pith. sign in

arxiv: 2606.20853 · v1 · pith:LFPNAKETnew · submitted 2026-06-18 · 💻 cs.DB

ReSequel: Robust LLM-assisted Query Rewriting and Optimization using Templatization and Sampling

Saeed Fathollahzadeh , Essam Mansour , Matthias Boehm This is my paper

Pith reviewed 2026-06-26 14:55 UTC · model grok-4.3

classification 💻 cs.DB
keywords SQL query rewritingLLM-assisted optimizationdatabase performancetemplatizationdata samplingquery verificationDBMSbenchmarks
0
0 comments X

The pith

ReSequel uses LLM query rewriting with templatization and sampling to deliver up to 16x speedups over native DBMSs.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents ReSequel as a method to improve SQL query performance by rewriting queries using large language models. It addresses challenges in traditional heuristic rewriting and pure LLM approaches by using metadata to create template-specific rules and verifying rewrites on sampled data. This allows adaptation to specific queries and databases while ensuring correctness. Experiments demonstrate significant speedups across various benchmarks and database systems. The approach acts as an outer layer on top of existing DBMSs without modifying their internal optimizers.

Core claim

ReSequel is an outer optimization layer on top of existing DBMSs that rewrites SQL queries using LLMs guided by template-specific rules inferred from catalog and statistical metadata. Rewritten query variants are generated, verified for result correctness, and ranked by runtime on sampled data to ensure both semantic equivalence and performance improvements.

What carries the argument

Templatization to infer template-specific rules from metadata that guide the LLM, combined with sampling-based generation, verification, and ranking of query variants.

If this is right

  • Workload-level speedups of up to 16x over native DBMSs.
  • Up to 22x speedups over LLM-based systems.
  • Individual queries can see speedups exceeding 600x.
  • Consistent gains across eight benchmarks including JOB, TPC-H, and three DBMSs: PostgreSQL, MySQL, DuckDB.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The method could potentially be adapted for other data query languages beyond SQL.
  • Integration with cost-based optimizers might further enhance the rewrites.
  • Long-term, this could shift focus from maintaining static rewrite rules to dynamic, data-driven ones.
  • Further testing on production workloads with varying data distributions would strengthen the approach.

Load-bearing premise

That verification and performance measurement on sampled data accurately predicts behavior on the full dataset without issues from data distribution shifts or rare values.

What would settle it

A rewritten query that passes verification and shows improvement on samples but fails to produce correct results or delivers no speedup when executed on the complete dataset.

Figures

Figures reproduced from arXiv: 2606.20853 by Essam Mansour, Matthias Boehm, Saeed Fathollahzadeh.

Figure 1
Figure 1. Figure 1: Comparison of ReSequel’s rewriting impact on host [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Challenges in Phase Ordering and Query Optimization Support in DBMS and Non-DBMS Systems. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: ReSequel Architecture and Workflow. User u Badges b Post p History h 11 12 13 [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Materialization Steps for STATS 𝑄#122 ( [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Example of Normalization, Masking, and Generalization of Operators for Two Queries / Parameter Materialization. [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Extracted Template from Queries #1 & #2 in [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Example Template Clustering and Schemas. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Overall Performance Impact on Various DBMSs Across 5 Datasets (LLM: Gemini-2.5-pro; 10% Downsampling). [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 11
Figure 11. Figure 11: Time Breakdown of Rewriting Across Eight Datasets (Downsampling Rate = 10%, DBMS = PostgreSQL). [PITH_FULL_IMAGE:figures/full_fig_p010_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Performance Impact on Three Datasets. Impact of Different LLMs [PITH_FULL_IMAGE:figures/full_fig_p010_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Impact of Queries with Messy Characteristics Across Datasets (Gemini-2.5-pro & PostgreSQL). [PITH_FULL_IMAGE:figures/full_fig_p011_13.png] view at source ↗
Figure 18
Figure 18. Figure 18: Validation and Top-1 Selection Overhead with [PITH_FULL_IMAGE:figures/full_fig_p011_18.png] view at source ↗
Figure 16
Figure 16. Figure 16: Individual Query Performance Impact across [PITH_FULL_IMAGE:figures/full_fig_p011_16.png] view at source ↗
Figure 19
Figure 19. Figure 19: Verification Rates Across Different Sample Sizes [PITH_FULL_IMAGE:figures/full_fig_p012_19.png] view at source ↗
read the original abstract

