Building database statistics across a join network using skew values
View Patent ↗A method that builds column statistics utilizing at least one skew value. The column statistics built using skew values, instead of column statistics built only through random sampling, may be used to more accurately reflect skew values across join networks, and thus enable a query optimizer to better select an access plan that is optimal under current runtime conditions.
1. A method of building database statistics, the method comprising:
(a) building a first column statistic for a first column in a first table, including detecting a skew value in the first column statistic;
(b) based upon the skew value detected in the first column statistic, using the skew value to build a second column statistic for a second column in a second table that is likely to be joined with the first column in a database query, wherein building the second column statistic comprises including the skew value with a random sample of values in the second column;
(c) storing the skew value for building at least one column statistic; and
(d) determining that the second column is more unique than the first column, wherein building the first column statistic is performed before building the second column statistic based upon the determination that the second column is more unique than the first column.
2. The method of claim 1 , wherein building the first column statistic comprises including the skew value with a random sample of values in the first column.
3. The method of claim 1 , wherein the first column statistic and the second column statistic include at least one of a frequent value list, a data distribution, a histogram, an index, and an encoded vector index.
4. The method of claim 1 , further comprising detecting at least one skew value in the second column statistic.
5. The method of claim 4 , further comprising rebuilding the first column statistic using the detected skew value in the second column statistic.