Avoid duplicate seek for single-row inserts

[timsehn]

Summary

• combine single-row insert and same-size replace fast-path dispatch into one seek
• keep the existing no-rechunk safety checks for inserts and existing same-size replacement behavior
• remove the separate insert fast path call from flush dispatch

Benchmark target

Latest merged PR #1035 sysbench comments highest multiplier:

• suite/key type: compositepk, 2-column INTEGER PRIMARY KEY(a,b) WITHOUT ROWID
• metric: types_delete_insert_ac
• mode: file-backed autocommit
• multiplier in generated comment: 4.98x

On current master the benchmark schema has been corrected, so the local targeted SQL used composite keys for sbtest_types.

Local timing

Targeted 100k-row corrected composite types_delete_insert_ac, Doltlite only, 5 samples each:

• clean master samples: 49,195 27,048 41,305 36,694 42,577 us, median 41,305 us
• this branch samples: 38,643 20,950 26,351 21,582 37,605 us, median 26,351 us

Tests

• make -j8 sqlite3.o libdoltlite.a
• make -j8 invariant_test corruption_test && ./invariant_test && ./corruption_test
• BENCH_ROWS=200 BENCH_RUNS=1 BENCH_MAX_MULTIPLIER=1000 BENCH_AVG_MAX_MULTIPLIER=1000 ./test/sysbench_compare_compositepk.sh
• make -j8 c-tests && test/run_c_tests.sh .

[github-actions]

Sysbench-Style Benchmark: Doltlite vs SQLite

In-Memory

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	24,051	35,747	1.49
oltp_range_select	10,525	14,270	1.36
oltp_sum_range	9,401	14,226	1.51
oltp_order_range	2,574	3,233	1.26
oltp_distinct_range	3,634	4,350	1.20
oltp_index_scan	3,893	6,815	1.75
select_random_points	10,256	17,657	1.72
select_random_ranges	2,925	5,246	1.79
covering_index_scan	4,095	4,428	1.08
groupby_scan	30,801	34,429	1.12
index_join	5,845	9,170	1.57
index_join_scan	3,338	5,352	1.60
types_table_scan	1,065,247	1,384,936	1.30
table_scan	1,223,967	1,548,277	1.26
oltp_read_only	103,639	135,753	1.31
Average			1.42

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	175,068	237,955	1.36
oltp_insert	15,059	27,265	1.81
oltp_update_index	51,058	103,477	2.03
oltp_update_non_index	33,803	67,975	2.01
oltp_delete_insert	45,206	78,453	1.74
oltp_write_only	22,526	49,076	2.18
types_delete_insert	24,038	44,614	1.86
oltp_read_write	70,178	125,601	1.79
Average			1.85

File-Backed

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	97,250	103,195	1.06
oltp_range_select	18,718	35,594	1.90
oltp_sum_range	18,519	35,247	1.90
oltp_order_range	3,474	6,226	1.79
oltp_distinct_range	4,568	7,291	1.60
oltp_index_scan	11,782	17,497	1.49
select_random_points	23,918	63,453	2.65
select_random_ranges	10,509	14,831	1.41
covering_index_scan	11,160	7,947	0.71
groupby_scan	32,999	50,499	1.53
index_join	10,441	17,476	1.67
index_join_scan	4,347	12,156	2.80
types_table_scan	1,206,826	2,704,626	2.24
table_scan	1,409,499	3,395,343	2.41
oltp_read_only	217,156	287,279	1.32
Average			1.77

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	184,022	262,503	1.43
oltp_insert	22,104	37,372	1.69
oltp_update_index	123,631	225,143	1.82
oltp_update_non_index	77,535	147,860	1.91
oltp_delete_insert	93,066	156,848	1.69
oltp_write_only	56,703	103,013	1.82
types_delete_insert	48,783	85,547	1.75
oltp_read_write	119,641	265,617	2.22
Average			1.79

File-Backed (autocommit)

Each statement runs as its own transaction — exposes per-commit
fixed costs that the wrapped-in-BEGIN/COMMIT tests amortize away.
SQLite uses WAL mode with synchronous=FULL in this section so
the comparison uses SQLite's durable WAL autocommit path.

