Improve wildcard and phrase search for the 1.9 release

docs/README.md

@@ -7,6 +7,7 @@
 
 - [Schema (SDL)](schema.md)
 - [Query language](query-language.md)
+- [Wildcard queries and expansion limits](wildcard-query.md)
 - [Tokenization and phrase queries](dynamic-tokenizer-and-phrase.md)
 - [Chunked text](chunked-text-fields.md) and [chunked BM25](chunked-bm25.md)
 - [Formula ranking, backfill, and diagnostics](candidate-rescoring.md)
@@ -63,6 +64,10 @@ status and fixture before treating a result as current behavior or performance.
 
 - [Full-text comparison](search-benchmark-current.md)
 - [Search review ledger](search-performance-review.md)
+- [Yonik Searchbench comparison](searchbench-comparison.md)
+- [Collector selection benchmark](collector-benchmark.md)
+- [IResearch optimization and Linux I/O audit](iresearch-optimization-audit.md)
+- [Topic-aware placement and redistribution proposal](topic-aware-placement.md)
 - [Range bitset word materialization](range-word-materialization.md): source comparison and measured filter-construction gains.
 - [Range block scan experiments](range-block-scans.md): header-based pruning and sequential decoding.
 - [Merge review](search-merge-review.md), [reorder review](reordering-performance-review.md), and [Rust hot paths](rust-hot-path-review.md)