Heuristic query rewriting has long complemented cost-based optimization to improve performance. Such rewrites transform SQL queries into semantically equivalent forms that are easier or faster to execute. Examples are standardizing expressions, eliminating redundancy, propagating constants, pushing down selections and projections, unnesting queries, and utilizing constraints. Modern DBMSs implement hundreds to thousands of such rules, but maintaining them is notoriously difficult. The interactions among rules are complex, and their static nature and application order prevent adaptation to specific query and database characteristics. Recent approaches that use large language models (LLMs) for query rewriting show promise but face challenges regarding the large search space, reliable query verification, and exploitation of metadata. We present ReSequel, an outer optimization layer on top of existing DBMSs to rewrite SQL queries using LLMs. ReSequel leverages catalog and statistical metadata to infer template-specific rules that guide the LLM toward effective query transformations. We generate, verify, and rank rewritten query variants on sampled data to ensure result correctness and runtime improvements. Our experiments cover eight benchmarks: JOB, TPC-H, Stats(-CEB), Public BI, IMDB, DSB, and StackOverflow; multiple DBMSs: PostgreSQL, MySQL, and DuckDB; as well as LLM-based query rewriting baselines. ReSequel yields workload-level speedups of up to 16x over native DBMSs and 22x over LLM-based systems, with individual queries exceeding 600x, across eight benchmarks and three DBMSs.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. ReSequel is presented as an outer optimization layer atop existing DBMSs that uses LLMs guided by catalog/statistical metadata and template-specific rules to generate SQL rewrites; rewrites are generated, verified for semantic equivalence, and ranked for runtime improvement by execution on sampled data, with experiments across eight benchmarks (JOB, TPC-H, Stats-CEB, Public BI, IMDB, DSB, StackOverflow) and three DBMSs (PostgreSQL, MySQL, DuckDB) claiming workload speedups up to 16× over native engines and 22× over prior LLM baselines, with individual queries exceeding 600×.

Significance. If the central empirical claims hold under rigorous full-data verification, the work would be significant for demonstrating a practical, metadata-aware LLM integration into query optimization that addresses rule-maintenance difficulties in traditional rewriters while providing measurable gains over both native and prior LLM approaches; the multi-DBMS, multi-benchmark scope and use of templatization to constrain the search space are strengths.

major comments (3)
  1. [§4] §4 (Experimental Methodology) and abstract: the reported speedups rest on sampled-data verification of both equivalence and runtime; the manuscript provides no quantitative characterization of sample sizes, sampling method (e.g., uniform vs. stratified), or failure rates of the verification step, leaving open whether rare predicates, low-selectivity joins, or skewed aggregates are adequately covered.
  2. [§3.3] Verification procedure (likely §3.3): semantic equivalence and runtime gains observed on a sample do not necessarily transfer to the full dataset when cardinality estimates, plan selection, or I/O behavior change with full data volume; the headline 16× workload and 600× single-query claims therefore require explicit evidence that sampled checks are sufficient proxies.
  3. [Results tables] Table or figure reporting per-benchmark results: without disclosure of how many rewrites were rejected by the verifier, how often the LLM produced syntactically invalid SQL, or the distribution of speedups conditional on successful verification, the aggregate speedup numbers cannot be interpreted as robust.
minor comments (2)
  1. [§3.1] Notation for template rules and metadata usage could be formalized with a small example query and its template to improve reproducibility.
  2. [Abstract and §4] The abstract lists eight benchmarks but the text should explicitly state whether all were run on all three DBMSs or whether some combinations were omitted.

Simulated Author's Rebuttal

3 responses · 0 unresolved

Thank you for the thorough review and constructive feedback on our manuscript. We appreciate the referee's recognition of the significance of our work and address each major comment point by point below. We will make revisions to improve the clarity and robustness of the experimental methodology section.

read point-by-point responses

  1. Referee: [§4] §4 (Experimental Methodology) and abstract: the reported speedups rest on sampled-data verification of both equivalence and runtime; the manuscript provides no quantitative characterization of sample sizes, sampling method (e.g., uniform vs. stratified), or failure rates of the verification step, leaving open whether rare predicates, low-selectivity joins, or skewed aggregates are adequately covered.