Reads

Reads have no commit cost; these are the same SQL files as the
File-Backed Reads section, included here for symmetry and to
catch any per-statement overhead doltlite pays on the read path.

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	53,392	103,450	1.94
oltp_range_select	14,316	35,284	2.46
oltp_sum_range	14,123	35,165	2.49
oltp_order_range	3,103	6,221	2.00
oltp_distinct_range	4,141	7,305	1.76
oltp_index_scan	7,428	17,490	2.35
select_random_points	19,127	62,928	3.29
select_random_ranges	6,120	14,728	2.41
covering_index_scan	6,704	7,875	1.17
groupby_scan	32,814	50,301	1.53
index_join	7,974	17,639	2.21
index_join_scan	4,040	12,156	3.01
types_table_scan	1,199,735	2,704,809	2.25
table_scan	1,411,218	3,377,019	2.39
oltp_read_only	153,966	289,538	1.88
Average			2.21

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert_ac	22,083	79,842	3.62
oltp_insert_ac	24,456	97,977	4.01
oltp_update_index_ac	26,832	111,079	4.14
oltp_update_non_index_ac	22,188	88,880	4.01
oltp_delete_insert_ac	26,946	108,686	4.03
oltp_write_only_ac	27,471	97,054	3.53
types_delete_insert_ac	21,899	97,608	4.46
oltp_read_write_ac	31,299	115,604	3.69
Average			3.94

100000 rows, median of 5 invocations per test, workload-only timing via host monotonic clock when available.

Performance Ceiling Check (6x individual, 5x average)

All tests within ceilings.

[github-actions]

Sysbench-Style Benchmark (composite PK): Doltlite vs SQLite

Companion to the classic Sysbench-Style Benchmark. Every workload here
runs against tables with a 2-column INTEGER PRIMARY KEY(a, b) WITHOUT ROWID.
Individual ratios gated at 6×; section averages gated at 5×.

In-Memory

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	34,001	44,556	1.31
oltp_range_select	20,535	25,420	1.24
oltp_sum_range	18,477	25,702	1.39
oltp_order_range	3,796	4,431	1.17
oltp_distinct_range	4,854	5,507	1.13
oltp_index_scan	4,574	7,095	1.55
select_random_points	26,878	36,773	1.37
select_random_ranges	7,338	9,176	1.25
covering_index_scan	4,325	4,754	1.10
groupby_scan	40,329	45,218	1.12
index_join	8,242	12,123	1.47
index_join_scan	4,077	6,977	1.71
types_table_scan	1,126,095	1,517,707	1.35
table_scan	1,296,677	1,732,511	1.34
oltp_read_only	151,972	190,393	1.25
Average			1.32

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	241,969	319,812	1.32
oltp_insert	18,923	33,638	1.78
oltp_update_index	69,023	130,340	1.89
oltp_update_non_index	52,534	87,483	1.67
oltp_delete_insert	50,152	97,555	1.95
oltp_write_only	27,343	57,917	2.12
types_delete_insert	32,657	57,882	1.77
oltp_read_write	99,229	165,542	1.67
Average			1.77

File-Backed

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	125,635	126,282	1.01
oltp_range_select	31,282	54,419	1.74
oltp_sum_range	29,191	53,715	1.84
oltp_order_range	4,870	8,227	1.69
oltp_distinct_range	5,899	9,321	1.58
oltp_index_scan	14,339	21,301	1.49
select_random_points	45,990	99,378	2.16
select_random_ranges	16,280	21,462	1.32
covering_index_scan	12,699	11,201	0.88
groupby_scan	43,259	68,802	1.59
index_join	13,566	27,991	2.06
index_join_scan	5,402	18,652	3.45
types_table_scan	1,335,666	3,412,771	2.56
table_scan	1,591,043	4,245,656	2.67
oltp_read_only	294,889	389,001	1.32
Average			1.82

