查询语法

查询http请求发送到查询节点(broker node)，然后查询节点转发至历史节点(historical node)或实时节点(realtime node)处理，最后合并结果返回

查询基本组件

Filter

过滤器，json对象，对维度进行筛选，表示满足filter的行是需要的数据，类似sql中的where子句

• selector filter 类似sql中的where key=value
  
  Shell

  1	"filter":{"type":"selector","dimension":...,"value":...}

• regex filter 使用正则，java正则表达式语法
  
  Shell

  1	"filter":{"type":"regex","dimension":...,"pattern":...}

• logical expression filter 支持and/or/not，允许嵌套
  
  Shell

  1 2 3

  "filter":{"type":"and","fields":[<filter>,...]} "filter":{"type":"or","fields":[<filter>,...]} "filter":{"type":"not","fields":<filter>}

• search filter 通过字符串匹配，有多种匹配方式，通过query.type指定
  
  Shell

  1	"filter":{"dimension":"product","query":{"type":"insensitive_contains","value":"foo"},"type":"search"}

• in filter 类似sql中的in
  
  Shell

  1	"filter":{"type":"in","dimension":...,"values":["foo",...]}

• bound filter 比较过滤器，支持字符串比较，默认就是字符串比较，要芀为数字需设置alphaNumeric为true
  
  Shell

  1 2 3 4

  "filter":{"type":"bound","dimension":"age","low":"21","upper":"31","alphaNumeric":true}  # 21<=age<=31 "filter":{"type":"bound","dimension":"age","low":"21",lowerStrict":true,"upper":"31","upperStrict":true,"alphaNumeric":true}  # 21<age<31 "filter":{"type":"bound","dimension":"age","upper":"31","upperStrict":true,"alphaNumeric":true} # age<31 "filter":{"type":"bound","dimension":"name","low":"foo","upper":"hoo"} # foo<=name<=hoo

• javascript filter 只支持一个入参，即filter里指定的维度值，返回true或false
  
  Shell

  1 2	# foo<=name<=hoo "filter":{"type":"javascript","dimension":"name","function":"function(x){return(x>='foo'&&x<='hoo'}"}

Aggregator

聚合器

• count aggregator 计算数据行数
  
  Shell

  1 2	{"type":"count","name":<output_name>} {"type":"longSum","name":<output_name>,"fieldName":"count"}  # 摄入了多少条原始数据

• sum aggregator 求和

  1、longSum 负责64位有符号整型的求和
  2、doubleSum 负责64位浮点数的求和

  Shell

  1 2	{"type":"longSum","name":<output_name>,"fieldName":<mertic_name>} {"type":"doubleSum","name":<output_name>,"fieldName":<mertic_name>}

• min/max aggregator 求最小/最大值

  1、doubleMin doubleMax
  2、longMin longMax

• cardinality aggregator 使用hyperloglog算法计算给定维度集合的基数 基数实际就是 count(distinct(value))
  对单个维度求基数，推荐hyperunique aggregator
  Shell

  1	{"type":"cardinality","name":<output_name>,"fieldNames":[<dimension1>,...],"byRow":<false|true>}

• hyperUnique aggregator 计算指定维度的基数
  Shell

  1	{"type":"hyperUnique","name":<output_name>,"fieldNames":<metric_name>}

• filtered aggretator 只对满足规则的维度进行聚合，可提升聚合效率
  Shell

  1	{"type":"filtered","filter":{"type":"selector","dimension":...,"value":...},"aggregator":<agg>}

• javascript aggregator 其中指定的列即为function的入参，执行性能会比较慢

Post-Aggregator

可以对Aggregator的结果进行二次加工并输出。最终的输出既包含Aggregator的结果，也包含Post-Aggregator的结果

• Arithmetic Post-Aggregator 对Agg的结果进行算术运算
• Field Accessor Post-Aggregator 返回指定Agg的值
• Constant Post-Aggregator 返回一个常数
• HyperUnique Cardinality Post-Aggregator 获取HyperUnique Aggretor的结果

Search Query

可以用在 filter 的 search 和 search查询中
定义了几种字符串匹配方式

• contains 包含 可通过 case_sensitive 指定是否区分大小写
• insensitive_contains 同 contains但不区分大小写
• fragment 可以指定多个值，任意匹配一个即可
  
  Shell

  1	{"type":"fragment","case_sensitive":false,"values":[xxx,...]}

Interval

指定时间区间 时间格式为ISO-8601 前闭后开
"intervals":["2016-08-28T00:00:00+08:00/2016-08-29T00:00:00+08:00"]

Context

在查询中可指定的参数，若不指定，会使用默认值

转载请注明：轻风博客 » druid笔记1-查询语法