def urls_add(self):
add_url_list = list(set(self.__request_json['urls_add'])) # 去重
# 已存在queue中的
exist_queue_url_list = []
res = Mongo.get()['queue_' + self.project_name].find({'url_md5': {'$in': [md5(l) for l in add_url_list]}},
{'url': 1})
for doc in res:
exist_queue_url_list.append(doc['url'])
# 已存在parsed中的
exist_parsed_url_list = []
res = Mongo.get()['parsed_' + self.project_name].find({'url_md5': {'$in': [md5(l) for l in add_url_list]}},
{'url': 1})
for doc in res: # todo 需要判断存在的时间, 允许重复抓取
exist_parsed_url_list.append(doc['url'])
# 加入队列
add_urls_data = []
for url in add_url_list:
if url not in exist_queue_url_list and url not in exist_parsed_url_list: # 不存在queue不存在parsed中才加入队列
add_urls_data.append(
{'domain': get_domain(url),
'url': url,
'url_md5': md5(url),
'flag_time': 0,
'add_time': int(time.time()),
'slave_ip': self.__request_address[0]})
add_urls_data and Mongo.get()['queue_' + self.project_name].insert(add_urls_data)
def toggle_project(project_id):
try:
project = Mongo.get()["projects"].find_one({"_id": ObjectId(project_id)})
if not project:
raise Exception("不存在的记录!")
except:
return jsonify({"success": False, "msg": "不存在的记录!"})
project["static"] = "抓取中" if project["static"] == "暂停中" else "暂停中"
Mongo.get()["projects"].update({"_id": ObjectId(project_id)}, {"$set": {"static": project["static"]}})
return jsonify({"success": True, "msg": "切换成功!"})
def get_projects():
project_dict = {}
for project in get_project_list():
project_dict[project['name']] = {
'_id': str(project['_id']),
'name': project['name'],
'static': project['static'],
'queue_len': Mongo.get()['queue_' + project['name']].count(),
'parsed_len': Mongo.get()['parsed_' + project['name']].count(),
'result_len': Mongo.get()['result_' + project['name']].count(),
}
return jsonify(project_dict)
def get_projects():
project_dict = {}
for project in get_project_list():
project_dict[project["name"]] = {
"_id": str(project["_id"]),
"name": project["name"],
"static": project["static"],
"queue_len": Mongo.get()["queue_" + project["name"]].count(),
"parsed_len": Mongo.get()["parsed_" + project["name"]].count(),
"result_len": Mongo.get()["result_" + project["name"]].count(),
}
return jsonify(project_dict)
def toggle_project(project_id):
try:
project = Mongo.get()['projects'].find_one({'_id': ObjectId(project_id)})
if not project:
raise Exception('不存在的记录!')
except:
return jsonify({'success': False, 'msg': '不存在的记录!'})
project['static'] = '抓取中' if project['static'] == '暂停中' else '暂停中'
Mongo.get()['projects'].update({'_id': ObjectId(project_id)},
{'$set': {'static': project['static']}})
return jsonify({'success': True, 'msg': '切换成功!'})
def get_projects():
project_dict = {}
for project in get_project_list():
project_dict[project['name']] = {
'_id': str(project['_id']),
'name': project['name'],
'static': project['static'],
'queue_len': Mongo.get()['queue_' + project['name']].count(),
'parsed_len': Mongo.get()['parsed_' + project['name']].count(),
'result_len': Mongo.get()['result_' + project['name']].count(),
}
return jsonify(project_dict)
def get_project_tasks(project_name):
res = []
for doc in Mongo.get()['parsed_' + project_name].find().sort('_id', -1).limit(100):
del doc['_id']
doc['slave_ip'] = mix_ip(doc['slave_ip'])
res.append(doc)
return json.dumps(res)
def toggle_project(project_id):
try:
project = Mongo.get()['projects'].find_one(
{'_id': ObjectId(project_id)})
if not project:
raise Exception('不存在的记录!')
