,COALESCE(ROUND(T2.neirong_uv/T1.dau,4),0) AS per_content_uv
,COALESCE(ROUND(T2.neirong_pv/T2.neirong_uv,4),0) AS per_content_pv
,COALESCE(CONCAT(ROUND(T4.retention_num1/T2.neirong_uv*100,4),'%'),0) AS retention_1
,COALESCE(CONCAT(ROUND(T4.retention_num7/T2.neirong_uv*100,4),'%'),0) AS retention_7
,COALESCE(CONCAT(ROUND(T4.retention_num30/T2.neirong_uv*100,4),'%'),0) AS retention_30
,COALESCE(T5.app_duration,0) AS avg_app_duration
,COALESCE(T3.neirong_stay,0) AS avg_content_stay
,COALESCE(T5.avg_opentimes,0) AS avg_open_times
,COALESCE(T9.search_stay,0) AS search_related_stay
,COALESCE(T9.welfare_stay,0) AS welfare_stay
,COALESCE(T9.question_stay,0) AS content_question_stay
,COALESCE(T9.ai_related_stay,0) AS ai_related_stay
,COALESCE(T9.diary_stay,0) AS content_diary_stay
,COALESCE(T9.home_stay,0) AS home_stay
,COALESCE(T9.conv_stay,0) AS conv_related_stay
,COALESCE(ROUND(T6.recommend_uv/T2.neirong_uv,4),0) AS recommend_rate
,COALESCE(ROUND(T6.recommend_pv/T6.recommend_uv,4),0) AS per_recommend_pv
,COALESCE(ROUND(T6.feeds_uv/T2.neirong_uv,4),0) AS feeds_rate
,COALESCE(ROUND(T6.feeds_pv/T6.feeds_uv,4),0) AS per_feeds_pv
,COALESCE(ROUND(T6.search_uv/T2.neirong_uv,4),0) AS search_rate
,COALESCE(ROUND(T6.search_pv/T6.search_uv,4),0) AS per_search_pv
,COALESCE(ROUND(T6.zone_uv/T2.neirong_uv,4),0) AS zone_rate
,COALESCE(ROUND(T6.zone_pv/T6.zone_uv,4),0) AS per_zone_pv
,COALESCE(ROUND(T6.content_uv/T2.neirong_uv,4),0) AS content_rate
,COALESCE(ROUND(T6.content_pv/T6.content_uv,4),0) AS per_from_content_pv
,COALESCE(ROUND(T6.blank_uv/T2.neirong_uv,4),0) AS blank_rate
,COALESCE(ROUND(T6.blank_pv/T6.blank_uv,4),0) AS per_blank_pv
,COALESCE(ROUND(T6.comment_uv/T2.neirong_uv,4),0) AS comment_rate
,COALESCE(ROUND(T6.comment_pv/T6.comment_uv,4),0) AS per_comment_pv
,COALESCE(ROUND(T6.org_uv/T2.neirong_uv,4),0) AS org_rate
,COALESCE(ROUND(T6.org_pv/T6.org_uv,4),0) AS per_org_pv
,COALESCE(ROUND(T6.category_uv/T2.neirong_uv,4),0) AS category_rate
,COALESCE(ROUND(T6.category_pv/T6.category_uv,4),0) AS per_category_pv
,COALESCE(ROUND(T6.my_diary_uv/T2.neirong_uv,4),0) AS my_diary_rate
,COALESCE(ROUND(T6.my_diary_pv/T6.my_diary_uv,4),0) AS per_my_diary_pv
,COALESCE(ROUND(T6.ai_uv/T2.neirong_uv,4),0) AS ai_rate
,COALESCE(ROUND(T6.ai_pv/T6.ai_uv,4),0) AS per_ai_pv
,null AS create_topic_num
,null AS create_reply_num
,COALESCE(T2.diary_uv,0) AS diary_uv
,COALESCE(T2.diary_pv,0) AS diary_pv
,COALESCE(ROUND(T2.diary_uv/T2.neirong_uv,4),0) AS diary_rate
,COALESCE(ROUND(T2.diary_pv/T2.diary_uv,4),0) AS per_diary_pv
,COALESCE(T3.diary_stay,0) AS diary_stay
,COALESCE(T2.post_uv,0) AS post_uv
,COALESCE(T2.post_pv,0) AS post_pv
,COALESCE(ROUND(T2.post_uv/T2.neirong_uv,4),0) AS post_rate
,COALESCE(ROUND(T2.post_pv/T2.post_uv,4),0) AS per_post_pv
,COALESCE(T3.post_stay,0) AS post_stay
,COALESCE(T2.question_uv,0) AS question_uv
,COALESCE(T2.question_pv,0) AS question_pv
,COALESCE(ROUND(T2.question_uv/T2.neirong_uv,4),0) AS question_rate
,COALESCE(ROUND(T2.question_pv/T2.question_uv,4),0) AS per_question_pv
,COALESCE(T3.question_stay,0) AS question_stay
,COALESCE(T2.question_answer_uv,0) AS question_answer_uv
,COALESCE(T2.question_answer_pv,0) AS question_answer_pv
