graphguard-anon commited on
Commit
6c33148
·
verified ·
1 Parent(s): 162e785

Add anonymized benchmark datasets and constraint suites for paper evaluation

Browse files
README.md ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - graph-machine-learning
5
+ - relational-databases
6
+ - knowledge-graphs
7
+ tags:
8
+ - database-integrity
9
+ - automated-reasoning
10
+ - first-order-logic
11
+ - craig-interpolation
12
+ - formal-verification
13
+ - sigmod-reproducibility
14
+ size_categories:
15
+ - 10GB<n<100GB
16
+ ---
17
+
18
+ # GraphGuard Benchmark Suite (Anonymized Submission)
19
+
20
+ This repository contains the complete cross-paradigm benchmark suites, schemas, constraints, and data instances for **GraphGuard**, an automated first-order theorem proving and Craig interpolation framework for database integrity verification.
21
+
22
+ ## Repository Contents
23
+
24
+ The repository is modularly structured into tier-based packages:
25
+
26
+ ### 1. Constraints Suite (`constraints.zip` ~0.5 MB)
27
+ Contains all integrity constraints evaluated across the three paradigms:
28
+ - **Property Graphs:** Mined and hand-crafted Graph Generating Dependencies (GGDs) for **CORDIS**, **DBLP**, **GDELT**, and **LDBC SNB**.
29
+ - **RDF Knowledge Graphs:** W3C SHACL shape schemas and OWL ontologies for **LUBM** and **WatDiv**.
30
+ - **Relational Schemas:** Relational dependencies (foreign keys, uniqueness EGDs, check constraints) for **TPC-H** and all **ChaseBench** scenarios (`doctors`, `doctors-fd`, `LUBM`, `Ontology-256`, `STB-128`, `deep`, and `correctness` suites).
31
+
32
+ ### 2. Standard Evaluation Benchmarks (`graph_benchmarks.zip`, `rdf_benchmarks.zip`, `relational_benchmarks.zip`)
33
+ - **Property Graphs (`graph_benchmarks.zip`):** Ready-to-evaluate node and edge CSV files for Neo4j and localized r-hop neighborhood extraction (`cordis_small`, `dblp_small`, `gdelt_small`, `ldbc_small_*`).
34
+ - **RDF Knowledge Graphs (`rdf_benchmarks.zip`):** LUBM multi-university instances (1 to 15 universities in OWL/RDF), WatDiv 100k triple sample (`watdiv_100k.nt`), and UBA synthetic data generator.
35
+ - **Relational Databases (`relational_benchmarks.zip`):** ChaseBench relational scenarios with tables, schemas, queries, and repair tools, alongside TPC-H decision-support tables at scale factors 0.01 to 0.1.
36
+
37
+ ### 3. Full-Scale Datasets (Optional, for heavy data scaling)
38
+ - `ldbc_scales.zip`: LDBC SNB SF 1 and SF 3 instances.
39
+ - `watdiv_10m.zip`: WatDiv full 10M triples dataset (`watdiv.10M.nt`).
40
+ - `dblp_raw.zip`: Complete DBLP publication XML database (`dblp.xml`, `dblp.dtd`).
41
+ - `tpch_scales.zip`: TPC-H tables at SF 0.3, SF 1, SF 3, and SF 10.
42
+
43
+ ---
44
+
45
+ ## Quick Download & Setup
46
+
47
+ You can download and extract these datasets directly into the codebase using the automated downloader script:
48
+
49
+ ```bash
50
+ # Download standard evaluation bundle (constraints + benchmark datasets):
51
+ python scripts/download_benchmarks.py
52
+
53
+ # Download only constraints (ultra-lightweight, ~300 KB):
54
+ python scripts/download_benchmarks.py --constraints-only
55
+
56
+ # Download specific datasets:
57
+ python scripts/download_benchmarks.py --dataset ldbc cordis
58
+ ```
59
+
60
+ All archives extract into the `datasets/` root directory, maintaining the exact folder paths expected by the evaluation pipelines (`run_all_benchmarks.py`, `run_eval.sh`, `run_chasebench_eval.sh`, and `run_lubm_scaling.sh`).
61
+
62
+ ---
63
+
64
+ ## Double-Blind Anonymity Compliance
65
+ This repository and its contents are fully anonymized for double-blind peer review. All local paths, personal metadata, and author identifications have been scrubbed in accordance with SIGMOD review guidelines.