docs/benchmark-results/collectors-2026-09-20.csv

@@ -0,0 +1,127 @@
+mode,fixture,k,algorithm,run,median_ns,lower_95_ns,upper_95_ns
+replay,bm25,10,heap,forward,39515.642489711936,39386.431547619046,39850.228600823044
+replay,bm25,10,heap,reverse,39306.20792642141,39187.473804240624,39355.03465662018
+replay,bm25,10,partition,forward,40378.012581168834,40300.745064102564,40576.22765837104
+replay,bm25,10,partition,reverse,40331.20761363636,40277.65635746606,40464.931249999994
+replay,bm25,10,loser_tree,forward,40703.54526994744,40585.32921245421,40983.482905982906
+replay,bm25,10,loser_tree,reverse,40940.42692307692,40841.43751204935,41313.692994505494
+replay,bm25,100,heap,forward,65890.77871817129,65455.179832559676,67814.83575148809
+replay,bm25,100,heap,reverse,73137.55103550295,72838.84615384616,73398.02243589744
+replay,bm25,100,partition,forward,51478.42261904762,50907.08275,52054.03965909091
+replay,bm25,100,partition,reverse,52493.31084120557,52388.82492690058,52724.83552631579
+replay,bm25,100,loser_tree,forward,71847.89354395604,71308.28422619047,72837.54910714286
+replay,bm25,100,loser_tree,reverse,71050.59659090909,70921.64841269841,71287.15622710623
+replay,bm25,1000,heap,forward,302907.0710180624,302253.9269480519,303911.0107142857
+replay,bm25,1000,heap,reverse,338297.9431818182,336406.8881752306,349206.5833333334
+replay,bm25,1000,partition,forward,115516.30475040257,114809.23093681916,116280.61388888888
+replay,bm25,1000,partition,reverse,134044.2004989496,130155.15188834154,140203.87074829932
+replay,bm25,1000,loser_tree,forward,347346.9734848485,346772.2428571428,347797.4257634828
+replay,bm25,1000,loser_tree,reverse,384665.1350877193,382161.15,387181.7318181818
+replay,improving,10,heap,forward,1927354.4204545454,1925226.4694444444,1931940.954347826
+replay,improving,10,heap,reverse,2257520.1166666667,2204796.0488505745,2260931.083333333
+replay,improving,10,partition,forward,1147094.9194078948,1137289.736111111,1155067.625
+replay,improving,10,partition,reverse,1276879.6283422462,1261305.4947916665,1313522.1833333333
+replay,improving,10,loser_tree,forward,2427187.7095238096,2423307.4541666666,2434081.9985714285
+replay,improving,10,loser_tree,reverse,2687434.2772727273,2592754.1799999997,2693464.3
+replay,improving,100,heap,forward,2884412.4356060605,2873111.25,2896138.517142857
+replay,improving,100,heap,reverse,3324969.6549707605,3256319.1136363633,3472792.344919786
+replay,improving,100,partition,forward,817892.4791666666,816569.5427489178,821563.5543478262
+replay,improving,100,partition,reverse,948630.7857142857,945023.0138888889,956221.3891941392
+replay,improving,100,loser_tree,forward,4276072.9375,4260562.5,4308059.875
+replay,improving,100,loser_tree,reverse,4923659.75,4731743.083333334,4995576.416666666
+replay,improving,1000,heap,forward,3803547.7333333334,3801858.535805627,3810335.3743589744
+replay,improving,1000,heap,reverse,4003068.5,3869324.4285714286,4206651.785714285
+replay,improving,1000,partition,forward,691638.7152777778,679697.4126984127,717763.4138257576
+replay,improving,1000,partition,reverse,770404.4895833333,744337.4624542124,773959.3537037037
+replay,improving,1000,loser_tree,forward,5932711.833333333,5922052.0,5946868.083333333
+replay,improving,1000,loser_tree,reverse,6543270.8,6533537.5,6699829.1
+replay,deteriorating,10,heap,forward,36859.167346938775,36645.45231331169,37078.5514520202
+replay,deteriorating,10,heap,reverse,56538.759114583336,52196.032799030174,61322.244999999995
+replay,deteriorating,10,partition,forward,38296.12146226416,38204.080338085885,38399.32884097035
+replay,deteriorating,10,partition,reverse,77053.9877277328,74788.38868561779,83035.52506265664
+replay,deteriorating,10,loser_tree,forward,38962.854853479854,38864.10203430031,39311.01037725031
+replay,deteriorating,10,loser_tree,reverse,49070.965248226945,48268.739614994935,60442.78219484883
+replay,deteriorating,100,heap,forward,38933.93972746331,38859.209014675056,38995.104309388036
+replay,deteriorating,100,heap,reverse,47526.9371968122,45327.53227365296,50262.38738738739
+replay,deteriorating,100,partition,forward,39995.56186868687,39642.04875,40210.557499999995
+replay,deteriorating,100,partition,reverse,45314.43052884615,44184.3925,45989.0103125
+replay,deteriorating,100,loser_tree,forward,40238.35784313726,40082.97507529658,40620.34749455338
+replay,deteriorating,100,loser_tree,reverse,48767.92530270656,46928.00545634921,49712.56018518518
+replay,deteriorating,1000,heap,forward,61527.09955495356,61063.70121703854,62191.833529411764
+replay,deteriorating,1000,heap,reverse,75952.72767857142,73572.73980331264,79128.69584980237
+replay,deteriorating,1000,partition,forward,55830.20755735492,55730.21655701754,55918.411864791284
+replay,deteriorating,1000,partition,reverse,72003.16311813187,68075.72231483686,72476.92564102565
+replay,deteriorating,1000,loser_tree,forward,56045.67907461214,55782.84766584767,56156.48568211068
+replay,deteriorating,1000,loser_tree,reverse,70712.8887394958,65167.42923809524,73313.83266666665
+replay,ties,10,heap,forward,145273.3596059113,144822.63775510204,145696.58214285714
+replay,ties,10,heap,reverse,194557.1458333333,189111.68306878308,203763.88535353536
+replay,ties,10,partition,forward,112782.69444444444,112081.40162037036,114679.54424242425
+replay,ties,10,partition,reverse,161494.20138888888,154579.4690625,166415.79861111112
+replay,ties,10,loser_tree,forward,148274.62797619047,148007.94047619047,149134.09989258862
+replay,ties,10,loser_tree,reverse,199826.83181818182,186615.3736614263,206855.11363636365
+replay,ties,100,heap,forward,158424.27756410255,157536.05726495726,159022.89476495725
+replay,ties,100,heap,reverse,191269.19675925927,182331.42402283708,193672.94973544974