,COALESCE(ROUND(T2.question_answer_uv/T2.neirong_uv,4),0) AS question_answer_rate
,COALESCE(ROUND(T2.question_answer_pv/T2.question_answer_uv,4),0) AS per_question_answer_pv
,COALESCE(T3.question_answer_stay,0) AS question_answer_stay
,COALESCE(T2.answer_uv,0) AS answer_uv
,COALESCE(T2.answer_pv,0) AS answer_pv
,COALESCE(ROUND(T2.answer_uv/T2.neirong_uv,4),0) AS answer_rate
,COALESCE(ROUND(T2.answer_pv/T2.answer_uv,4),0) AS per_answer_pv
,COALESCE(T3.answer_stay,0) AS answer_stay
,COALESCE(T2.video_uv,0) AS video_uv
,COALESCE(T2.video_pv,0) AS video_pv
,COALESCE(ROUND(T2.video_uv/T2.neirong_uv,4),0) AS video_rate
,COALESCE(ROUND(T2.video_pv/T2.video_uv,4),0) AS per_video_pv
,COALESCE(T3.video_stay,0) AS video_stay
,COALESCE(T2.wiki_uv,0) AS wiki_uv
,COALESCE(T2.wiki_pv,0) AS wiki_pv
,COALESCE(ROUND(T2.wiki_uv/T2.neirong_uv,4),0) AS wiki_rate
,COALESCE(ROUND(T2.wiki_pv/T2.wiki_uv,4),0) AS per_wiki_pv
,COALESCE(T3.wiki_stay,0) AS wiki_stay
,COALESCE(T2.article_uv,0) AS article_uv
,COALESCE(T2.article_pv,0) AS article_pv
,COALESCE(ROUND(T2.article_uv/T2.neirong_uv,4),0) AS article_rate
,COALESCE(ROUND(T2.article_pv/T2.article_uv,4),0) AS per_article_pv
,COALESCE(T3.article_stay,0) AS article_stay
FROM
(--基础维度/dau
SELECT partition_date,device_os_type,active_type,t2.channel,count(distinct device_id) AS dau
FROM
(
SELECT partition_date
,m.device_id, device_os_type
,case WHEN active_type = '4' THEN '老活跃设备'
WHEN active_type in ('1','2') then '新增设备' END as active_type
,array(CASE WHEN first_channel_source_type like '%xinyouxingkong%' or a.device_id is not NULL THEN '可疑'
WHEN (partition_date>='20190601' and tmp.col2 = 'AI')
or (partition_date < '20200301' AND partition_date>='20190601' and first_channel_source_type like 'promotion_toutiao_jy%')
or (partition_date>='20200601' and ((first_channel_source_type like 'promotion_toutiao_jy%') or (first_channel_source_type like 'dyand%') or (first_channel_source_type like 'douyin%')))
THEN 'AI' ELSE '其他' END , '合计') as channel
FROM online.ml_device_day_active_status m
LEFT JOIN
(SELECT col1,col2 --col1:子渠道，col2：是否属于AI，col3:标识
FROM pm.tl_pm_ydl
WHERE col3='0204_danlei_channel')tmp
on first_channel_source_type=tmp.col1
LEFT JOIN
(SELECT DISTINCT device_id
FROM al.al_pm_ct_dv_deviceappversionrollbackfrom20200101_d
WHERE partition_day = regexp_replace(DATE_SUB(current_date,2) ,'-',''))a
ON m.device_id = a.device_id
where partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date<regexp_replace((current_date),'-','')
AND active_type in ('1','2','4')
and first_channel_source_type not in ('yqxiu1','yqxiu2','yqxiu3','yqxiu4','yqxiu5','mxyc1','mxyc2','mxyc3'
WHEN active_type in ('1','2') then '新增设备' END as active_type
,array(CASE WHEN first_channel_source_type like '%xinyouxingkong%' or a.device_id is not NULL THEN '可疑'
WHEN (partition_date>='20190601' and tmp.col2 = 'AI')
or (partition_date < '20200301' AND partition_date>='20190601' and first_channel_source_type like 'promotion_toutiao_jy%')
or (partition_date>='20200601' and ((first_channel_source_type like 'promotion_toutiao_jy%') or (first_channel_source_type like 'dyand%') or (first_channel_source_type like 'douyin%')))
