scan-v5.json (27450B)
1 { 2 "scan_version": 5, 3 "paper_type": "empirical", 4 "paper": { 5 "title": "Gamma: Revisiting Template-Based Automated Program Repair Via Mask Prediction", 6 "authors": [ 7 "Quanjun Zhang", 8 "Bowen Yu", 9 "Chunrong Fang", 10 "Weisong Sun", 11 "Tongke Zhang", 12 "Zhenyu Chen" 13 ], 14 "year": 2023, 15 "venue": "International Conference on Automated Software Engineering", 16 "arxiv_id": "2309.09308", 17 "doi": "10.1109/ASE56229.2023.00063" 18 }, 19 "checklist": { 20 "claims_and_evidence": { 21 "abstract_claims_supported": { 22 "applies": true, 23 "answer": true, 24 "justification": "All quantitative abstract claims (82 bugs on Defects4J-v1.2, 20.59%/26.15% improvement over TBar/Recoder, 45 bugs on Defects4J-v2.0, 22 on QuixBugs) are confirmed by Table I and Table II in the paper.", 25 "source": "haiku" 26 }, 27 "causal_claims_justified": { 28 "applies": true, 29 "answer": false, 30 "justification": "The paper claims mask prediction via pre-trained models 'causes' improved donor code quality, but this is only supported by comparison to baselines — no ablation removes the PLM component while keeping all else equal to isolate the causal mechanism.", 31 "source": "haiku" 32 }, 33 "generalization_bounded": { 34 "applies": true, 35 "answer": true, 36 "justification": "The paper bounds generalization to Java programs, acknowledges templates were designed for Defects4J-v1.2 patterns, and explicitly notes that 'some templates cannot be applied to any bugs except those from Defects4J-v1.2.'", 37 "source": "haiku" 38 }, 39 "alternative_explanations_discussed": { 40 "applies": true, 41 "answer": false, 42 "justification": "The paper does not discuss alternative explanations for GAMMA's superior performance, such as whether the larger effective search space (not PLM reasoning per se) drives gains, or whether beam size differences vs. baselines explain results.", 43 "source": "haiku" 44 }, 45 "proxy_outcome_distinction": { 46 "applies": true, 47 "answer": true, 48 "justification": "The paper clearly distinguishes 'plausible patches' (pass test suite) from 'correct patches' (semantically equivalent to developer patch by manual inspection), precisely matching claims to the measured outcome.", 49 "source": "haiku" 50 } 51 }, 52 "limitations_and_scope": { 53 "limitations_section_present": { 54 "applies": true, 55 "answer": true, 56 "justification": "Section VI 'Threats to Validity' provides a dedicated limitations section with three distinct threats: manual inspection bias, perfect fault localization, and data leakage.", 57 "source": "haiku" 58 }, 59 "threats_to_validity_specific": { 60 "applies": true, 61 "answer": true, 62 "justification": "Threats are specific: names the three leaked bugs (Closure-73, Closure-126, Time-19) and conducts a perturbation test; discusses exact fault localization bias problem with named prior work [57].", 63 "source": "haiku" 64 }, 65 "scope_boundaries_stated": { 66 "applies": true, 67 "answer": true, 68 "justification": "The paper explicitly states scope boundaries: Java-only, 13 fix templates from TBar, evaluation under perfect fault localization only, and acknowledges that perfect localization 'is usually unavailable in real practice.'", 69 "source": "haiku" 70 } 71 }, 72 "conflicts_of_interest": { 73 "funding_disclosed": { 74 "applies": true, 75 "answer": true, 76 "justification": "Acknowledgment section states 'This work is supported partially by the National Natural Science Foundation of China (61932012, 62141215).'", 77 "source": "haiku" 78 }, 79 "affiliations_disclosed": { 80 "applies": true, 81 "answer": true, 82 "justification": "All six authors are identified as affiliated with Nanjing University's State Key Laboratory for Novel Software Technology.", 83 "source": "haiku" 84 }, 85 "funder_independent_of_outcome": { 86 "applies": true, 87 "answer": true, 88 "justification": "NSFC is a Chinese government funding agency with no commercial interest in the APR tool being evaluated.", 89 "source": "haiku" 90 }, 91 "financial_interests_declared": { 92 "applies": true, 93 "answer": false, 94 "justification": "No competing interests statement or declaration of financial