+replay,ties,100,partition,forward,125054.31923076924,123380.84815415822,126176.36887254902
+replay,ties,100,partition,reverse,168717.9088235294,157976.24855072465,174601.15674603175
+replay,ties,100,loser_tree,forward,168467.0848765432,167588.2,169061.127228164
+replay,ties,100,loser_tree,reverse,259754.85625,255489.5625,266773.8894607843
+replay,ties,1000,heap,forward,354454.73148148146,351741.7691197691,356490.9666666667
+replay,ties,1000,heap,reverse,387409.4937254902,385767.35833333334,403647.4244444445
+replay,ties,1000,partition,forward,191395.3623737374,187124.04545454547,192862.95881006867
+replay,ties,1000,partition,reverse,215587.74655881937,212982.91910331382,217529.74170437406
+replay,ties,1000,loser_tree,forward,410887.5764705882,409895.35583333333,412327.92500000005
+replay,ties,1000,loser_tree,reverse,455186.65454545454,452074.6059523809,457247.2
+replay,clustered,10,heap,forward,41516.34364478115,40904.0352230235,42663.50347222222
+replay,clustered,10,heap,reverse,49155.541253306874,48181.134722222225,49336.78956228956
+replay,clustered,10,partition,forward,41969.29300471345,41864.54096466737,42128.091755319154
+replay,clustered,10,partition,reverse,50191.41722408027,48766.014274691355,50951.81481481482
+replay,clustered,10,loser_tree,forward,44631.651679841896,44552.57381748686,44744.87675831202
+replay,clustered,10,loser_tree,reverse,52227.587516087515,51934.32528957529,53784.001528677996
+replay,clustered,100,heap,forward,69373.88187229437,69240.1684833024,69566.18477304715
+replay,clustered,100,heap,reverse,77327.47649456523,77185.00047348485,80602.65104166666
+replay,clustered,100,partition,forward,50563.091124661245,50414.85756097561,50654.047933604335
+replay,clustered,100,partition,reverse,55985.227020202015,55817.6909141583,58091.93560606061
+replay,clustered,100,loser_tree,forward,81052.5882936508,80834.8822167756,81442.49939123377
+replay,clustered,100,loser_tree,reverse,90453.48472222223,89826.46169871796,92947.58699712643
+replay,clustered,1000,heap,forward,292577.28792517004,291810.27639751555,295090.6038961039
+replay,clustered,1000,heap,reverse,332390.87026143796,322121.26473063976,338677.16282051284
+replay,clustered,1000,partition,forward,109030.39692982455,108632.1906604747,109262.0817669173
+replay,clustered,1000,partition,reverse,130703.92098214285,126192.74553571429,133076.75531135531
+replay,clustered,1000,loser_tree,forward,358643.7870370371,357415.9517045454,360053.56327160494
+replay,clustered,1000,loser_tree,reverse,394679.6875,387402.1363478261,396441.1
+replay,short,10,heap,forward,382.88037522095453,376.96589020122485,398.6473491800367
+replay,short,10,heap,reverse,461.67594103757204,458.83830465569645,469.03517764543835
+replay,short,10,partition,forward,535.9522305013758,531.5048394476499,541.7262655040433
+replay,short,10,partition,reverse,638.7539588062211,615.5377545951317,683.917256830601
+replay,short,10,loser_tree,forward,637.7437525926975,631.9158046424202,647.4423122694276
+replay,short,10,loser_tree,reverse,683.300741355974,651.0807981563358,700.0568547362568
+replay,short,100,heap,forward,2080.6235712304842,2061.7285637727764,2102.022463379779
+replay,short,100,heap,reverse,2582.8162076271187,2440.259299423863,2611.7127275580665
+replay,short,100,partition,forward,1962.1091320816863,1955.9287845849803,1970.3574193017128
+replay,short,100,partition,reverse,2271.05851010101,2256.1186613386612,2284.3007954545456
+replay,short,100,loser_tree,forward,2709.0677053451905,2702.9229189485213,2722.5034251589695
+replay,short,100,loser_tree,reverse,2766.574433656958,2759.66109955309,2873.3630470796734
+replay,short,1000,heap,forward,2086.184443062505,2075.8592242019304,2106.0416244853886
+replay,short,1000,heap,reverse,2121.795488277956,2037.0621037836622,2157.832349468713
+replay,short,1000,partition,forward,1905.5393196856908,1792.02712744245,2033.9370159410482
+replay,short,1000,partition,reverse,1725.679718700533,1692.6827746818599,1728.7563046454197
+replay,short,1000,loser_tree,forward,2169.763476195736,2138.6431374038075,2212.0702713649252
+replay,short,1000,loser_tree,reverse,2241.539607097592,2160.275683002394,2250.8144681680806
+pruned,clustered,10,heap,forward,4625.274340068438,4488.440443213296,4715.8748592128095
+pruned,clustered,10,heap,reverse,4716.2780739161335,4691.9218905472635,4815.652112998227
+pruned,clustered,10,partition,forward,5368.813873626374,5252.453526979843,5384.057969294081
+pruned,clustered,10,partition,reverse,5375.838482142857,5365.1167133520075,5394.435288220551
+pruned,clustered,10,loser_tree,forward,7849.256310386474,7660.434135610766,7904.996630434783
+pruned,clustered,10,loser_tree,reverse,7894.331967663606,7874.323550817708,8093.476280269949
+pruned,clustered,100,heap,forward,34851.15984848485,33096.322878787876,35667.83867243867
+pruned,clustered,100,heap,reverse,39224.45735217164,38227.775025394774,39623.827142857146
+pruned,clustered,100,partition,forward,14393.420708955224,14010.33862153762,14882.715298507463
+pruned,clustered,100,partition,reverse,14860.283072299924,14807.179406474821,14899.513322675193
+pruned,clustered,100,loser_tree,forward,47827.40104166667,46402.03234126984,48661.06542397661
+pruned,clustered,100,loser_tree,reverse,47363.41858552632,46024.7932330827,47781.94028340081
+pruned,clustered,1000,heap,forward,267523.6328125,253148.75982142857,275183.8229166666
+pruned,clustered,1000,heap,reverse,288762.71081349207,282846.34621848736,291171.27976190473
+pruned,clustered,1000,partition,forward,78951.92307692308,77441.42439082656,79295.60576923078
+pruned,clustered,1000,partition,reverse,85734.74166666667,82977.26416666666,86183.44990079364
+pruned,clustered,1000,loser_tree,forward,435214.23333333334,421009.7,453387.6751304348
+pruned,clustered,1000,loser_tree,reverse,350604.01375000004,344713.76969696966,352747.63666666666

