flink图表：analyze流程图

1、analyzetable流程图以tpc-ds生成的表作为例子，先从短小的reason表开始。它一共有3个字段，分别是r_reason_sk，r_reason_id，r_reason_desc。

2、流程图为一条直线，如下图所示，分别A、B、C三个方块。

3、A方块表达了从外部注入表的相关信息，table名，table中的字段名，长度等等。

4、B方块readfields，首先读入了三个字段信息，然后table整个传入buffer缓冲中，table保存在default_catalog,default_database,以reason表明保存。接着，select: (1 AS $f0, r_reason_sk, r_reason_id, CAST(CHAR_LENGTH(r_reason_id)) AS $f3, CHAR_LENGTH(r_reason_id) AS $f4, r_reason_desc, CAST(CHAR_LENGTH(r_reason_desc)) AS $f6, CHAR_LENGTH(r_reason_desc) AS $f7)，然后，select:(Partial_COUNT(*) AS count1$0, Partial_COUNT(APPROXIMATE DISTINCT r_reason_sk) AS $f1, Partial_COUNT(r_reason_sk) AS count$1, Partial_MAX(r_reason_sk) AS max$2, Partial_MIN(r_reason_sk) AS min$3, Partial_COUNT(APPROXIMATE DISTINCT r_reason_id) AS $f5, Partial_COUNT(r_reason_id) AS count$4, Partial_AVG($f3) AS (sum$5, count$6), Partial_MAX($f4) AS max$7, Partial_COUNT(APPROXIMATE DISTINCT r_reason_desc) AS $f9, Partial_COUNT(r_reason_desc) AS count$8, Partial_AVG($f6) AS (sum$9, count$10), Partial_MAX($f7) AS max$11)，每个字段按照PARTIAL_COUNT(APPROXIMATE DISTINCT),PARTIAL_COUNT,PARTIAL_MAX,PARTIAL_MIN的顺序进行select。

5、C方块Final_COUNT(count1$0) AS $f0, Final_COUNT(APPROXIMATE DISTINCT $f1) AS $f1, Final_COUNT(count$1) AS $f2, Final_MAX(max$2) AS $f3, Final_MIN(min$3) AS $f4, Final_COUNT(APPROXIMATE DISTINCT $f5) AS $f5, Final_COUNT(count$4) AS $f6, Final_AVG(sum$5, count$6) AS $f7, Final_MAX(max$7) AS $f8, Final_COUNT(APPROXIMATE DISTINCT $f9) AS $f9, Final_COUNT(count$8) AS $f10, Final_AVG(sum$9, count$10) AS $f11, Final_MAX(max$11) AS $f12),传进来了一堆参数，唯一能识别出的就是$f开头的都是B方块中的APPROXIMATE DISTINCT参数。select: ($f0 AS EXPR$0, $f1 AS EXPR$1, ($f0 - $f2) AS EXPR$2, 8.0:DOUBLE AS EXPR$3, 8.0 AS EXPR$4, $f3 AS EXPR$5, $f4 AS EXPR$6, $f5 AS EXPR$7, ($f0 - $f6) AS EXPR$8, $f7 AS EXPR$9, $f8 AS EXPR$10, null:INTEGER AS EXPR$11, null:INTEGER AS EXPR$12, $f9 AS EXPR$13, ($f0 - $f10) AS EXPR$14, $f11 AS EXPR$15, $f12 AS EXPR$16, null:INTEGER AS EXPR$17, null:INTEGER AS EXPR$18))然后把他们分门别类的传入expr中，最后用sink collection传出。