THEN 'AI' ELSE '其他' END , '合计') as channel
FROM online.ml_device_day_active_status m
LEFT JOIN
(SELECT col1,col2 --col1:子渠道，col2：是否属于AI，col3:标识
FROM pm.tl_pm_ydl
WHERE col3='0204_danlei_channel')tmp
on first_channel_source_type=tmp.col1
LEFT JOIN
(SELECT DISTINCT device_id
FROM al.al_pm_ct_dv_deviceappversionrollbackfrom20200101_d
WHERE partition_day = regexp_replace(DATE_SUB(current_date,2) ,'-',''))a
ON m.device_id = a.device_id
where partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date<regexp_replace((current_date),'-','')
AND active_type in ('1','2','4')
and first_channel_source_type not in ('yqxiu1','yqxiu2','yqxiu3','yqxiu4','yqxiu5','mxyc1','mxyc2','mxyc3'
AND first_channel_source_type not like 'promotion\_jf\_%'
)t1
JOIN
(
SELECT partition_date,page_name,cl_id
FROM online.bl_hdfs_maidian_updates
WHERE partition_date >= regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date < regexp_replace((current_date),'-','')
AND action = 'page_view'
AND page_name IN ('diary_detail','topic_detail','post_detail','user_post_detail','doctor_post_detail','question_detail','answer_detail','question_answer_detail',
WHEN active_type in ('1','2') then '新增设备' END as active_type
,array(CASE WHEN first_channel_source_type like '%xinyouxingkong%' or a.device_id is not NULL THEN '可疑'
WHEN (partition_date>='20190601' and tmp.col2 = 'AI')
or (partition_date < '20200301' AND partition_date>='20190601' and first_channel_source_type like 'promotion_toutiao_jy%')
or (partition_date>='20200601' and ((first_channel_source_type like 'promotion_toutiao_jy%') or (first_channel_source_type like 'dyand%') or (first_channel_source_type like 'douyin%')))
THEN 'AI' ELSE '其他' END , '合计') as channel
FROM online.ml_device_day_active_status m
LEFT JOIN
(SELECT col1,col2 --col1:子渠道，col2：是否属于AI，col3:标识
FROM pm.tl_pm_ydl
WHERE col3='0204_danlei_channel')tmp
on first_channel_source_type=tmp.col1
LEFT JOIN
(SELECT DISTINCT device_id
FROM al.al_pm_ct_dv_deviceappversionrollbackfrom20200101_d
WHERE partition_day = regexp_replace(DATE_SUB(current_date,2) ,'-',''))a
ON m.device_id = a.device_id
where partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date<regexp_replace((current_date),'-','')
AND active_type in ('1','2','4')
and first_channel_source_type not in ('yqxiu1','yqxiu2','yqxiu3','yqxiu4','yqxiu5','mxyc1','mxyc2','mxyc3'
AND first_channel_source_type not like 'promotion\_jf\_%'
)t1
JOIN
(
SELECT partition_date,page_name,cl_id,page_stay
FROM online.bl_hdfs_maidian_updates
WHERE partition_date >= regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date < regexp_replace((current_date),'-','')
AND action = 'page_view'
AND page_name IN ('diary_detail','topic_detail','post_detail','user_post_detail','doctor_post_detail','question_detail','answer_detail','question_answer_detail',
where partition_day=regexp_replace(DATE_SUB(current_date,1) ,'-','')
)spam_pv
on t2.cl_id=spam_pv.device_id
WHERE spam_pv.device_id IS NULL
)t3
LATERAL VIEW explode(t3.channel) t4 AS channel
GROUP BY partition_date,device_os_type,active_type,t4.channel
)T3
ON T1.partition_date=T3.partition_date
AND T1.device_os_type=T3.device_os_type