    Authors: We acknowledge that the current manuscript does not provide sufficient quantitative details on the sampling procedure used for verification. In the revised manuscript, we will expand §4 to include: (1) the sampling method, which is uniform random sampling of a fixed number of rows (10,000 rows for large tables, or 1% for smaller ones); (2) sample sizes for each benchmark; and (3) failure rates, which were approximately 15% for equivalence checks across the experiments. We will also add a discussion on how the metadata-guided templatization helps in covering diverse query patterns, including those with rare predicates. revision: yes

  2. Referee: [§3.3] Verification procedure (likely §3.3): semantic equivalence and runtime gains observed on a sample do not necessarily transfer to the full dataset when cardinality estimates, plan selection, or I/O behavior change with full data volume; the headline 16× workload and 600× single-query claims therefore require explicit evidence that sampled checks are sufficient proxies.

    Authors: This point highlights a potential limitation of sampling-based verification. While we believe the sampled verification is a reasonable proxy because the rewrites are semantically equivalent and the runtime improvements are measured on the same DBMS, we agree that explicit evidence is valuable. In the revision, we will add a new experiment section where we verify a random sample of 50 queries on the full dataset to confirm that the speedups are consistent. We will report the correlation between sample and full-data speedups. However, full verification for all queries is not feasible due to time constraints, which is why sampling is used in the first place. revision: partial

  3. Referee: [Results tables] Table or figure reporting per-benchmark results: without disclosure of how many rewrites were rejected by the verifier, how often the LLM produced syntactically invalid SQL, or the distribution of speedups conditional on successful verification, the aggregate speedup numbers cannot be interpreted as robust.

    Authors: We agree that additional transparency on the verification outcomes would strengthen the results. We will revise the results section to include a new table or subsection reporting: the average number of rewrites generated per query, the percentage of rewrites rejected due to failed equivalence or runtime regression, the rate of invalid SQL (which was low at under 5% due to our templatization), and the distribution of speedups (e.g., median, quartiles) for accepted rewrites. This will allow readers to better interpret the aggregate numbers. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical results on external benchmarks

full rationale

The paper describes a system (ReSequel) that applies LLMs to SQL rewriting guided by templates and metadata, then verifies/ranks candidates via execution on sampled data before reporting full-dataset speedups. All load-bearing claims are direct experimental measurements (workload speedups up to 16×, single-query up to 600×) on eight named external benchmarks across three DBMSs. No equations, fitted parameters, or self-referential derivations appear; the sampling step is a methodological choice whose validity is an empirical question, not a definitional reduction. No self-citation chains are invoked as uniqueness theorems or load-bearing premises. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The system rests on the domain assumption that LLMs can be steered by metadata-derived rules to produce semantically equivalent rewrites and that sampling suffices for verification; no free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption LLMs guided by template-specific rules inferred from catalog and statistical metadata will produce semantically equivalent and faster query rewrites
    This premise is required for the outer optimization layer to function as described.
  • domain assumption Verification of equivalence and performance on sampled data generalizes to the full dataset
    Invoked when the abstract states that rewrites are generated, verified, and ranked on sampled data.

pith-pipeline@v0.9.1-grok · 5807 in / 1481 out tokens · 30559 ms · 2026-06-26T14:55:01.974884+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

110 extracted references · 57 canonical work pages · 1 internal anchor

  1. [1]

    Apache Calcite is a dynamic data management framework

    2025. Apache Calcite is a dynamic data management framework. https: //calcite.apache.org/

    2025

  2. [2]

    Groq Cloud

    2025. Groq Cloud. https://console.groq.com/

    2025

  3. [3]

    Peter Akioyamen, Zixuan Yi, and Ryan Marcus. 2024. The Unreasonable Effectiveness of LLMs for Query Optimization.NeurIPS(2024). https: //doi.org/10.48550/ARXIV.2411.02862

    work page doi:10.48550/arxiv.2411.02862 2024

  4. [4]

    Astrahan et al

    Morton M. Astrahan et al. 1976. System R: Relational Approach to Database Management.ACM Trans. Database Syst.1, 2 (1976), 97–137. https://doi.org/10. 1145/320455.320457

    arXiv 1976

  5. [5]

    Qiushi Bai, Sadeem Alsudais, and Chen Li. 2023. QueryBooster: Improving SQL Performance Using Middleware Services for Human-Centered Query Rewriting. PVLDB16, 11 (2023), 2911–2924. https://doi.org/10.14778/3611479.3611497

    work page doi:10.14778/3611479.3611497 2023

  6. [6]

    Eirik Bakke and David R. Karger. 2016. Expressive Query Construction through Direct Manipulation of Nested Relational Results. InSIGMOD. 1377–1392. https: //doi.org/10.1145/2882903.2915210

    work page doi:10.1145/2882903.2915210 2016

  7. [7]

