Error in TPCH-Q1

[rgankema]

I was looking at the benchmark queries to get an idea of what more complex Weld queries look like, and I think I stumbled across a bug in the following part of the query:

merge(b, { 
    i64(e.$0*2c + e.$1),
    {
        e.$2,
        e.$3,
        sum_disc_price,
        sum_disc_price * (1.0f + e.$6),
        e.$4,
        1
    }
})

Here values are grouped by returnflag * 2 + linestatus. However, as the value for returnflag ranges from 0 to 1 (inclusive), and that for linestatus from 0 to 2 (inclusive), this means that any value with (returnflag=0, linestatus=2) will be mapped to the same group values with (returnflag=1, linestatus=0), as both will be mapped to index 2 == (0*2+2) == (1*2+0).

Changing the mapping to i64(e.$0*3c + e.$1) should solve the issue.