AND T1.active_type=T3.active_type
AND T1.channel=T3.channel
LEFT JOIN
(--内容用户留存
SELECT regexp_replace(partition_date,'-','') AS partition_date
,device_os_type,active_type,t5.channel
,int(count(DISTINCT CASE WHEN date_add(partition_date,1) = retention_date THEN device_id END)) AS retention_num1
,int(count(DISTINCT CASE WHEN date_add(partition_date,6) = retention_date THEN device_id END)) AS retention_num7
,int(count(DISTINCT CASE WHEN date_add(partition_date,29) = retention_date THEN device_id END)) AS retention_num30
FROM
(
SELECT t1.partition_date,device_os_type,active_type,channel,t3.device_id,t3.partition_date as retention_date
FROM
(
SELECT
concat_ws('-',substr(partition_date,1,4),substr(partition_date,5,2),substr(partition_date,7,2)) as partition_date
,m.device_id, device_os_type
,case WHEN active_type = '4' THEN '老活跃设备'
WHEN active_type in ('1','2') then '新增设备' END as active_type
,array(CASE WHEN first_channel_source_type like '%xinyouxingkong%' or a.device_id is not NULL THEN '可疑'
WHEN (partition_date>='20190601' and tmp.col2 = 'AI')
or (partition_date < '20200301' AND partition_date>='20190601' and first_channel_source_type like 'promotion_toutiao_jy%')
or (partition_date>='20200601' and ((first_channel_source_type like 'promotion_toutiao_jy%') or (first_channel_source_type like 'dyand%') or (first_channel_source_type like 'douyin%')))
THEN 'AI' ELSE '其他' END , '合计') as channel
FROM online.ml_device_day_active_status m
LEFT JOIN
(SELECT col1,col2 --col1:子渠道，col2：是否属于AI，col3:标识
FROM pm.tl_pm_ydl
WHERE col3='0204_danlei_channel')tmp
on first_channel_source_type=tmp.col1
LEFT JOIN
(SELECT DISTINCT device_id
FROM al.al_pm_ct_dv_deviceappversionrollbackfrom20200101_d
WHERE partition_day = regexp_replace(DATE_SUB(current_date,2) ,'-',''))a
ON m.device_id = a.device_id
where partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date<regexp_replace((current_date),'-','')
AND active_type in ('1','2','4')
and first_channel_source_type not in ('yqxiu1','yqxiu2','yqxiu3','yqxiu4','yqxiu5','mxyc1','mxyc2','mxyc3'
AND first_channel_source_type not like 'promotion\_jf\_%'
)t1
JOIN
(--内容用户
SELECT cl_id
,concat_ws('-',substr(partition_date,1,4),substr(partition_date,5,2),substr(partition_date,7,2)) AS partition_date
FROM online.bl_hdfs_maidian_updates
WHERE partition_date >= regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date < regexp_replace((current_date),'-','')
AND action = 'page_view'
AND page_name IN ('diary_detail','topic_detail','post_detail','user_post_detail','doctor_post_detail','question_detail','answer_detail','question_answer_detail',
WHEN active_type in ('1','2') then '新增设备' END as active_type
,array(CASE WHEN first_channel_source_type like '%xinyouxingkong%' or a.device_id is not NULL THEN '可疑'
WHEN (partition_date>='20190601' and tmp.col2 = 'AI')
or (partition_date < '20200301' AND partition_date>='20190601' and first_channel_source_type like 'promotion_toutiao_jy%')
or (partition_date>='20200601' and ((first_channel_source_type like 'promotion_toutiao_jy%') or (first_channel_source_type like 'dyand%') or (first_channel_source_type like 'douyin%')))
THEN 'AI' ELSE '其他' END , '合计') as channel