    Zhuowei Bao, Benny Kimelfeld, and Yunyao Li. 2012. Automatic suggestion of query-rewrite rules for enterprise search. InSIGIR. 591–600. https://doi.org/10. 1145/2348283.2348363

    arXiv 2012

  8. [8]

    Alexander Baumstark, Muhammad Attahir Jibril, and Kai-Uwe Sattler. 2023. Adaptive query compilation in graph databases.Distributed Parallel Databases 41, 3 (2023), 359–386. https://doi.org/10.1007/S10619-023-07430-4

    work page doi:10.1007/s10619-023-07430-4 2023

  9. [9]

    Mior, and Daniel Lemire

    Edmon Begoli, Jesús Camacho-Rodríguez, Julian Hyde, Michael J. Mior, and Daniel Lemire. 2018. Apache Calcite: A Foundational Framework for Optimized Query Processing Over Heterogeneous Data Sources. InSIGMOD. 221–230. https://doi.org/10.1145/3183713.3190662

    work page doi:10.1145/3183713.3190662 2018

  10. [10]

    InProceedings of the 1995 ACM SIGMOD International Conference on Management of Data, pages 1–10, 1995

    Hal Berenson, Philip A. Bernstein, Jim Gray, Jim Melton, Elizabeth J. O’Neil, and Patrick E. O’Neil. 1995. A Critique of ANSI SQL Isolation Levels. InSIGMOD. 1–10. https://doi.org/10.1145/223784.223785

    work page doi:10.1145/223784.223785 1995

  11. [11]

    Altan Birler, Alfons Kemper, and Thomas Neumann. 2024. Robust Join Pro- cessing with Diamond Hardened Joins.PVLDB17, 11 (2024), 3215–3228. https://doi.org/10.14778/3681954.3681995

    work page doi:10.14778/3681954.3681995 2024

  12. [12]

    Burdick, Alexandre V

    Matthias Böhm, Douglas R. Burdick, Alexandre V. Evfimievski, Berthold Rein- wald, Frederick R. Reiss, Prithviraj Sen, Shirish Tatikonda, and Yuanyuan Tian

  13. [13]

    Bull.37, 3 (2014), 52–62

    SystemML’s Optimizer: Plan Generation for Large-Scale Machine Learn- ing Programs.Data Eng. Bull.37, 3 (2014), 52–62. http://sites.computer.org/ debull/A14sept/p52.pdf

    2014

  14. [14]

    Nicolas Bruno and Surajit Chaudhuri. 2006. To Tune or not to Tune? A Light- weight Physical Design Alerter. InVLDB. 499–510. http://dl.acm.org/citation. cfm?id=1164171

    2006

  15. [15]

    Sunil Chakkappen, Shreya Kunjibettu, Daniel Mcgreer, Masoomeh Kishi, Hong Su, Mohamed Ziauddin, Mohamed Zaït, Zhan Li, and Yuying Zhang. 2025. Automatic Indexing in Oracle.PVLDB18, 12 (2025), 4924–4937. https://doi.org/ 10.14778/3750601.3750616

    work page doi:10.14778/3750601.3750616 2025

  16. [16]

    Chamberlin, Morton M

    Donald D. Chamberlin, Morton M. Astrahan, Kapali P. Eswaran, Patricia P. Griffiths, Raymond A. Lorie, James W. Mehl, Phyllis Reisner, and Bradford W. Wade. 1976. SEQUEL 2: A Unified Approach to Data Definition, Manipulation, and Control.IBM J. Res. Dev.20, 6 (1976), 560–575. https://doi.org/10.1147/RD. 206.0560

    work page doi:10.1147/rd 1976

  17. [17]

    Chamberlin and Raymond F

    Donald D. Chamberlin and Raymond F. Boyce. 1974. SEQUEL: A Structured English Query Language. InSIGMOD. 249–264. https://doi.org/10.1145/800296. 811515

    work page doi:10.1145/800296 1974

  18. [18]

    Ioannidis

    Konstantinos Chasialis, Yannis Foufoulas, Alkis Simitsis, and Yannis E. Ioannidis

  19. [19]

    Optimizing UDF Queries in SQL Data Engines. InEDBT. https://doi.org/ 10.48786/EDBT.2026.21

    work page doi:10.48786/edbt.2026.21 2026

  20. [20]

    Surajit Chaudhuri and Kyuseok Shim. 1994. Including Group-By in Query Optimization. InVLDB. 354–366. http://www.vldb.org/conf/1994/P354.PDF

    1994

  21. [21]