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	251,113	346,245	1.38
oltp_insert	25,377	49,961	1.97
oltp_update_index	158,696	292,217	1.84
oltp_update_non_index	104,768	186,522	1.78
oltp_delete_insert	108,089	205,460	1.90
oltp_write_only	63,467	132,922	2.09
types_delete_insert	60,968	108,693	1.78
oltp_read_write	167,890	350,376	2.09
Average			1.85

File-Backed (autocommit)

Each statement runs as its own transaction — exposes per-commit
fixed costs that the wrapped-in-BEGIN/COMMIT tests amortize away.
SQLite uses WAL mode with synchronous=FULL in this section so
the comparison uses SQLite's durable WAL autocommit path.

Reads

Reads have no commit cost; these are the same SQL files as the
File-Backed Reads section, included here for symmetry and to
catch any per-statement overhead doltlite pays on the read path.

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	70,124	130,394	1.86
oltp_range_select	26,237	54,727	2.09
oltp_sum_range	24,656	54,928	2.23
oltp_order_range	4,424	8,378	1.89
oltp_distinct_range	5,506	9,327	1.69
oltp_index_scan	8,825	21,758	2.47
select_random_points	40,499	100,246	2.48
select_random_ranges	11,005	21,648	1.97
covering_index_scan	7,142	11,093	1.55
groupby_scan	42,767	67,094	1.57
index_join	11,095	27,978	2.52
index_join_scan	4,909	18,249	3.72
types_table_scan	1,341,460	3,436,590	2.56
table_scan	1,598,182	4,271,081	2.67
oltp_read_only	218,837	383,586	1.75
Average			2.20

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert_ac	15,568	58,884	3.78
oltp_insert_ac	19,689	78,904	4.01
oltp_update_index_ac	19,351	91,505	4.73
oltp_update_non_index_ac	16,606	73,663	4.44
oltp_delete_insert_ac	17,453	82,513	4.73
oltp_write_only_ac	20,078	83,366	4.15
types_delete_insert_ac	15,644	69,953	4.47
oltp_read_write_ac	25,727	101,690	3.95
Average			4.28

100000 rows, median of 5 invocations per test, workload-only timing via host monotonic clock when available.

Performance Ceiling Check (6x individual, 5x average)

All tests within ceilings.

[github-actions]

Sysbench-Style Benchmark (BLOB PK): Doltlite vs SQLite

Companion to the classic Sysbench-Style Benchmark. Every workload here
runs against tables with a 16-byte big-endian BLOB PRIMARY KEY.
Individual ratios gated at 6×; section averages gated at 5×.

In-Memory

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	30,232	45,609	1.51
oltp_range_select	12,270	19,957	1.63
oltp_sum_range	11,622	19,246	1.66
oltp_order_range	2,870	3,831	1.33
oltp_distinct_range	3,932	4,891	1.24
oltp_index_scan	4,315	7,651	1.77
select_random_points	17,742	28,281	1.59
select_random_ranges	4,151	6,373	1.54
covering_index_scan	4,257	5,238	1.23
groupby_scan	32,275	39,141	1.21
index_join	6,696	12,755	1.90
index_join_scan	4,132	7,908	1.91
types_table_scan	1,140,774	1,597,725	1.40
table_scan	1,343,509	1,753,043	1.30
oltp_read_only	116,551	168,920	1.45
Average			1.51

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	239,516	338,322	1.41
oltp_insert	19,606	38,121	1.94
oltp_update_index	66,147	140,982	2.13
oltp_update_non_index	48,004	90,463	1.88
oltp_delete_insert	47,531	107,811	2.27
oltp_write_only	28,115	62,795	2.23
types_delete_insert	32,342	59,269	1.83
oltp_read_write	83,602	162,435	1.94
Average			1.96