except:
return jsonify({'success': False, 'msg': '不存在的记录!'})
project['static'] = '抓取中' if project['static'] == '暂停中' else '暂停中'
Mongo.get()['projects'].update({'_id': ObjectId(project_id)},
{'$set': {
'static': project['static']
}})
return jsonify({'success': True, 'msg': '切换成功!'})
def get_project_tasks(project_name):
res = []
for doc in Mongo.get()["parsed_" + project_name].find().sort("_id", -1).limit(100):
del doc["_id"]
doc["slave_ip"] = mix_ip(doc["slave_ip"])
res.append(doc)
return json.dumps(res)
def save_project():
return jsonify({'success': False, 'msg': '禁止修改!'})
form_data = json.loads(request.data) # todo 需要验证表单数据
name_r = re.compile(r'^[0-9a-zA-Z_-]+$')
if not name_r.search(form_data['name']):
return jsonify({'success': False, 'msg': '计划名称必须满足正则规则: ^[0-9a-zA-Z_-]+$ '})
exists_project = list(Mongo.get()['projects'].find({'name': form_data['name']}, {'_id': 1, 'add_time': 1}).limit(1))
if 'edit' not in form_data and exists_project:
return jsonify({'success': False, 'msg': '计划名称已经存在!'})
# 新增计划或更新计划
data = {
'name': form_data['name'],
'init_url': form_data['init_url'],
'desc': form_data['desc'] if 'desc' in form_data else '',
'code': form_data['code'],
'static': '暂停中',
'update_time': int(time.time()),
'add_time': exists_project[0]['add_time'] if exists_project else int(time.time()),
}
Mongo.get()['projects'].update({'name': form_data['name']}, data, True)
# 当是新计划时的初始化
if 'edit' not in form_data:
Mongo.get()['queue_' + form_data['name']].insert(
{
'url': form_data['init_url'],
'url_md5': md5(form_data['init_url']),
'flag_time': 0,
'add_time': int(time.time()),
'slave_ip': '0.0.0.0'
})
# 在没创建集合前设置索引mongodb会自动创建该集合并赋索引
Mongo.get()['parsed_' + form_data['name']].ensure_index('url_md5', unique=True)
Mongo.get()['queue_' + form_data['name']].ensure_index('url_md5', unique=True)
# 有新计划加入, 重启全部slave
restart_slave_list = GlobalHelper.get('restart_slave_list') or []
for slave_record in Mongo.get()['slave_record'].find():
restart_slave_list.append(slave_record['ip'])
GlobalHelper.set('restart_slave_list', list(set(restart_slave_list)))
return jsonify({'success': True, 'msg': '保存成功!'})
def save_project():
form_data = json.loads(request.data) # todo 需要验证表单数据
name_r = re.compile(r"^[0-9a-zA-Z_-]+$")
if not name_r.search(form_data["name"]):
return jsonify({"success": False, "msg": "计划名称必须满足正则规则: ^[0-9a-zA-Z_-]+$ "})
exists_project = list(Mongo.get()["projects"].find({"name": form_data["name"]}, {"_id": 1, "add_time": 1}).limit(1))
if "edit" not in form_data and exists_project:
return jsonify({"success": False, "msg": "计划名称已经存在!"})
# 新增计划或更新计划
data = {
"name": form_data["name"],
"init_url": form_data["init_url"],
"desc": form_data["desc"] if "desc" in form_data else "",
"code": form_data["code"],
"static": "暂停中",
"update_time": int(time.time()),
"add_time": exists_project[0]["add_time"] if exists_project else int(time.time()),
}
Mongo.get()["projects"].update({"name": form_data["name"]}, data, True)
# 当是新计划时的初始化
if "edit" not in form_data:
Mongo.get()["queue_" + form_data["name"]].insert(
{
"url": form_data["init_url"],
"url_md5": md5(form_data["init_url"]),
"flag_time": 0,
"add_time": int(time.time()),
"slave_ip": "0.0.0.0",
}
)
# 在没创建集合前设置索引mongodb会自动创建该集合并赋索引
Mongo.get()["parsed_" + form_data["name"]].ensure_index("url_md5", unique=True)
Mongo.get()["queue_" + form_data["name"]].ensure_index("url_md5", unique=True)
# 有新计划加入, 重启全部slave
restart_slave_list = GlobalHelper.get("restart_slave_list") or []
for slave_record in Mongo.get()["slave_record"].find():
restart_slave_list.append(slave_record["ip"])
GlobalHelper.set("restart_slave_list", list(set(restart_slave_list)))
return jsonify({"success": True, "msg": "保存成功!"})
def __storage_record(self):
for ip, data in self.slave_record.items():