    Transaction Processing Council. 2025. https://www.tpc.org/tpch

    2025

  22. [22]

    Bailu Ding, Surajit Chaudhuri, Johannes Gehrke, and Vivek Narasayya. 2021. DSB: A decision support benchmark for workload-driven and traditional data- base systems.PVLDB14, 13 (2021), 3376–3388. https://doi.org/10.14778/3484224. 3484234

    work page doi:10.14778/3484224 2021

  23. [23]

    Haoran Ding, Zhaoguo Wang, Yicun Yang, Dexin Zhang, Zhenglin Xu, Haibo Chen, Ruzica Piskac, and Jinyang Li. 2023. Proving Query Equivalence Using Linear Integer Arithmetic.Proc. ACM Manag. Data1, 4 (2023), 227:1–227:26. https://doi.org/10.1145/3626768

    work page doi:10.1145/3626768 2023

  24. [24]

    Jens Dittrich and Joris Nix. 2020. The Case for Deep Query Optimisation. In CIDR. http://cidrdb.org/cidr2020/papers/p3-dittrich-cidr20.pdf

    2020

  25. [25]

    Jens Dittrich, Joris Nix, and Christian Schön. 2021. The next 50 Years in Database Indexing or: The Case for Automatically Generated Index Structures.PVLDB 15, 3 (2021), 527–540. https://doi.org/10.14778/3494124.3494136

    work page doi:10.14778/3494124.3494136 2021

  26. [26]

    Rui Dong, Jie Liu, Yuxuan Zhu, Cong Yan, Barzan Mozafari, and Xinyu Wang

  27. [27]

    VLDB Endow.16, 11 (2023), 3151–3164

    SlabCity: Whole-Query Optimization Using Program Synthesis.PVLDB 16, 11 (2023), 3151–3164. https://doi.org/10.14778/3611479.3611515

    work page doi:10.14778/3611479.3611515 2023

  28. [28]

    Markus Dreseler, Martin Boissier, Tilmann Rabl, and Matthias Uflacker. 2020. Quantifying TPC-H Choke Points and Their Optimizations.VLDB13, 8 (2020), 1206–1220. https://doi.org/10.14778/3389133.3389138

    work page doi:10.14778/3389133.3389138 2020

  29. [29]

    Evfimievski, Shirish Tatikonda, Berthold Reinwald, and Prithviraj Sen

    Tarek Elgamal, Shangyu Luo, Matthias Boehm, Alexandre V. Evfimievski, Shirish Tatikonda, Berthold Reinwald, and Prithviraj Sen. 2017. SPOOF: Sum-Product Optimization and Operator Fusion for Large-Scale Machine Learning. InCIDR. http://cidrdb.org/cidr2017/papers/p3-elgamal-cidr17.pdf

    2017

  30. [30]

    Jingzhi Fang, Yanyan Shen, Yue Wang, and Lei Chen. 2020. Optimizing DNN Computation Graph using Graph Substitutions.PVLDB13, 11 (2020), 2734–2746. http://www.vldb.org/pvldb/vol13/p2734-fang.pdf

    2020

  31. [31]

    Amela Fejza, Pierre Genevès, and Nabil Layaïda. 2024. Efficient Enumeration of Recursive Plans in Transformation-based Query Optimizers.PVLDB17, 11 (2024), 3095–3108. https://doi.org/10.14778/3681954.3681986

    work page doi:10.14778/3681954.3681986 2024

  32. [32]

    Philipp Fent, Altan Birler, and Thomas Neumann. 2023. Practical planning and execution of groupjoin and nested aggregates.VLDB J.32, 6 (2023), 1165–1190. https://doi.org/10.1007/S00778-022-00765-X

    work page doi:10.1007/s00778-022-00765-x 2023

  33. [33]

    Sofoklis Floratos, Ahmad Ghazal, Jason Sun, Jianjun Chen, and Xiaodong Zhang

  34. [34]

    DBSpinner: Making a Case for Iterative Processing in Databases. InICDE. 2399–2410. https://doi.org/10.1109/ICDE51399.2021.00273

    work page doi:10.1109/icde51399.2021.00273 2021

  35. [35]

    Yannis Foufoulas and Alkis Simitsis. 2023. Efficient Execution of User-Defined Functions in SQL Queries.VLDB16, 12 (2023), 3874–3877. https://doi.org/10. 14778/3611540.3611574

    arXiv 2023

  36. [36]