FROM online.ml_device_day_active_status m
LEFT JOIN
(SELECT col1,col2 --col1:子渠道，col2：是否属于AI，col3:标识
FROM pm.tl_pm_ydl
WHERE col3='0204_danlei_channel')tmp
on first_channel_source_type=tmp.col1
LEFT JOIN
(SELECT DISTINCT device_id
FROM al.al_pm_ct_dv_deviceappversionrollbackfrom20200101_d
WHERE partition_day = regexp_replace(DATE_SUB(current_date,2) ,'-',''))a
ON m.device_id = a.device_id
where partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date<regexp_replace((current_date),'-','')
AND active_type in ('1','2','4')
and first_channel_source_type not in ('yqxiu1','yqxiu2','yqxiu3','yqxiu4','yqxiu5','mxyc1','mxyc2','mxyc3'
AND first_channel_source_type not like 'promotion\_jf\_%'
)t1
JOIN
(--内容用户
SELECT partition_date,cl_id
FROM online.bl_hdfs_maidian_updates
WHERE partition_date >= regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date < regexp_replace((current_date),'-','')
AND action = 'page_view'
AND page_name IN ('diary_detail','topic_detail','post_detail','user_post_detail','doctor_post_detail','question_detail','answer_detail','question_answer_detail',
WHEN active_type in ('1','2') then '新增设备' END as active_type
,array(CASE WHEN first_channel_source_type like '%xinyouxingkong%' or a.device_id is not NULL THEN '可疑'
WHEN (partition_date>='20190601' and tmp.col2 = 'AI')
or (partition_date < '20200301' AND partition_date>='20190601' and first_channel_source_type like 'promotion_toutiao_jy%')
or (partition_date>='20200601' and ((first_channel_source_type like 'promotion_toutiao_jy%') or (first_channel_source_type like 'dyand%') or (first_channel_source_type like 'douyin%')))
THEN 'AI' ELSE '其他' END , '合计') as channel
FROM online.ml_device_day_active_status m
LEFT JOIN
(SELECT col1,col2 --col1:子渠道，col2：是否属于AI，col3:标识
FROM pm.tl_pm_ydl
WHERE col3='0204_danlei_channel')tmp
on first_channel_source_type=tmp.col1
LEFT JOIN
(SELECT DISTINCT device_id
FROM al.al_pm_ct_dv_deviceappversionrollbackfrom20200101_d
WHERE partition_day = regexp_replace(DATE_SUB(current_date,2) ,'-',''))a
ON m.device_id = a.device_id
where partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date<regexp_replace((current_date),'-','')
AND active_type in ('1','2','4')
and first_channel_source_type not in ('yqxiu1','yqxiu2','yqxiu3','yqxiu4','yqxiu5','mxyc1','mxyc2','mxyc3'
AND first_channel_source_type not like 'promotion\_jf\_%'
)t1
LEFT JOIN
( --不同页面进入内容
SELECT partition_date,cl_id
,case when referrer like 'search_result%' then 'search'
when referrer ='' then 'blank'
when referrer = 'zone_v3' then 'zone_v3'
when referrer='all_case_service_comment' then 'comment'
when referrer in ('organization_detail','expert_detail') then 'org'
when referrer = 'category' then 'category'
when referrer ='my_diary' then 'my_diary'
when referrer in ('face_detect_result','report_result') then 'ai'
when referrer in ('diary_detail','topic_detail','post_detail','user_post_detail','doctor_post_detail','question_detail','answer_detail','question_answer_detail') then 'content'
else null end as referrer
FROM online.bl_hdfs_maidian_updates
WHERE partition_date >= regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date < regexp_replace((current_date),'-','')