res = Mongo.get().slave_record.update(
{'ip': ip},
{'ip': ip, 'data': data}, True) # 有着更新, 无则插入
if not res['updatedExisting'] and 'upserted' in res: # 插入时
self.slave_record[ip]['_id'] = str(res['upserted'])
def get_project_tasks(project_name):
res = []
for doc in Mongo.get()['parsed_' + project_name].find().sort(
'_id', -1).limit(100):
del doc['_id']
doc['slave_ip'] = mix_ip(doc['slave_ip'])
res.append(doc)
return json.dumps(res)
def __init__(self, project_name, request_json, request_address):
self.project_name = project_name
temp = list(
Mongo.get()['projects'].find({'name': self.project_name}).limit(1))
self.project = temp[0] if temp else {}
self.__request_json = request_json
self.__request_address = request_address
self.__slave_record = SlaveRecord.get_instance()
def get_project_list():
"""
todo 测试是否支持中文
:return:
"""
# get_project_list.cache = [v for v in Mongo.get().projects.find({}, {'_id': 0})]
get_project_list.cache = [v for v in Mongo.get().projects.find()]
return get_project_list.cache
def get_json_results(project_name):
res = make_response(
json.dumps(
list(Mongo.get()['result_' + project_name].find({}, {'_id': 0}))),
200)
# header("Content-Disposition:attachment;filename={$fileName}");
res.headers['Content-Disposition'] = 'attachment'
res.headers['filename'] = project_name + '/json'
return res
def get_slave_tasks(slave_id):
res = []
try:
slave_record = Mongo.get()['slave_record'].find_one({'_id': ObjectId(slave_id)})
if not slave_record:
raise Exception('不存在的记录!')
except:
return json.dumps(res)
for project in get_project_list():
for doc in Mongo.get()['parsed_' + project['name']].find({'slave_ip': slave_record['ip']}).sort('_id',
-1).limit(20):
del doc['_id']
doc['slave_ip'] = mix_ip(doc['slave_ip'])
res.append(doc)
res.sort(key=lambda x: x['add_time'], reverse=True)
return json.dumps(res)
def toggle_slave(slave_id):
try:
slave_record = Mongo.get()['slave_record'].find_one({'_id': ObjectId(slave_id)})
if not slave_record:
raise Exception('不存在的记录!')
except:
return jsonify({'success': False, 'msg': '不存在的记录!'})
slave_record['data']['static'] = '抓取中' if slave_record['data']['static'] == '暂停中' else '暂停中'
try:
Mongo.get()['slave_record'].update({'_id': ObjectId(slave_id)},
{'$set': {'data.static': slave_record['data']['static']}})
global_salve_record = GlobalHelper.get('salve_record')
global_salve_record[slave_record['ip']]['static'] = slave_record['data']['static']
GlobalHelper.set('salve_record', global_salve_record)
except:
print traceback.format_exc()
return jsonify({'success': True, 'msg': '切换成功!'})
def get_results(project_name, page=1):
res = []
pagination = paginate(Mongo.get()['result_' + project_name].find().sort('_id', -1), page, 20)
if pagination:
for doc in pagination.result():
del doc['_id']
res.append(doc)
return json.dumps({'result': res, 'render_json': pagination.render_json(10)})
def get_results(project_name, page=1):
res = []
pagination = paginate(Mongo.get()["result_" + project_name].find().sort("_id", -1), page, 20)
if pagination:
for doc in pagination.result():
del doc["_id"]
res.append(doc)
return json.dumps({"result": res, "render_json": pagination.render_json(10)})
def __storage_record(self):
for ip, data in self.slave_record.items():
res = Mongo.get().slave_record.update({'ip': ip}, {
'ip': ip,
'data': data
}, True) # 有着更新, 无则插入
if not res['updatedExisting'] and 'upserted' in res: # 插入时
self.slave_record[ip]['_id'] = str(res['upserted'])
def code_ctrl(self):
result = []
for project in Mongo.get().projects.find({}, {
'_id': 0,
'code': 1,
'name': 1,
'init_url': 1
}):
result.append(project)
return result
def toggle_slave(slave_id):
try:
slave_record = Mongo.get()["slave_record"].find_one({"_id": ObjectId(slave_id)})
if not slave_record:
raise Exception("不存在的记录!")
