sync.Pool usage

[marniks7]

Hi,
I was testing different bitmap libraries (by the way as a replacement for hash map)
The results i received for 8 and operations for 500k entities where first bitmap has 10k entities in a row with 1 bit, Clone of the bitmap is used on each operation

goos: linux
goarch: amd64
pkg: bitmap-usage/benchmark/500k-large-groups/kelindar
cpu: Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz
BenchmarkFindPriceV2_11position            	   26214	     41542 ns/op	   67760 B/op	       8 allocs/op
BenchmarkFindPriceV2_3824position          	   45291	     40293 ns/op	   73904 B/op	       8 allocs/op
BenchmarkFindPriceV2_3824position_OptStats 	   43438	     35956 ns/op	   73952 B/op	      13 allocs/op
BenchmarkFindPriceV2_9701position          	   41796	     36919 ns/op	   65712 B/op	       7 allocs/op
BenchmarkFindPriceV2_MultiplePricesErr     	   44133	     24858 ns/op	   73856 B/op	       6 allocs/op
PASS
ok  	bitmap-usage/benchmark/500k-large-groups/kelindar	51.610s

This is what i received for https://github.com/kelindar/column

goos: linux
goarch: amd64
pkg: bitmap-usage/benchmark/500k-large-groups/kelindar-column
cpu: Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz
BenchmarkFindPriceV2_11position            	   40641	     25217 ns/op	     216 B/op	       8 allocs/op
BenchmarkFindPriceV2_3824position          	   45633	     25901 ns/op	     216 B/op	       8 allocs/op
BenchmarkFindPriceV2_9701position          	   44162	     26637 ns/op	     217 B/op	       8 allocs/op
BenchmarkFindPriceV2_MultiplePricesErr     	   48385	     24083 ns/op	     177 B/op	       6 allocs/op
PASS
ok  	bitmap-usage/benchmark/500k-large-groups/kelindar-column	107.145s

In your kelindar-column you have fill bitmap.Bitmap // The fill-list and Txn with sync.Pool usage which allow to avoid bitmap allocation. I believe the technique is worth mentioning for potential users of this library

Results with sync.Pool:

goos: linux
goarch: amd64
pkg: bitmap-usage/benchmark/500k-large-groups/kelindar
cpu: Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz
BenchmarkFindPriceV2_11position            	   79882	     15122 ns/op	     176 B/op	       6 allocs/op
BenchmarkFindPriceV2_3824position          	   85920	     14792 ns/op	     176 B/op	       6 allocs/op
BenchmarkFindPriceV2_3824position_OptStats 	   68959	     15563 ns/op	     225 B/op	      11 allocs/op
BenchmarkFindPriceV2_9701position          	   56557	     19224 ns/op	     177 B/op	       6 allocs/op
BenchmarkFindPriceV2_MultiplePricesErr     	  106196	     11107 ns/op	     128 B/op	       4 allocs/op
PASS
ok  	bitmap-usage/benchmark/500k-large-groups/kelindar	61.455s

if you are curious here is hashmap results - iteration and comparison for 10k entities

goos: linux
goarch: amd64
pkg: bitmap-usage/benchmark/500k-large-groups/map
cpu: Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz
BenchmarkHttpClientServer_FindPrice            	    5232	    220943 ns/op	    8283 B/op	     112 allocs/op
BenchmarkFindPrice_11position                  	    9979	    118803 ns/op	      64 B/op	       1 allocs/op
BenchmarkFindPrice_3824position                	    9849	    119427 ns/op	      64 B/op	       1 allocs/op
BenchmarkFindPrice_3824position_Optimized      	   25146	     46990 ns/op	      64 B/op	       1 allocs/op
BenchmarkFindPrice_9701position                	    9825	    120534 ns/op	      64 B/op	       1 allocs/op
BenchmarkFindPrice_9701position_Optimized      	    9945	    117033 ns/op	      64 B/op	       1 allocs/op
BenchmarkFindPrice_MultiplePricesErr           	    9799	    119765 ns/op	       0 B/op	       0 allocs/op
BenchmarkFindPrice_MultiplePricesErr_Optimized 	    9814	    119057 ns/op	       0 B/op	       0 allocs/op
PASS
ok  	bitmap-usage/benchmark/500k-large-groups/map	47.843s

[kelindar]

@marniks7 Thanks, I'll add this to documentation. Any other suggestions on what to improve in bitmap or column packages? By the way, great work on marniks7/bitmap-usage!