Hi everyone! I am writing the requirements document for four query functions and would like input from the community before the semantics are fixed. - INTEGRAL: the time-weighted integral of a time series over the query range. The current library UDF UDAFIntegral sums the trapezoid areas between adjacent points and takes a time unit parameter. - SAMPLE(N): returns N points sampled at random from the query result. The current built-in EQUAL_SIZE_BUCKET_RANDOM_SAMPLE samples per bucket by a proportion parameter; there is no function that takes N points from the whole result set. - DERIVATIVE: the rate of change per unit time. It is already a built-in UDTF (UDTFCommonDerivative); it divides the value difference by the time difference in milliseconds and has no unit parameter. - TZ: time zone handling in queries. There is no such function yet. We want to define how a query states a time zone and what it changes: the display of timestamps, the computation of day and week boundaries, or both. Has anyone worked on functions like these? Looking forward any suggestion. ---
我正在编写四个查询函数的需求文档，希望在语义确定之前听取社区的意见。 - INTEGRAL：时间序列在查询区间上的时间加权积分。现有的 library UDF UDAFIntegral 累加相邻两点之间的梯形面积，并带一个时间单位参数。 - SAMPLE(N)：从查询结果中随机抽取 N 个点。现有的内置函数 EQUAL_SIZE_BUCKET_RANDOM_SAMPLE 按 proportion 参数在每个桶内抽样；没有从整个结果集中抽取 N 个点的函数。 - DERIVATIVE：单位时间的变化率。它已经是内置 UDTF（UDTFCommonDerivative）；它用值差除以毫秒时间差，没有单位参数。 - TZ：查询中的时区处理。目前没有这个函数。我们想定义查询如何指定时区，以及它改变什么：时间戳的显示、天和周边界的计算，或两者都改变。 社区里有人做过类似的函数吗？想了解这些边界情况是如何定义的; 期待大家的建议 Best regards Yaobin.Chen