except:
return jsonify({"success": False, "msg": "不存在的记录!"})
slave_record["data"]["static"] = "抓取中" if slave_record["data"]["static"] == "暂停中" else "暂停中"
try:
Mongo.get()["slave_record"].update(
{"_id": ObjectId(slave_id)}, {"$set": {"data.static": slave_record["data"]["static"]}}
)
global_salve_record = GlobalHelper.get("salve_record")
global_salve_record[slave_record["ip"]]["static"] = slave_record["data"]["static"]
GlobalHelper.set("salve_record", global_salve_record)
except:
print traceback.format_exc()
return jsonify({"success": True, "msg": "切换成功!"})
def get_slave_tasks(slave_id):
res = []
try:
slave_record = Mongo.get()["slave_record"].find_one({"_id": ObjectId(slave_id)})
if not slave_record:
raise Exception("不存在的记录!")
except:
return json.dumps(res)
for project in get_project_list():
for doc in (
Mongo.get()["parsed_" + project["name"]].find({"slave_ip": slave_record["ip"]}).sort("_id", -1).limit(20)
):
del doc["_id"]
doc["slave_ip"] = mix_ip(doc["slave_ip"])
res.append(doc)
res.sort(key=lambda x: x["add_time"], reverse=True)
return json.dumps(res)
def get_urls(self):
# SerHandle.__init_project(self.project_name)
response_url_list = []
ids = []
# print self.__slave_record.slave_record[self.__request_address[0]]['deny_domains']
# todo need to test
deny_domains = [x['domain'] for x in
self.__slave_record.slave_record[self.__request_address[0]]['deny_domains']]
for doc in Mongo.get()['queue_' + self.project_name].find(
{'domain': {'$nin': deny_domains}, 'flag_time': {'$lt': int(time.time() - 300)}}).limit(3) \
.sort('_id', pymongo.ASCENDING): # 取标识时间早于当前时间300秒之前的url
ids.append(doc['_id'])
response_url_list.append(doc['url'])
ids and Mongo.get()['queue_' + self.project_name].update({'_id': {'$in': ids}},
{'$set': {'flag_time': int(time.time())}},
multi=True)
return response_url_list
def get_slave_tasks(slave_id):
res = []
try:
slave_record = Mongo.get()['slave_record'].find_one(
{'_id': ObjectId(slave_id)})
if not slave_record:
raise Exception('不存在的记录!')
except:
return json.dumps(res)
for project in get_project_list():
for doc in Mongo.get()['parsed_' + project['name']].find({
'slave_ip':
slave_record['ip']
}).sort('_id', -1).limit(20):
del doc['_id']
doc['slave_ip'] = mix_ip(doc['slave_ip'])
res.append(doc)
res.sort(key=lambda x: x['add_time'], reverse=True)
return json.dumps(res)
def get_results(project_name, page=1):
res = []
pagination = paginate(
Mongo.get()['result_' + project_name].find().sort('_id', -1), page, 20)
if pagination:
for doc in pagination.result():
del doc['_id']
res.append(doc)
return json.dumps({
'result': res,
'render_json': pagination.render_json(10)
})
def restart_slave(slave_id):
try:
slave_record = Mongo.get()["slave_record"].find_one({"_id": ObjectId(slave_id)})
if not slave_record:
raise Exception("不存在的记录!")
except:
return jsonify({"success": False, "msg": "不存在的记录!"})
restart_slave_list = GlobalHelper.get("restart_slave_list") or []
restart_slave_list.append(slave_record["ip"])
GlobalHelper.set("restart_slave_list", list(set(restart_slave_list)))
return jsonify({"success": True, "msg": "重启中!"})
def restart_slave(slave_id):
try:
slave_record = Mongo.get()['slave_record'].find_one({'_id': ObjectId(slave_id)})
if not slave_record:
raise Exception('不存在的记录!')