    Freitag, Maximilian Bandle, Tobias Schmidt, Alfons Kemper, and Thomas Neumann

    Michael J. Freitag, Maximilian Bandle, Tobias Schmidt, Alfons Kemper, and Thomas Neumann. 2020. Adopting Worst-Case Optimal Joins in Relational Database Systems.VLDB13, 11 (2020), 1891–1904. http://www.vldb.org/pvldb/ vol13/p1891-freitag.pdf

    2020

  37. [37]

    Jonathan Fürst, Catherine Kosten, Farhad Nooralahzadeh, Yi Zhang, and Kurt Stockinger. 2025. Evaluating the Data Model Robustness of Text-to-SQL Systems Based on Real User Queries. InEDBT. https://doi.org/10.48786/EDBT.2025.13

    work page doi:10.48786/edbt.2025.13 2025

  38. [38]

    Rainer Gemulla, Philipp Rösch, and Wolfgang Lehner. 2008. Linked Bernoulli Synopses: Sampling along Foreign Keys. InSSDBM, Vol. 5069. 6–23. https: //doi.org/10.1007/978-3-540-69497-7_4

    work page doi:10.1007/978-3-540-69497-7_4 2008

  39. [39]

    Parke Godfrey, Jarek Gryz, Andrzej Hoppe, Wenbin Ma, and Calisto Zuzarte

  40. [40]

    Query Rewrites with Views for XML in DB2. InICDE. 1339–1350. https: //doi.org/10.1109/ICDE.2009.131

    work page doi:10.1109/icde.2009.131 2009

  41. [41]

    Google. 2025. https://gemini.google.com

    2025

  42. [42]

    Goetz Graefe. 1995. The Cascades Framework for Query Optimization.Data Eng. Bull.18, 3 (1995), 19–29. http://sites.computer.org/debull/95SEP-CD.pdf

    1995

  43. [43]

    Luca Gretscher and Jens Dittrich. 2025. How to Optimize SQL Queries? A Comparison Between Split, Holistic, and Hybrid Approaches.PVLDB18, 11 (2025), 3910–3922. https://doi.org/10.14778/3749646.3749663

    work page doi:10.14778/3749646.3749663 2025

  44. [44]

    CWI Database Architectures Group. 2025. https://github.com/cwida/public_bi_ benchmark

    2025

  45. [45]

    Immanuel Haffner and Jens Dittrich. 2023. A simplified Architecture for Fast, Adaptive Compilation and Execution of SQL Queries. InEDBT. https://doi.org/ 10.48786/EDBT.2023.01

    work page doi:10.48786/edbt.2023.01 2023

  46. [46]

    Yuxing Han et al. 2021. Cardinality Estimation in DBMS: A Comprehensive Benchmark Evaluation.PVLDB15, 4 (2021), 752–765. https://doi.org/10.14778/ 3503585.3503586

    arXiv 2021

  47. [47]

    Cinda Heeren, H. V. Jagadish, and Leonard Pitt. 2003. Optimal indexing using near-minimal space. InPODS. 244–251. https://doi.org/10.1145/773153.773177

    work page doi:10.1145/773153.773177 2003

  48. [48]

    Gerald Held, Michael Stonebraker, and Eugene Wong. 1975. INGRES: A Rela- tional Data Base System. InAFIPS, Vol. 44. 409–416. https://doi.org/10.1145/ 1499949.1500029

    arXiv 1975

  49. [49]

    Waddington, and Adam Mor- rison

    Moshik Hershcovitch, Artem Khyzha, Daniel G. Waddington, and Adam Mor- rison. 2022. Elastic Indexes: Dynamic Space vs. Query Efficiency Tuning for In-Memory Database Indexing. InEDBT. https://doi.org/10.48786/EDBT.2022.18

    work page doi:10.48786/edbt.2022.18 2022

  50. [50]

    Zezhou Huang, Rathijit Sen, Jiaxiang Liu, and Eugene Wu. 2023. JoinBoost: Grow Trees Over Normalized Data Using Only SQL.PVLDB16, 11 (2023), 3071–3084. https://doi.org/10.14778/3611479.3611509

    work page doi:10.14778/3611479.3611509 2023

  51. [51]

    Zhihao Jia, Oded Padon, James Thomas, Todd Warszawski, Matei Zaharia, and Alex Aiken. 2019. TASO: optimizing deep learning computation with automatic generation of graph substitutions. InSOSP. 47–62. https://doi.org/10.1145/ 3341301.3359630

    arXiv 2019

  52. [52]

