Query Execution Optimization in Spark SQL

Algorithm 1

Algorithm to construct variable-width distribution histogram.

	Construct variable-width distribution histogram by frequency distribution histogram
	input: Hfre = {<attr1, freq1>, <attr2, freq2>, …, <attrn, freqn>}
	output: Hwidth = {<start1, end1, times1>,<start2, end2, times2>, …, <startm, endm, timesm>}
	procedure
	i ⟵ 1; Hwidth ⟵ {}
	start ⟵ attr1; end ⟵ attr1;
	max ⟵ freq1; T ⟵ freq1
	while i ≤ n do
	i ⟵ i + 1
	if |max-freqi|/freqi < 0.05 then
	end ⟵ attri
	T ⟵ T + freqi
	if freqi > max then
	max ⟵ freqi
	end if
	else
	Hwidth ⟵ Hwidth + <start, end, T>
	start ⟵ attri; end ⟵ attri
	max ⟵ freqi; T ⟵ freqi
	end if
	end while
	end procedure