except:
return jsonify({'success': False, 'msg': '不存在的记录!'})
restart_slave_list = GlobalHelper.get('restart_slave_list') or []
restart_slave_list.append(slave_record['ip'])
GlobalHelper.set('restart_slave_list', list(set(restart_slave_list)))
return jsonify({'success': True, 'msg': '重启中!'})
def urls_parsed(self):
urls_data = []
url_list = []
for url in self.__request_json['urls_parsed']:
self.__slave_record.add_parsed_record(self.__request_address[0])
url_list.append(url)
urls_data.append(
{'url': url, 'url_md5': md5(url), 'add_time': int(time.time()), 'slave_ip': self.__request_address[0]})
Mongo.get()['queue_' + self.project_name].remove({'url_md5': {'$in': [md5(l) for l in url_list]}},
multi=True) # 删除抓取完毕的队列
try:
urls_data and Mongo.get()['parsed_' + self.project_name].insert(urls_data)
except:
try:
for single_url in urls_data:
single_url and Mongo.get()['parsed_' + self.project_name].insert_one(single_url)
except Exception, error:
print traceback.format_exc()
print error
print u'下面链接重复抓取的并重复保存到parsed_*中的记录'
print single_url, '\r\n\r\n'
def restart_slave(slave_id):
try:
slave_record = Mongo.get()['slave_record'].find_one(
{'_id': ObjectId(slave_id)})
if not slave_record:
raise Exception('不存在的记录!')
except:
return jsonify({'success': False, 'msg': '不存在的记录!'})
restart_slave_list = GlobalHelper.get('restart_slave_list') or []
restart_slave_list.append(slave_record['ip'])
GlobalHelper.set('restart_slave_list', list(set(restart_slave_list)))
return jsonify({'success': True, 'msg': '重启中!'})
def toggle_slave(slave_id):
try:
slave_record = Mongo.get()['slave_record'].find_one(
{'_id': ObjectId(slave_id)})
if not slave_record:
raise Exception('不存在的记录!')
except:
return jsonify({'success': False, 'msg': '不存在的记录!'})
slave_record['data'][
'static'] = '抓取中' if slave_record['data']['static'] == '暂停中' else '暂停中'
try:
Mongo.get()['slave_record'].update(
{'_id': ObjectId(slave_id)},
{'$set': {
'data.static': slave_record['data']['static']
}})
global_salve_record = GlobalHelper.get('salve_record')
global_salve_record[
slave_record['ip']]['static'] = slave_record['data']['static']
GlobalHelper.set('salve_record', global_salve_record)
except:
print traceback.format_exc()
return jsonify({'success': True, 'msg': '切换成功!'})
def api_test():
try:
client = MongoClient('localhost', 27017)
res = client['pyfetch']['result_cnbeta'].find().limit(10)
# print GlobalHelper.get('salve_record')
# time.sleep(10)
pagination = paginate(Mongo.get()['result_cnbeta'].find(), 1, 30)
if pagination:
for row in pagination.current_page():
print row
# print pagination.next()
# print pagination.prev()
except:
print traceback.format_exc()
return jsonify({'fd': 1})
def api_test():
try:
client = MongoClient('localhost', 27017)
res = client['pyfetch']['result_cnbeta'].find().limit(10)
# print GlobalHelper.get('salve_record')
# time.sleep(10)
pagination = paginate(Mongo.get()['result_cnbeta'].find(), 1, 30)
if pagination:
for row in pagination.current_page():
print row
# print pagination.next()
# print pagination.prev()
except:
print traceback.format_exc()
return jsonify({'fd': 1})
def __init__(self):
self.__init_format = {
'_id': '',
'ip': '',
'parsed_count': 0,
'connected_count': 0,
'last_connected_time': 0,
'work_time_count': 1,
'deny_domains': [],
'error_domains': {},
'static': '抓取中'}
self.slave_record = {}
self.deny_urls_temp = {}
if not self.slave_record:
for item in Mongo.get().slave_record.find():
item['data']['_id'] = str(item['_id'])
item['data']['ip'] = str(item['ip'])