File-Backed

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	104,652	119,406	1.14
oltp_range_select	21,819	48,896	2.24
oltp_sum_range	21,128	48,400	2.29
oltp_order_range	3,844	7,840	2.04
oltp_distinct_range	4,924	8,952	1.82
oltp_index_scan	12,840	21,283	1.66
select_random_points	33,304	82,513	2.48
select_random_ranges	11,418	17,464	1.53
covering_index_scan	11,690	12,910	1.10
groupby_scan	35,305	61,869	1.75
index_join	13,016	33,102	2.54
index_join_scan	6,550	23,180	3.54
types_table_scan	1,255,188	3,571,135	2.85
table_scan	1,476,246	4,254,377	2.88
oltp_read_only	238,517	361,128	1.51
Average			2.09

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	249,646	363,026	1.45
oltp_insert	36,139	58,663	1.62
oltp_update_index	172,757	308,326	1.78
oltp_update_non_index	105,259	187,845	1.78
oltp_delete_insert	113,683	214,882	1.89
oltp_write_only	71,208	134,766	1.89
types_delete_insert	63,178	115,971	1.84
oltp_read_write	155,863	346,999	2.23
Average			1.81

File-Backed (autocommit)

Each statement runs as its own transaction — exposes per-commit
fixed costs that the wrapped-in-BEGIN/COMMIT tests amortize away.
SQLite uses WAL mode with synchronous=FULL in this section so
the comparison uses SQLite's durable WAL autocommit path.

Reads

Reads have no commit cost; these are the same SQL files as the
File-Backed Reads section, included here for symmetry and to
catch any per-statement overhead doltlite pays on the read path.

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	61,165	120,386	1.97
oltp_range_select	17,249	49,136	2.85
oltp_sum_range	16,607	48,912	2.95
oltp_order_range	3,489	7,901	2.26
oltp_distinct_range	4,596	9,009	1.96
oltp_index_scan	8,484	21,266	2.51
select_random_points	28,787	84,530	2.94
select_random_ranges	7,260	17,452	2.40
covering_index_scan	7,680	13,019	1.70
groupby_scan	34,978	62,005	1.77
index_join	11,000	33,532	3.05
index_join_scan	6,173	23,451	3.80
types_table_scan	1,265,086	3,598,708	2.84
table_scan	1,499,077	4,290,122	2.86
oltp_read_only	173,300	365,031	2.11
Average			2.53

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert_ac	22,550	85,545	3.79
oltp_insert_ac	23,980	100,792	4.20
oltp_update_index_ac	28,453	114,617	4.03
oltp_update_non_index_ac	22,613	99,466	4.40
oltp_delete_insert_ac	25,517	106,718	4.18
oltp_write_only_ac	25,102	103,070	4.11
types_delete_insert_ac	23,158	93,426	4.03
oltp_read_write_ac	31,331	119,385	3.81
Average			4.07

100000 rows, median of 5 invocations per test, workload-only timing via host monotonic clock when available.

Performance Ceiling Check (6x individual, 5x average)

All tests within ceilings.

[github-actions]

Sysbench-Style Benchmark (TEXT PK): Doltlite vs SQLite

Companion to the classic Sysbench-Style Benchmark. Every workload here
runs against tables with a 32-char hex TEXT PRIMARY KEY (UUID-shaped).
Individual ratios gated at 6×; section averages gated at 5×.