and action='page_view'
AND page_name IN ('diary_detail','topic_detail','post_detail','user_post_detail','doctor_post_detail','question_detail','answer_detail','question_answer_detail',
WHEN active_type in ('1','2') then '新增设备' END as active_type
,array(CASE WHEN first_channel_source_type like '%xinyouxingkong%' or a.device_id is not NULL THEN '可疑'
WHEN (partition_date>='20190601' and tmp.col2 = 'AI')
or (partition_date < '20200301' AND partition_date>='20190601' and first_channel_source_type like 'promotion_toutiao_jy%')
or (partition_date>='20200601' and ((first_channel_source_type like 'promotion_toutiao_jy%') or (first_channel_source_type like 'dyand%') or (first_channel_source_type like 'douyin%')))
THEN 'AI' ELSE '其他' END , '合计') as channel
FROM online.ml_device_day_active_status m
LEFT JOIN
(SELECT col1,col2 --col1:子渠道，col2：是否属于AI，col3:标识
FROM pm.tl_pm_ydl
WHERE col3='0204_danlei_channel')tmp
on first_channel_source_type=tmp.col1
LEFT JOIN
(SELECT DISTINCT device_id
FROM al.al_pm_ct_dv_deviceappversionrollbackfrom20200101_d
WHERE partition_day = regexp_replace(DATE_SUB(current_date,2) ,'-',''))a
ON m.device_id = a.device_id
where partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date<regexp_replace((current_date),'-','')
AND active_type in ('1','2','4')
and first_channel_source_type not in ('yqxiu1','yqxiu2','yqxiu3','yqxiu4','yqxiu5','mxyc1','mxyc2','mxyc3'
WHEN active_type in ('1','2') then '新增设备' END as active_type
,array(CASE WHEN first_channel_source_type like '%xinyouxingkong%' or a.device_id is not NULL THEN '可疑'
WHEN (partition_date>='20190601' and tmp.col2 = 'AI')
or (partition_date < '20200301' AND partition_date>='20190601' and first_channel_source_type like 'promotion_toutiao_jy%')
or (partition_date>='20200601' and ((first_channel_source_type like 'promotion_toutiao_jy%') or (first_channel_source_type like 'dyand%') or (first_channel_source_type like 'douyin%')))
THEN 'AI' ELSE '其他' END , '合计') as channel
FROM online.ml_device_day_active_status m
LEFT JOIN
(SELECT col1,col2 --col1:子渠道，col2：是否属于AI，col3:标识
FROM pm.tl_pm_ydl
WHERE col3='0204_danlei_channel')tmp
on first_channel_source_type=tmp.col1
LEFT JOIN
(SELECT DISTINCT device_id
FROM al.al_pm_ct_dv_deviceappversionrollbackfrom20200101_d
WHERE partition_day = regexp_replace(DATE_SUB(current_date,2) ,'-',''))a
ON m.device_id = a.device_id
where partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date<regexp_replace((current_date),'-','')
AND active_type in ('1','2','4')
and first_channel_source_type not in ('yqxiu1','yqxiu2','yqxiu3','yqxiu4','yqxiu5','mxyc1','mxyc2','mxyc3'
AND first_channel_source_type not like 'promotion\_jf\_%'
)t1
LEFT JOIN
(--找出user_id当天活跃的第一个设备id
SELECT user_id,partition_date,
if(size(device_list) > 0, device_list [ 0 ], '') AS device_id
FROM online.ml_user_updates
WHERE partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date<regexp_replace((current_date),'-','')
)t2
ON t1.partition_date=t2.partition_date
AND t1.device_id=t2.device_id
LEFT JOIN
( --有评论过日记帖的设备,排除疑似广告
SELECT t1.user_id,reply_date,t1.id,'topic_reply' as type
FROM
(
