[parquet] Fix file index result filter the row ranges missing rowgroup offset problem

[Tan-JiaLiang]

Purpose

Fix #4780 filter row ranges missing rowgroup offset.

Tests

API and Format

Documentation

[Tan-JiaLiang]

Hi, @JingsongLi Can you please take a look? I made a mistake in #4780, it have been merged into v1.0 and need to have a fix.

[JingsongLi]

@Tan-JiaLiang Thanks! I will take a look!

[Tan-JiaLiang]

I ran the benchmark again

generate random values bound in [0, 1000) for BSI index field, and test predicate EQ with a random value（index result was no filter any ranges）

read	Best/Avg Time(ms)	Row Rate(K/s)	Per Row(ns)	Relative
without-bsi-index-1000-761	3982 / 4210	753.4	1327.3	1.0X
without-bsi-index-1000-391	3771 / 3810	795.6	1256.9	1.1X
without-bsi-index-1000-932	3848 / 3910	779.6	1282.6	1.0X
with-bsi-index-1000-761	3856 / 3885	778.1	1285.2	1.0X
with-bsi-index-1000-391	4062 / 4116	738.5	1354.0	1.0X
with-bsi-index-1000-932	4128 / 4148	726.7	1376.1	1.0X

generate random values bound in [0, 10000) for BSI index field, and test predicate EQ with a random value（index result maybe more filter 5-10 ranges）

read	Best/Avg Time(ms)	Row Rate(K/s)	Per Row(ns)	Relative
without-bsi-index-10000-2830	4270 / 4348	702.6	1423.2	1.0X
without-bsi-index-10000-1992	4055 / 4126	739.7	1351.8	1.1X
without-bsi-index-10000-1052	3933 / 4107	762.8	1311.0	1.1X
with-bsi-index-10000-2830	3046 / 3094	985.0	1015.3	1.4X
with-bsi-index-10000-1992	3327 / 3372	901.7	1109.0	1.3X
with-bsi-index-10000-1052	2840 / 2975	1056.3	946.7	1.5X

generate random values bound in [0, 20000) for BSI index field, and test predicate EQ with a random value（index result maybe more filter 15-20 ranges）

read	Best/Avg Time(ms)	Row Rate(K/s)	Per Row(ns)	Relative
without-bsi-index-20000-16842	4181 / 4447	717.5	1393.8	1.0X
without-bsi-index-20000-10432	4324 / 4484	693.8	1441.4	1.0X
without-bsi-index-20000-14386	4270 / 4420	702.5	1423.5	1.0X
with-bsi-index-20000-16842	2077 / 2213	1444.4	692.3	2.0X
with-bsi-index-20000-10432	2265 / 2408	1324.8	754.8	1.8X
with-bsi-index-20000-14386	2372 / 2442	1264.6	790.7	1.8X

values bound in [0, 50000), and test predicate EQ with a random value（index result maybe more filter 30-35 ranges）

read	Best/Avg Time(ms)	Row Rate(K/s)	Per Row(ns)	Relative
without-bsi-index-50000-12835	4280 / 4384	700.9	1426.8	1.0X
without-bsi-index-50000-4585	4456 / 4526	673.2	1485.5	1.0X
without-bsi-index-50000-28631	4300 / 4383	697.6	1433.4	1.0X
with-bsi-index-50000-12835	1117 / 1141	2686.9	372.2	3.8X
with-bsi-index-50000-4585	1256 / 1278	2389.1	418.6	3.4X
with-bsi-index-50000-28631	1322 / 1388	2269.7	440.6	3.2X

generate random values bound in [0, 100000) for BSI index field, and test predicate EQ with a random value（index result maybe more filter 30-45 ranges）

read	Best/Avg Time(ms)	Row Rate(K/s)	Per Row(ns)	Relative
without-bsi-index-100000-81044	4195 / 4216	715.2	1398.2	1.0X
without-bsi-index-100000-85026	4183 / 4332	717.3	1394.2	1.0X
without-bsi-index-100000-34524	4436 / 4705	676.2	1478.8	0.9X
with-bsi-index-100000-81044	739 / 764	4061.0	246.2	5.7X
with-bsi-index-100000-85026	512 / 582	5862.7	170.6	8.2X
with-bsi-index-100000-34524	524 / 603	5728.7	174.6	8.0X

generate random values bound in [0, 200000) for BSI index field, and test predicate EQ with a random value（index result maybe more filter 25-45 ranges）

read	Best/Avg Time(ms)	Row Rate(K/s)	Per Row(ns)	Relative
without-bsi-index-200000-131987	4100 / 4168	731.7	1366.8	1.0X
without-bsi-index-200000-119728	4110 / 4345	729.9	1370.0	1.0X
without-bsi-index-200000-18498	4293 / 4407	698.9	1430.9	1.0X
with-bsi-index-200000-131987	452 / 492	6637.3	150.7	9.1X
with-bsi-index-200000-119728	595 / 648	5040.5	198.4	6.9X
with-bsi-index-200000-18498	390 / 417	7699.7	129.9	10.5X

generate random values bound in [0, 500000) for BSI index field, and test predicate EQ with a random value（index result maybe filter all ranges）

read	Best/Avg Time(ms)	Row Rate(K/s)	Per Row(ns)	Relative
without-bsi-index-500000-90083	4162 / 4225	720.8	1387.3	1.0X
without-bsi-index-500000-277761	3935 / 4165	762.4	1311.6	1.1X
without-bsi-index-500000-290711	3949 / 4166	759.8	1316.2	1.1X
with-bsi-index-500000-90083	27 / 28	113039.1	8.8	156.8X
with-bsi-index-500000-277761	220 / 237	13616.6	73.4	18.9X
with-bsi-index-500000-290711	23 / 24	131814.8	7.6	182.9X

[JingsongLi]

+1