Poor performance when using rows to scan in data

[jimmiebtlr]

Describe the bug
When using rows to read row by row, the performance is around 10x worse than using raw sql.

Note, I also gave it a try using raw sql, and using qrm.ScanOneRowToDest, with the same poor performance seen there.

Some additional background

• Using a struct with 3 not huge jet models (total probably 30 fields, 10 of which are duplicate with an AS clause, roughtly 10 fields used total out of them).
• Reducing fields to 1 model with 11 fields cuts the time 50%
• I'm returning 1M rows.
• Similar code that uses Query or QueryContext instead of Rows is returning in around 25% the time.

My guess is the overhead of reflecting on every row is driving this behavior.

Environment (please complete the following information):

• OS: macosx
• Database: mysql
• Database driver: mysql
• Jet version 2.5

Code snippet

q, args := query.Sql()
rows, err := db.Query(q, args...)
// handle err and defer close

for rows.Next() {
	rowData := generator()
	err = qrm.ScanOneRowToDest(rows, rowData)
	// handle err
}

Expected behavior
I expect the results to take < 40% current time.

[go-jet]

Hi @jimmiebtlr. Managed to reproduce it. New scanContext is unnecessarily created with each Scan call.
I've made a fix on the rows-perf branch. Rows scan should now be several times faster.

[jimmiebtlr]

Tested rows-perf branch and performance is much improved thanks!!

[go-jet]

Fixed with Release 2.7.1