constraints.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03d45fa5992f0f9c09e6955ad6ce2f0e4db59e5f74ad71a1c9dec80895c53981
3
+ size 421478
dblp_raw.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76045d1a1dac2c91e6b1b86fb991eab99328df4672a1587042afa5c8421588be
3
+ size 1056396055
graph_benchmarks.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61278b579fdb8f97f6775629e1893e2bd0b7f0c3fc05f73c554da4d3d88d819d
3
+ size 170478403
ldbc_scales.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84699564471a0259a102db26a27fb76b89cc19e3555ac9b8f7d7144e76585367
3
+ size 2255841826
manifest.json ADDED
@@ -0,0 +1,94 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "repository": "graphguard-benchmarks",
3
+ "description": "Anonymized GraphGuard benchmark and constraint artifacts",
4
+ "packages": {
5
+ "constraints": {
6
+ "filename": "constraints.zip",
7
+ "tier": "constraints",
8
+ "description": "Graph Generating Dependencies (GGDs), SHACL shapes, Relational rules, and ChaseBench dependencies",
9
+ "file_count": 519,
10
+ "uncompressed_bytes": 1665545,
11
+ "uncompressed_mb": 1.59,
12
+ "compressed_bytes": 421478,
13
+ "compressed_mb": 0.4,
14
+ "sha256": "03d45fa5992f0f9c09e6955ad6ce2f0e4db59e5f74ad71a1c9dec80895c53981"
15
+ },
16
+ "graph_benchmarks": {
17
+ "filename": "graph_benchmarks.zip",
18
+ "tier": "benchmark",
19
+ "description": "Ready-to-use extracted Property Graph CSVs for CORDIS, DBLP, GDELT, and LDBC (SF 0.003 - 0.3)",
20
+ "file_count": 14,
21
+ "uncompressed_bytes": 976340967,
22
+ "uncompressed_mb": 931.11,
23
+ "compressed_bytes": 170478403,
24
+ "compressed_mb": 162.58,
25
+ "sha256": "61278b579fdb8f97f6775629e1893e2bd0b7f0c3fc05f73c554da4d3d88d819d"
26
+ },
27
+ "rdf_benchmarks": {
28
+ "filename": "rdf_benchmarks.zip",
29
+ "tier": "benchmark",
30
+ "description": "RDF Knowledge Graphs: LUBM 1-15 universities OWL/CSVs, WatDiv 100k triples, and UBA generator",
31
+ "file_count": 38,
32
+ "uncompressed_bytes": 35259307,
33
+ "uncompressed_mb": 33.63,
34
+ "compressed_bytes": 2323979,
35
+ "compressed_mb": 2.22,
36
+ "sha256": "50de80675d6b93b3b4b25028dd3c6630673c2e0b36483db95a8a9866d0d3c995"
37
+ },
38
+ "relational_benchmarks": {
39
+ "filename": "relational_benchmarks.zip",
40
+ "tier": "benchmark",
41
+ "description": "Relational benchmark data: ChaseBench scenarios (data, schemas, queries, tools) & TPC-H SF 0.01 - 0.1",
42
+ "file_count": 3564,
43
+ "uncompressed_bytes": 1416173130,
44
+ "uncompressed_mb": 1350.57,
45
+ "compressed_bytes": 977490624,
46
+ "compressed_mb": 932.21,
47
+ "sha256": "ed2a2cb7b887f27e1112f982cb9c2de8a5933a076bc4efa27fe6ca94e3b9c73e"
48
+ },
49
+ "ldbc_scales": {
50
+ "filename": "ldbc_scales.zip",
51
+ "tier": "full",
52
+ "description": "LDBC SNB larger scale factor subgraphs (ldbc_small_1, ldbc_small_3) and raw graph CSV dumps",
53
+ "file_count": 590,
54
+ "uncompressed_bytes": 10153498134,
55
+ "uncompressed_mb": 9683.13,
56
+ "compressed_bytes": 2255841826,
57
+ "compressed_mb": 2151.34,
58
+ "sha256": "84699564471a0259a102db26a27fb76b89cc19e3555ac9b8f7d7144e76585367"
59
+ },
60
+ "watdiv_10m": {
61
+ "filename": "watdiv_10m.zip",
62
+ "tier": "full",
63
+ "description": "WatDiv full 10M triples dataset (watdiv.10M.nt)",
64
+ "file_count": 2,
65
+ "uncompressed_bytes": 1548944211,
66
+ "uncompressed_mb": 1477.19,
67
+ "compressed_bytes": 88186076,
68
+ "compressed_mb": 84.1,
69
+ "sha256": "469d608225236e5ade76828438ffb142f739206a9d42c824941784bee29c071f"
70
+ },
71
+ "dblp_raw": {
72
+ "filename": "dblp_raw.zip",
73
+ "tier": "full",
74
+ "description": "DBLP complete XML dump and DTD specification",
75
+ "file_count": 2,
76
+ "uncompressed_bytes": 5138736228,
77
+ "uncompressed_mb": 4900.68,
78
+ "compressed_bytes": 1056396055,
79
+ "compressed_mb": 1007.46,
80
+ "sha256": "76045d1a1dac2c91e6b1b86fb991eab99328df4672a1587042afa5c8421588be"
81
+ },
82
+ "tpch_scales": {
83
+ "filename": "tpch_scales.zip",
84
+ "tier": "full",
85
+ "description": "TPC-H larger scale factor tables (sf_0.3, sf_1, sf_3, sf_10)",
86
+ "file_count": 32,
87
+ "uncompressed_bytes": 16169411720,
88
+ "uncompressed_mb": 15420.35,
89
+ "compressed_bytes": 4631644870,
90
+ "compressed_mb": 4417.08,
91
+ "sha256": "d5e04814f2a8a03c2e62075e55cf423248c693785d54137c41126b2161bfd38f"
92
+ }
93
+ }
94
+ }
rdf_benchmarks.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50de80675d6b93b3b4b25028dd3c6630673c2e0b36483db95a8a9866d0d3c995
3
+ size 2323979
relational_benchmarks.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ed2a2cb7b887f27e1112f982cb9c2de8a5933a076bc4efa27fe6ca94e3b9c73e
3
+ size 977490624
tpch_scales.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5e04814f2a8a03c2e62075e55cf423248c693785d54137c41126b2161bfd38f
3
+ size 4631644870
watdiv_10m.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:469d608225236e5ade76828438ffb142f739206a9d42c824941784bee29c071f
3
+ size 88186076