    Zhihao Jia, James Thomas, Todd Warszawski, Mingyu Gao, Matei Zaharia, and Alex Aiken. 2019. Optimizing DNN Computation with Relaxed Graph Substitutions. InMLSys. https://proceedings.mlsys.org/paper_files/paper/2019/ hash/

    2019

  53. [53]

    David Justen, Daniel Ritter, Campbell Fraser, Andrew Lamb, Nga Tran, Allison Lee, Thomas Bodner, Mhd Yamen Haddad, Steffen Zeuch, Volker Markl, and Matthias Boehm. 2024. POLAR: Adaptive and Non-invasive Join Order Selection via Plans of Least Resistance.PVLDB17, 6 (2024), 1350–1363. https://doi.org/ 10.14778/3648160.3648175

    work page doi:10.14778/3648160.3648175 2024

  54. [54]

    George Katsogiannis-Meimarakis and Georgia Koutrika. 2021. A Deep Dive into Deep Learning Approaches for Text-to-SQL Systems. InSIGMOD. 2846–2851. https://doi.org/10.1145/3448016.3457543

    work page doi:10.1145/3448016.3457543 2021

  55. [55]

    Kester, Manos Athanassoulis, and Stratos Idreos

    Michael S. Kester, Manos Athanassoulis, and Stratos Idreos. 2017. Access Path Selection in Main-Memory Optimized Data Systems: Should I Scan or Should I Probe?. InSIGMOD. 715–730. https://doi.org/10.1145/3035918.3064049

    work page doi:10.1145/3035918.3064049 2017

  56. [56]

    Steffen Kläbe and Kai-Uwe Sattler. 2023. Patched Multi-Key Partitioning for Robust Query Performance. InEDBT. https://doi.org/10.48786/EDBT.2023.26

    work page doi:10.48786/edbt.2023.26 2023

  57. [57]

    Jan Kossmann, Thorsten Papenbrock, and Felix Naumann. 2023. Correction to: Data dependencies for query optimization: a survey.VLDB J.32, 2 (2023), 471. https://doi.org/10.1007/S00778-021-00710-4

    work page doi:10.1007/s00778-021-00710-4 2023

  58. [58]

    Warner, Vikas Arora, and Susan Kotsovolos

    Muralidhar Krishnaprasad, Zhen Hua Liu, Anand Manikutty, James W. Warner, Vikas Arora, and Susan Kotsovolos. 2004. Query Rewrite for XML in Oracle XML DB. InVLDB. 1122–1133. https://doi.org/10.1016/B978-012088469-8.50098-X

    work page doi:10.1016/b978-012088469-8.50098-x 2004

  59. [59]

    Boncz, Alfons Kemper, and Thomas Neumann

    Viktor Leis, Andrey Gubichev, Atanas Mirchev, Peter A. Boncz, Alfons Kemper, and Thomas Neumann. 2015. How Good Are Query Optimizers, Really?PVLDB 9, 3 (2015), 204–215. https://doi.org/10.14778/2850583.2850594

    work page doi:10.14778/2850583.2850594 2015

  60. [60]

    Viktor Leis, Kan Kundhikanjana, Alfons Kemper, and Thomas Neumann. 2015. Efficient Processing of Window Functions in Analytical SQL Queries.PVLDB8, 10 (2015), 1058–1069. https://doi.org/10.14778/2794367.2794375

    work page doi:10.14778/2794367.2794375 2015

  61. [61]

    Viktor Leis, Bernhard Radke, Andrey Gubichev, Alfons Kemper, and Thomas Neumann. 2017. Cardinality Estimation Done Right: Index-Based Join Sampling. InCIDR. http://cidrdb.org/cidr2017/papers/p9-leis-cidr17.pdf

    2017

  62. [62]

    Viktor Leis, Bernhard Radke, Andrey Gubichev, Atanas Mirchev, Peter Boncz, Alfons Kemper, and Thomas Neumann. 2018. Query optimization through the looking glass, and what we found running the Join Order Benchmark.VLDB J. 27, 5 (2018), 643–668. https://doi.org/10.1007/S00778-017-0480-7

    work page doi:10.1007/s00778-017-0480-7 2018

  63. [63]

    Zhaodonghui Li, Haitao Yuan, Huiming Wang, Gao Cong, and Lidong Bing

  64. [64]

    https://doi.org/10

    LLM-R2: A Large Language Model Enhanced Rule-Based Rewrite System for Boosting Query Efficiency.PVLDB18, 1 (2024), 53–65. https://doi.org/10. 14778/3696435.3696440

    arXiv 2024

  65. [65]