self.slave_record[item['ip']] = item['data']
self.refresh_connect_status()
def __init__(self):
self.__init_format = {
'_id': '',
'ip': '',
'parsed_count': 0,
'connected_count': 0,
'last_connected_time': 0,
'work_time_count': 1,
'deny_domains': [],
'error_domains': {},
'static': '抓取中'
}
self.slave_record = {}
self.deny_urls_temp = {}
if not self.slave_record:
for item in Mongo.get().slave_record.find():
item['data']['_id'] = str(item['_id'])
item['data']['ip'] = str(item['ip'])
self.slave_record[item['ip']] = item['data']
self.refresh_connect_status()
def code_ctrl(self):
result = []
for project in Mongo.get().projects.find({}, {'_id': 0, 'code': 1, 'name': 1, 'init_url': 1}):
result.append(project)
return result
def get_json_results(project_name):
res = make_response(json.dumps(list(Mongo.get()['result_' + project_name].find({}, {'_id': 0}))), 200)
# header("Content-Disposition:attachment;filename={$fileName}");
res.headers['Content-Disposition'] = 'attachment'
res.headers['filename'] = project_name + '/json'
return res
def get_project_by_name(name):
res = list(Mongo.get().projects.find({"name": name}, {"_id": 0}))
if not res:
return jsonify({})
return jsonify(res[0])
def get_json_results(project_name):
res = make_response(json.dumps(list(Mongo.get()["result_" + project_name].find({}, {"_id": 0}))), 200)
# header("Content-Disposition:attachment;filename={$fileName}");
res.headers["Content-Disposition"] = "attachment"
res.headers["filename"] = project_name + "/json"
return res
def get_project_by_name(name):
res = list(Mongo.get().projects.find({'name': name}, {'_id': 0}))
if not res:
return jsonify({})
return jsonify(res[0])
def result_save(self):
Mongo.get()['result_' + self.project_name].insert(self.__request_json['save'])
def save_project():
form_data = json.loads(request.data) # todo 需要验证表单数据
name_r = re.compile(r'^[0-9a-zA-Z_-]+$')
if not name_r.search(form_data['name']):
return jsonify({
'success': False,
'msg': '计划名称必须满足正则规则: ^[0-9a-zA-Z_-]+$ '
})
exists_project = list(Mongo.get()['projects'].find(
{
'name': form_data['name']
}, {
'_id': 1,
'add_time': 1
}).limit(1))
if 'edit' not in form_data and exists_project:
return jsonify({'success': False, 'msg': '计划名称已经存在!'})
# 新增计划或更新计划
data = {
'name':
form_data['name'],
'init_url':
form_data['init_url'],
'desc':
form_data['desc'] if 'desc' in form_data else '',
'code':
form_data['code'],
'static':
'暂停中',
'update_time':
int(time.time()),
'add_time':
exists_project[0]['add_time'] if exists_project else int(time.time()),
}
Mongo.get()['projects'].update({'name': form_data['name']}, data, True)
# 当是新计划时的初始化
if 'edit' not in form_data:
Mongo.get()['queue_' + form_data['name']].insert({
'url':
form_data['init_url'],
'url_md5':
md5(form_data['init_url']),
'flag_time':
0,
'add_time':
int(time.time()),
'slave_ip':
'0.0.0.0'
})
# 在没创建集合前设置索引mongodb会自动创建该集合并赋索引
Mongo.get()['parsed_' + form_data['name']].ensure_index('url_md5',
unique=True)
Mongo.get()['queue_' + form_data['name']].ensure_index('url_md5',
unique=True)
# 有新计划加入, 重启全部slave
restart_slave_list = GlobalHelper.get('restart_slave_list') or []
for slave_record in Mongo.get()['slave_record'].find():
restart_slave_list.append(slave_record['ip'])
GlobalHelper.set('restart_slave_list', list(set(restart_slave_list)))
return jsonify({'success': True, 'msg': '保存成功!'})
def get_project_by_name(name):
res = list(Mongo.get().projects.find({'name': name}, {'_id': 0}))
if not res:
return jsonify({})
return jsonify(res[0])