SELECT user_id,regexp_replace(substr(reply_date,1,10),'-','') as reply_date,problem_id,id
FROM online.tl_hdfs_topicreply_view
WHERE partition_date = regexp_replace(DATE_SUB(current_date,1) ,'-','')
AND is_spam = 'false' --排除疑似广告
-- and diary_id is not null 这个表的diary_id有问题，需要join problem表来判断是不是属于日记
and regexp_replace(substr(reply_date,1,10),'-','') >= regexp_replace(DATE_SUB(current_date,60) ,'-','')
and regexp_replace(substr(reply_date,1,10),'-','') < regexp_replace((current_date),'-','')
)t1
JOIN
(
SELECT id,diary_id
FROM online.tl_hdfs_problem_view
WHERE partition_date = regexp_replace(DATE_SUB(current_date,1) ,'-','')
group by id,diary_id
)t2
on t2.id=t1.problem_id
--group by t1.user_id,reply_date
UNION ALL
--有评论过回答的设备,排除疑似广告
SELECT
t1.user_id,t1.reply_date,t1.id as id,'answer_reply' as type
FROM
(
SELECT user_id,regexp_replace(substr(create_time,1,10),'-','') as reply_date,answer_id,id
FROM online.tl_hdfs_answer_reply_view
WHERE partition_date = regexp_replace(DATE_SUB(current_date,1) ,'-','')
AND (is_fake is NULL or is_fake = 'false')
AND answer_id is not NULL
and regexp_replace(substr(create_time,1,10),'-','') >= regexp_replace(DATE_SUB(current_date,60) ,'-','')
and regexp_replace(substr(create_time,1,10),'-','') < regexp_replace((current_date),'-','')
)t1
JOIN
(
SELECT id,question_id
FROM online.tl_hdfs_answer_view
WHERE partition_date =regexp_replace(DATE_SUB(current_date,1) ,'-','')
group by id,question_id
)t2
ON t2.id = t1.answer_id
UNION ALL
--有评论过用户帖的设备
SELECT user_id,regexp_replace(substr(create_time,1,10),'-','') as reply_date,id,'tractate_reply' as type
FROM online.tl_hdfs_api_tractate_reply_view
WHERE partition_date =regexp_replace(DATE_SUB(current_date,1) ,'-','')
and regexp_replace(substr(create_time,1,10),'-','') >=regexp_replace(DATE_SUB(current_date,60) ,'-','')
and regexp_replace(substr(create_time,1,10),'-','') < regexp_replace((current_date),'-','')
)t3
ON t2.partition_date = t3.reply_date
AND t2.user_id = t3.user_id
JOIN --限制用户是在app进行的回复
(
SELECT a.partition_date,user_id
FROM
(
SELECT partition_date,user_id,action
FROM online.bl_hdfs_operation_updates
WHERE partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date< regexp_replace((current_date),'-','')
)a
JOIN
(
SELECT code
FROM dim.dim_community_action_type
WHERE communityuserbehavior_type_name = '回帖'
)type
ON a.action = code
GROUP BY a.partition_date,user_id
)t4
ON t3.user_id = t4.user_id
AND t3.reply_date = t4.partition_date
LEFT JOIN
( --医生账号
SELECT distinct user_id
FROM online.tl_hdfs_doctor_view
WHERE partition_date = regexp_replace(DATE_SUB(current_date,1) ,'-','')
--马甲账号/模特用户
UNION ALL
SELECT user_id
FROM ml.ml_c_ct_ui_user_dimen_d
WHERE partition_day = regexp_replace(DATE_SUB(current_date,1) ,'-','')
AND (is_puppet = 'true' or is_classifyuser = 'true')
UNION ALL
--公司内网覆盖用户
select distinct user_id
from dim.dim_device_user_staff
UNION ALL
--登陆过医生设备
SELECT distinct t1.user_id
FROM
(
SELECT user_id, v.device_id as device_id
FROM online.ml_user_history_detail
LATERAL VIEW EXPLODE(device_history_list) v AS device_id
WHERE partition_date = regexp_replace(DATE_SUB(current_date,1) ,'-','')
) t1
JOIN
(
SELECT device_id
FROM online.ml_device_history_detail