    Daniel Lindner, Daniel Ritter, and Felix Naumann. 2024. Enabling Data Dependency-based Query Optimization.CoRRabs/2406.06886 (2024). https: //doi.org/10.48550/ARXIV.2406.06886

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2406.06886 2024

  66. [66]

    Jie Liu and Barzan Mozafari. 2026. GenRewrite: Query Rewriting via Large Language Models.SIGMOD4, 1 (2026), 1–26. https://dl.acm.org/doi/pdf/10. 1145/3786684

    2026

  67. [67]

    Ilyas, Benny Kimelfeld, and Sudeepa Roy

    Ester Livshits, Rina Kochirgan, Segev Tsur, Ihab F. Ilyas, Benny Kimelfeld, and Sudeepa Roy. 2021. Properties of Inconsistency Measures for Databases. In SIGMOD. 1182–1194. https://doi.org/10.1145/3448016.3457310

    work page doi:10.1145/3448016.3457310 2021

  68. [68]

    Guy Lohman. 2002. The DB2 Universal Database Optimizer. Guest Lecture. https://cs.uwaterloo.ca/~ilyas/CS448W14/ibm.pdf

    2002

  69. [69]

    Guy Lohman. 2014. Is Query Optimization a “Solved” Problem? ACM SIGMOD Blog. https://wp.sigmod.org/?p=1075

    2014

  70. [70]

    Toby Mao. 2024. SQLGlot. GitHub repository. Retrieved April 29, 2026 from https://github.com/tobymao/sqlglot

    2024

  71. [71]

    Simmen, Guy M

    Volker Markl, Vijayshankar Raman, David E. Simmen, Guy M. Lohman, and Hamid Pirahesh. 2004. Robust Query Processing through Progressive Optimiza- tion. InSIGMOD. 659–670. https://doi.org/10.1145/1007568.1007642

    work page doi:10.1145/1007568.1007642 2004

  72. [72]

    Guido Moerkotte. 2025. Building Query Compilers. https://pi3.informatik.uni- mannheim.de/~moer/querycompiler.pdf Last Accessed: October 03, 2025

    2025

  73. [73]

    Guido Moerkotte and Thomas Neumann. 2011. Accelerating Queries with Group-By and Join by Groupjoin.PVLDB4, 11 (2011), 843–851. http://www. vldb.org/pvldb/vol4/p843-moerkotte.pdf

    2011

  74. [74]

    Magnus Müller, Lucas Woltmann, and Wolfgang Lehner. 2023. Enhanced Fea- turization of Queries with Mixed Combinations of Predicates for ML-based Cardinality Estimation. InEDBT. https://doi.org/10.48786/EDBT.2023.22

    work page doi:10.48786/edbt.2023.22 2023

  75. [75]

    Parimarjan Negi, Ryan Marcus, Andreas Kipf, Hongzi Mao, Nesime Tatbul, Tim Kraska, and Mohammad Alizadeh. 2021. Flow-Loss: Learning Cardinality Estimates That Matter.PVLDB14, 11 (jul 2021), 2019–2032. https://doi.org/10. 14778/3476249.3476259

    arXiv 2021

  76. [76]

    Thomas Neumann and Alfons Kemper. 2015. Unnesting Arbitrary Queries. In BTW, Vol. P-241. 383–402. https://dl.gi.de/handle/20.500.12116/2418

    2015

  77. [77]

    Thomas Neumann and Viktor Leis. 2024. A Critique of Modern SQL and a Proposal Towards a Simple and Expressive Query Language. InCIDR. https: //www.cidrdb.org/cidr2024/papers/p48-neumann.pdf

    2024

  78. [78]

    Oleksii Vasyliev. 2025. https://pgtune.leopard.in.ua

    2025

  79. [79]

    Amy Ousterhout, Steven McCanne, Henri Dubois-Ferrière, Silvery Fu, Sylvia Ratnasamy, and Noah Treuhaft. 2023. Zed: Leveraging Data Types to Pro- cess Eclectic Data. InCIDR. https://www.cidrdb.org/cidr2023/papers/p52- ousterhout.pdf

    2023

  80. [80]

    Simone Papicchio, Paolo Papotti, and Luca Cagliero. 2023. QATCH: Benchmarking SQL-centric tasks with Table Representation Learn- ing Models on Your Data. InNeurIPS. http://papers.nips.cc/paper_ files/paper/2023/hash/62a24b69b820d30e9e5ad4f15ff7bf72-Abstract- Datasets_and_Benchmarks.html

    2023

Showing first 80 references.