docs/benchmark-results/collectors-2026-09-20.json

@@ -0,0 +1,71 @@
+{
+  "date": "2026-09-20",
+  "production_revision": "15b8a3683fd563f03f1b4f8b14310c761303734e",
+  "benchmark_working_diff_sha256": "97b260e00f71b7957e46b65131105aab6ddd4d61d2753eeb369b9b8fb16deedc",
+  "source_sha256": {
+    "hermes-core/benches/collector_selection/main.rs": "76c0f776af06a3ba2347cbac19dbce4c3feae7c3487ce90cc1366fd6afdff507",
+    "hermes-core/benches/collector_selection/collectors.rs": "97627ab76081419a679d2787fbf5b8c56b12a5186b2d4bc6503798de56d4e38c"
+  },
+  "environment": {
+    "host": "macOS-15.6.1-arm64-arm-64bit-Mach-O",
+    "machine": "arm64",
+    "processor": "arm",
+    "cpu_model": "Apple M4",
+    "logical_cpus": 10,
+    "rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: aarch64-apple-darwin\nrelease: 1.98.1\nLLVM version: 22.1.8",
+    "cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
+    "flags": {
+      "RUSTFLAGS": "",
+      "CARGO_ENCODED_RUSTFLAGS": "",
+      "CARGO_BUILD_TARGET": "",
+      "CARGO_TARGET_DIR": "",
+      "CARGO_BUILD_JOBS": "",
+      "CRITERION_HOME": "",
+      "RAYON_NUM_THREADS": "",
+      "HERMES_PIN_MODE": "",
+      "HERMES_PIN_METADATA_BUDGET_MB": ""
+    }
+  },
+  "runs": {
+    "forward": {
+      "command": [
+        "cargo",
+        "bench",
+        "--locked",
+        "-p",
+        "hermes-core",
+        "--bench",
+        "collector_selection",
+        "--",
+        "--save-baseline",
+        "collectors-forward"
+      ],
+      "elapsed_seconds": 101.84161683311686,
+      "collector_reverse": false
+    },
+    "reverse": {
+      "command": [
+        "cargo",
+        "bench",
+        "--locked",
+        "-p",
+        "hermes-core",
+        "--bench",
+        "collector_selection",
+        "--",
+        "--save-baseline",
+        "collectors-reverse"
+      ],
+      "elapsed_seconds": 100.68795216735452,
+      "collector_reverse": true
+    }
+  },
+  "samples_per_case": 30,
+  "warmup_seconds": 0.3,
+  "target_measurement_seconds": 1,
+  "case_count_per_run": 63,
+  "metric": "Criterion median point estimate and its within-run 95% bootstrap confidence interval, in nanoseconds",
+  "scope": "Synthetic score replay and exact-block-bound feedback; includes collector allocation, screening, admission, sorting and destruction. Excludes scoring, decode, I/O, positions, exact-count collection, seeded/shared thresholds and end-to-end query execution.",
+  "noise": "Active desktop background processes; repeat slowed several cases. Small differences inconclusive. No agent compilation/tests overlapped timed runs.",
+  "memory": "Backing capacities: heap 12*k bytes, partition 24*k, loser tree 28*k. Object sizes 48/48/72 bytes. Excludes input fixtures, allocator overhead, output and oracle; not RSS."
+}