(ResponseType.INTERNAL_ERR)
return response.get_json()
def user_rating():
"""
打分
/api/user-rating
"""
try:
token = request.headers.get('Authorization', '')
# by convention jwt token is like "Bearer <token string>"
uname = verify_auth_token(token[7:])
if not uname:
raise ValueError('Invalid Token')
user = DB['users'].find_one({'username': uname})
user_rates = user.get('rates', [])
user_eval = user.get('evaluation', [])
if not user_rates:
user_rates = []
if not user_eval:
user_eval = []
response = Response(ResponseType.SUCCESS)
response.update_attr(
'rates', {str(rate['cid']): rate['rate']
for rate in user_rates})
response.update_attr(
'evaluation',
{str(rate['cid']): rate['rate']
for rate in user_eval})
except ValueError as ve:
response = Response.get_custom_response(ResponseType.FAILURE, str(ve))
except Exception as e:
print(e)
response = Response(ResponseType.INTERNAL_ERR)
return response.get_json()
def rating():
"""
打分
/api/rating?course_id=<int>&rate=<int>&is_evaluation=<bool>
"""
try:
rate_params = RatingParams(request.args)
if not rate_params.validate():
raise ValidationError
rate = rate_params.rate.data
course_id = rate_params.course_id.data
is_evalution = rate_params.is_evaluation.data
token = request.headers.get('Authorization', '')
# by convention jwt token is like "Bearer <token string>"
uname = verify_auth_token(token[7:])
if not uname:
raise ValueError('Invalid Token')
if is_evalution:
# check whether this rate exists
if not DB['users'].find_one({
'username': uname,
'evaluation.cid': course_id
}):
DB['users'].update({'username': uname}, {
'$push': {
'evaluation': {
'cid': course_id,
'rate': rate
}
}
})
else:
DB['users'].update(
{
'username': uname,
'evaluation.cid': course_id
}, {'$set': {
'evaluation.$.rate': rate
}})
else:
# check whether this rate exists
if not DB['users'].find_one({
'username': uname,
'rates.cid': course_id
}):
DB['users'].update(
{'username': uname},
{'$push': {
'rates': {
'cid': course_id,
'rate': rate
}
}})
else:
DB['users'].update({
'username': uname,
'rates.cid': course_id
}, {'$set': {
'rates.$.rate': rate
}})
response = Response(ResponseType.SUCCESS)
except ValidationError:
response = Response(ResponseType.VALIDATION_ERR)
except ValueError as ve:
response = Response.get_custom_response(ResponseType.FAILURE, str(ve))
except Exception as e:
print(e)
response = Response(ResponseType.INTERNAL_ERR)
return response.get_json()
def recommend_courses():
"""
recommend courses
/api/recommend?page=<int>&page_size=<int>
"""
try:
recommend_params = RecommendCoursesParams(request.args)
if not recommend_params.validate():
raise ValidationError
page = recommend_params.page.data
page_size = recommend_params.page_size.data
mode = recommend_params.mode.data
token = request.headers.get('Authorization', '')
# auth validation
uname = verify_auth_token(token[7:])
if not uname:
raise ValueError('Invalid Token')
# find user
user = DB['users'].find_one({'username': uname})
if not user:
raise ValueError('User Not Found')
# create user model
user_rates = user.get('rates', None)
if not user_rates:
return Response.get_custom_response(
ResponseType.FAILURE, "unable to recommend").get_json()
course_vectors = COURSE_VECTORS_DICT[str(mode)]
user_model = recommend(user_rates, course_vectors)
# store user model
DB['users'].update({'username': uname},
{'$set': {
'usermodel': user_model
}})
user_courses = [x['cid'] for x in user_model]
# change the order the full rating items
user_rates_dict = {x['cid']: x['rate'] for x in user_rates}
rate5 = [c for c in user_courses if user_rates_dict.get(c, 0) == 5]
rate4 = [c for c in user_courses if user_rates_dict.get(c, 0) == 4]
other_rates = [
c for c in user_courses if user_rates_dict.get(c, 0) < 4
]
user_courses = rate5 + rate4 + other_rates
courses = [
after_pop(DB['courses'].find_one({'cid': a}), '_id')
for a in user_courses[(page - 1) * page_size:page * page_size]
]
response = Response(ResponseType.SUCCESS)
response.update_attr('courses', courses)
response.update_attr('page', page)
response.update_attr('pageSize', page_size)
response.update_attr('totalSize', len(user_courses))
except ValidationError:
response = Response(ResponseType.VALIDATION_ERR)
except ValueError as ve:
response = Response.get_custom_response(ResponseType.FAILURE, str(ve))
except Exception as e:
print('unknown err', e)
response = Response(ResponseType.INTERNAL_ERR)
return response.get_json()
def search():
'''
得到一个或者关键词之后, 如果这个或者这些关键词命中了
返回命中和未命中的关键词, 以及命中的关键词分别的kId
目前只做一个关键词的好了
来了一个新词之后, 在所有关键词里面匹配, 精准命中就是直接返回
如果不中, 那么就计算相似度, 然后返回相似度top5的关键词
相似度计算可以使用编辑距离自动机或者是BK树, 不想写, 好麻烦
如果之后要做到实时推荐, 输入一个词出一个词, 那么就需要这样做
'''
# 从url的请求参数中获取keyword, like /api/search?keywords=xxx;xxx
keywords = request.args.get('keywords', '').lower()
# keywords为一个string, 多个keywords使用;分隔, 因为一个keyword可能是phrase
keywords = keywords.split(';')
print(keywords)
# 同时因为mongodb不会parse指定的key, 所以不存在sql注入的安全风险
mongo_col = DB['keywords'] # keywords collection
fail_list = []
matched_list = []
match = True
wids = []
recommend = []
try:
for k in keywords:
res = mongo_col.find_one({'keyword': k})
if res:
matched_list.append(res['keyword'])
wids.append(res['wid'])
else:
match = False
fail_list.append(k)
# 将匹配失败的词汇和所有的keywords进行相似度匹配
match_num = 5
if fail_list:
best_matches = sorted(
KEYWORDS,
reverse=True,
key=lambda kwobj: max(fail_list,
key=lambda fail_kw: Levenshtein.ratio(
kwobj['keyword'], fail_kw)))
recommend = [x['keyword'] for x in best_matches[:match_num]]
response = Response(ResponseType.SUCCESS)
response.update_attr('keywords', matched_list)
response.update_attr('match', match)
response.update_attr('wids', wids)
response.update_attr('recommend', recommend)
except Exception as e:
print(e)
response = Response(ResponseType.FAILURE)
return response.get_json()
# @app.route('/api/root', methods=['GET'], strict_slashes=False)
# def get_root():
# rootId = 1
# response = Response(ResponseType.SUCCESS)
# response.update_attr('rootId', rootId)
# except Exception as e:
# print(e)
# response = Response(ResponseType.FAILURE)
# return response.get_json()