WHERE partition_date = regexp_replace(DATE_SUB(current_date,1) ,'-','')
AND is_login_doctor = '1'
) t2
ON t1.device_id = t2.device_id
)t5
ON t3.user_id=t5.user_id
where (t5.user_id is null or t5.user_id = '')
)t6
LATERAL VIEW explode(t6.channel) t7 AS channel
GROUP BY partition_date,device_os_type,active_type,t7.channel
)T8
ON T1.partition_date=T8.partition_date
AND T1.device_os_type=T8.device_os_type
AND T1.active_type=T8.active_type
AND T1.channel=T8.channel
LEFT JOIN
(--部分页面的单设备页面浏览时长
SELECT partition_date
,device_os_type
,active_type
,t5.channel
,round(sum(CASE WHEN page_name like 'search%' THEN page_stay else 0 END)/count(distinct cl_id)/60,4) AS search_stay
,round(sum(CASE WHEN page_name = 'welfare_detail' THEN page_stay else 0 END)/count(distinct cl_id)/60,4) AS welfare_stay
,round(sum(CASE WHEN page_name = 'question_detail' THEN page_stay else 0 END)/count(distinct cl_id)/60,4) AS question_stay
,round(sum(CASE WHEN page_name in ('report_result','face_scan','face_detect_result','face_scan_loading','face_institute_report') THEN page_stay else 0 END)/count(distinct cl_id)/60,4) AS ai_related_stay
,round(sum(CASE WHEN page_name = 'diary_detail' THEN page_stay else 0 END)/count(distinct cl_id)/60,4) AS diary_stay
,round(sum(CASE WHEN page_name = 'home' THEN page_stay else 0 END)/count(distinct cl_id)/60,4) AS home_stay
,round(sum(CASE WHEN page_name in ('conversation_detail','consult_home') THEN page_stay else 0 END)/count(distinct cl_id)/60,4) AS conv_stay
WHEN active_type in ('1','2') then '新增设备' END as active_type
,array(CASE WHEN first_channel_source_type like '%xinyouxingkong%' or a.device_id is not NULL THEN '可疑'
WHEN (partition_date>='20190601' and tmp.col2 = 'AI')
or (partition_date < '20200301' AND partition_date>='20190601' and first_channel_source_type like 'promotion_toutiao_jy%')
or (partition_date>='20200601' and ((first_channel_source_type like 'promotion_toutiao_jy%') or (first_channel_source_type like 'dyand%') or (first_channel_source_type like 'douyin%')))
THEN 'AI' ELSE '其他' END , '合计') as channel
FROM online.ml_device_day_active_status m
LEFT JOIN
(SELECT col1,col2 --col1:子渠道，col2：是否属于AI，col3:标识
FROM pm.tl_pm_ydl
WHERE col3='0204_danlei_channel')tmp
on first_channel_source_type=tmp.col1
LEFT JOIN
(SELECT DISTINCT device_id
FROM al.al_pm_ct_dv_deviceappversionrollbackfrom20200101_d
WHERE partition_day = regexp_replace(DATE_SUB(current_date,2) ,'-',''))a
ON m.device_id = a.device_id
where partition_date>=regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date<regexp_replace((current_date),'-','')
AND active_type in ('1','2','4')
and first_channel_source_type not in ('yqxiu1','yqxiu2','yqxiu3','yqxiu4','yqxiu5','mxyc1','mxyc2','mxyc3'
AND first_channel_source_type not like 'promotion\_jf\_%'
)t1
JOIN
(--内容用户
SELECT partition_date,cl_id
FROM online.bl_hdfs_maidian_updates
WHERE partition_date >= regexp_replace(DATE_SUB(current_date,60) ,'-','')
AND partition_date < regexp_replace((current_date),'-','')
AND action = 'page_view'
AND page_name IN ('diary_detail','topic_detail','post_detail','user_post_detail','doctor_post_detail','question_detail','answer_detail','question_answer_detail',
