Table Generation Guaranteed Values

[stanbrub]

During table generation for benchmark tests, random is often used as a lazy way to provide a non-sequential distribution of data. The problem is that some tests require looking up generated values which may or may not be present. Consider the following snippet...

result = source.partition_by(['column3']).get_constituent(['random1'])

"random1" is a "column3" column value that is randomly generated. Depending on the scale selected, there is no guarantee that "random1" will exist as a value in the "column3" column.

Possible solutions:

• Replace random() on table generation with a random that always injects the first value in the defined range, then does random from then on
• Don't do random on columns at all. Do incremental data with overlapping ranges (ex. col1=[1-100], col2=[1-101] then shuffle the rows)