In-Memory

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	31,015	46,239	1.49
oltp_range_select	13,751	20,469	1.49
oltp_sum_range	12,282	20,003	1.63
oltp_order_range	3,059	3,931	1.29
oltp_distinct_range	4,071	5,068	1.24
oltp_index_scan	4,759	8,043	1.69
select_random_points	18,402	29,959	1.63
select_random_ranges	4,177	6,721	1.61
covering_index_scan	4,869	5,447	1.12
groupby_scan	32,541	39,568	1.22
index_join	6,974	13,052	1.87
index_join_scan	4,545	8,037	1.77
types_table_scan	1,230,983	1,655,183	1.34
table_scan	1,512,557	1,807,207	1.19
oltp_read_only	128,252	178,896	1.39
Average			1.46

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	233,296	345,506	1.48
oltp_insert	22,368	39,894	1.78
oltp_update_index	76,789	156,852	2.04
oltp_update_non_index	48,868	94,332	1.93
oltp_delete_insert	55,105	111,169	2.02
oltp_write_only	30,558	64,827	2.12
types_delete_insert	33,641	63,298	1.88
oltp_read_write	85,211	168,077	1.97
Average			1.90

File-Backed

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	108,110	128,238	1.19
oltp_range_select	22,980	53,700	2.34
oltp_sum_range	22,293	53,168	2.38
oltp_order_range	3,946	8,334	2.11
oltp_distinct_range	5,066	9,316	1.84
oltp_index_scan	13,364	22,211	1.66
select_random_points	34,963	88,953	2.54
select_random_ranges	11,724	18,514	1.58
covering_index_scan	12,965	13,643	1.05
groupby_scan	36,205	65,617	1.81
index_join	15,243	35,162	2.31
index_join_scan	8,061	25,252	3.13
types_table_scan	1,337,334	3,996,239	2.99
table_scan	1,572,982	4,794,824	3.05
oltp_read_only	249,230	391,596	1.57
Average			2.10

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	248,660	383,986	1.54
oltp_insert	49,950	62,350	1.25
oltp_update_index	200,735	338,008	1.68
oltp_update_non_index	126,684	202,841	1.60
oltp_delete_insert	142,374	229,782	1.61
oltp_write_only	121,153	142,343	1.17
types_delete_insert	72,260	126,747	1.75
oltp_read_write	211,786	357,582	1.69
Average			1.54

File-Backed (autocommit)

Each statement runs as its own transaction — exposes per-commit
fixed costs that the wrapped-in-BEGIN/COMMIT tests amortize away.
SQLite uses WAL mode with synchronous=FULL in this section so
the comparison uses SQLite's durable WAL autocommit path.

Reads

Reads have no commit cost; these are the same SQL files as the
File-Backed Reads section, included here for symmetry and to
catch any per-statement overhead doltlite pays on the read path.

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	62,687	130,677	2.08
oltp_range_select	18,786	54,561	2.90
oltp_sum_range	17,682	54,485	3.08
oltp_order_range	3,775	8,518	2.26
oltp_distinct_range	4,740	9,482	2.00
oltp_index_scan	9,254	22,397	2.42
select_random_points	30,966	89,089	2.88
select_random_ranges	7,444	18,800	2.53
covering_index_scan	8,510	13,668	1.61
groupby_scan	35,402	66,077	1.87
index_join	13,074	36,737	2.81
index_join_scan	7,949	26,644	3.35
types_table_scan	1,319,857	4,036,013	3.06
table_scan	1,587,038	4,754,230	3.00
oltp_read_only	185,062	389,062	2.10
Average			2.53

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert_ac	22,189	93,314	4.21
oltp_insert_ac	26,877	105,911	3.94
oltp_update_index_ac	30,928	128,512	4.16
oltp_update_non_index_ac	25,996	107,368	4.13
oltp_delete_insert_ac	26,642	114,150	4.28
oltp_write_only_ac	26,887	107,146	3.99
types_delete_insert_ac	22,601	106,214	4.70
oltp_read_write_ac	34,551	127,313	3.68
Average			4.14

100000 rows, median of 5 invocations per test, workload-only timing via host monotonic clock when available.

Performance Ceiling Check (6x individual, 5x average)

All tests within ceilings.