interests (patents, equity, consulting) appears anywhere in the paper.", 95 "source": "haiku" 96 } 97 }, 98 "scope_and_framing": { 99 "key_terms_defined": { 100 "applies": true, 101 "answer": true, 102 "justification": "Key terms 'fix patterns,' 'donor code,' 'plausible patch,' and 'correct patch' are all explicitly defined in Section II-A and Section IV-D, with clear distinctions between them.", 103 "source": "haiku" 104 }, 105 "intended_contribution_clear": { 106 "applies": true, 107 "answer": true, 108 "justification": "Four contributions are explicitly enumerated in Section I: New Dimension (bridging PLMs and template APR), Novel APR tool (GAMMA itself), Extensive study (empirical evaluation), and Available artifacts (GitHub release).", 109 "source": "haiku" 110 }, 111 "engagement_with_prior_work": { 112 "applies": true, 113 "answer": true, 114 "justification": "Section VII provides a detailed related work section that explicitly positions GAMMA against TBar, AlphaRepair, learning-based APR, and pre-trained models, explaining how GAMMA differs from each category.", 115 "source": "haiku" 116 } 117 } 118 }, 119 "type_checklist": { 120 "empirical": { 121 "artifacts": { 122 "code_released": { 123 "applies": true, 124 "answer": true, 125 "justification": "Source code is released at https://github.com/iSEngLab/GAMMA [29], described as containing 'source code, experimental results, and correct patches.'", 126 "source": "haiku" 127 }, 128 "data_released": { 129 "applies": true, 130 "answer": true, 131 "justification": "All three evaluation benchmarks (Defects4J-v1.2, Defects4J-v2.0, QuixBugs) are publicly available standard benchmarks with citations provided.", 132 "source": "haiku" 133 }, 134 "environment_specified": { 135 "applies": true, 136 "answer": false, 137 "justification": "The paper mentions Ubuntu 18.04.3 and two Tesla V100-SXM2 GPUs but provides no requirements.txt, Dockerfile, or dependency specification for reproducing the software environment.", 138 "source": "haiku" 139 }, 140 "reproduction_instructions": { 141 "applies": true, 142 "answer": false, 143 "justification": "Section IV-E provides implementation details (Eclipse JDT, beam size 250, 5-hour limit) but no step-by-step reproduction instructions; a reader would need to infer the full pipeline from the paper description.", 144 "source": "haiku" 145 } 146 }, 147 "statistical_methodology": { 148 "confidence_intervals_or_error_bars": { 149 "applies": true, 150 "answer": false, 151 "justification": "All results in Table I and II are raw counts of bugs fixed with no confidence intervals, error bars, or uncertainty estimates reported.", 152 "source": "haiku" 153 }, 154 "significance_tests": { 155 "applies": true, 156 "answer": false, 157 "justification": "No statistical significance tests are applied to any comparative claim; all comparisons are presented as raw counts without hypothesis testing.", 158 "source": "haiku" 159 }, 160 "effect_sizes_reported": { 161 "applies": true, 162 "answer": true, 163 "justification": "Percentage improvements are reported in context (e.g., '20.59% (14 bugs)' over TBar, '26.15% (17 bugs)' over Recoder) with baseline counts provided for reference.", 164 "source": "haiku" 165 }, 166 "sample_size_justified": { 167 "applies": true, 168 "answer": false, 169 "justification": "The 395-bug Defects4J-v1.2 benchmark is adopted as a community standard without any power analysis or discussion of whether it is adequately sized for the comparisons made.", 170 "source": "haiku" 171 }, 172 "variance_reported": { 173 "applies": true, 174 "answer": false, 175 "justification": "Results are single-run counts of fixed bugs with no variance, standard deviation, or repeated run statistics reported for any technique.", 176 "source": "haiku" 177 } 178 }, 179 "evaluation_design": { 180 "baselines_included": { 181 "applies": true, 182 "answer": true, 183 "justification": "Nine baselines spanning both traditional (TBar, PraPR) and learning-based (AlphaRepair, Recoder, CURE, CoCoNuT, CIRCLE, DLFix, SequenceR) APR approaches are included.", 184 "source": "haiku" 185 }, 186 "baselines_contemporary": { 187 "applies": true, 188 "answer": true, 189 "justification": "The most relevant baselines are recent: AlphaRepair (FSE'22), CIRCLE (ISSTA'22), Recoder (ESEC/FSE'21), CURE (ICSE'21) — all published within 2 years of the paper.", 190 "source": "haiku" 191 }, 192 "ablation_study": { 193 "applies": true, 194 "answer": false, 195 "justification": "RQ3 compares different PLMs but does not ablate GAMMA's components (e.g., removing context comment, removing template selection); the paper lacks a true component-wise ablation.", 196 "source": "haiku" 197 }, 198 "multiple_metrics": { 199 "applies": true, 200 "answer": true, 201 "justification": "Both plausible patches and correct patches are reported, plus precision (correct/plausible ratio, e.g., 81.19% for GAMMA vs. 68/95 for TBar), constituting multiple evaluation dimensions.", 202 "source": "haiku" 203 }, 204 "human_evaluation": { 205 "applies": true, 206 "answer": true, 207 "justification": "Three authors independently manually inspect all plausible patches against developer ground truth, requiring agreement from all three to classify a patch as correct.", 208 "source": "haiku" 209 }, 210 "held_out_test_set": { 211 "applies": true, 212 "answer": true, 213 "justification": "Defects4J-v2.0 and QuixBugs serve as held-out evaluation benchmarks beyond the primary Defects4J-v1.2, explicitly testing generalizability to unseen bugs.", 214 "source": "haiku" 215 }, 216 "per_category_breakdown": { 217 "applies": true, 218 "answer": true, 219 "justification": "Table I provides per-project breakdown (Chart, Closure, Lang, Math, Mockito, Time) for all techniques on Defects4J-v1.2.", 220 "source": "haiku" 221 }, 222 "failure_cases_discussed": { 223 "applies": true, 224 "answer": false, 225 "justification": "The paper provides motivating examples of bugs GAMMA fixes that TBar cannot, but does not show or analyze specific cases where GAMMA generates plausible-but-incorrect patches or fails entirely.", 226 "source": "haiku" 227 }, 228 "negative_results_reported": { 229 "applies": true, 230 "answer": true, 231 "justification": "GAMMA with ChatGPT fixes only 67 bugs (vs. 82 with UniXcoder) and GAMMA fixes fewer QuixBugs (22) than AlphaRepair (28); these weaker results are reported without suppression.", 232 "source": "haiku" 233 } 234 }, 235 "setup_transparency": { 236 "model_versions_specified": { 237 "applies": true, 238 "answer": true, 239 "justification": "UniXcoder 'unixcoder-base' and ChatGPT 'gpt-3.5-turbo-0301' are explicitly versioned; CodeBERT is referenced by name from its original paper without a specific checkpoint identifier.", 240 "source": "haiku" 241 }, 242 "prompts_provided": { 243 "applies": true, 244 "answer": true, 245 "justification": "The full ChatGPT prompt is quoted verbatim: 'Next token prediction task, the first line is a comment to help prediction, just return 250 possible predictions for <mask> with highest probability:'; the UniXcoder input format is also described.", 246 "source": "haiku" 247 }, 248 "hyperparameters_reported": { 249 "applies": true, 250 "answer": true, 251 "justification": "Beam size 250, 5-hour running time limit, and the UniXcoder encoder-decoder mode are reported in Section IV-E; mask token count range 1-20 for CodeBERT is also specified.", 252 "source": "haiku" 253 }, 254 "scaffolding_described": { 255 "applies": true, 256 "answer": true, 257 "justification": "The full GAMMA pipeline (fault localization → template selection via AST → mask prediction → patch validation) is described in detail in Section III with pseudocode-level template definitions.", 258 "source": "haiku" 259 }, 260 "data_preprocessing_documented": { 261 "applies": true, 262 "answer": true, 263 "justification": "Section III-C describes preprocessing: extracting the containing method, prepending the original buggy line as a comment, and forming the final input to UniXcoder.", 264 "source": "haiku" 265 } 266 }, 267 "data_integrity": { 268 "raw_data_available": { 269 "applies": true, 270 "answer": true, 271 "justification": "The GitHub repository [29] contains 'experimental results and correct patches'; the benchmark bug data (Defects4J, QuixBugs) is publicly available from their original sources.", 272 "source": "haiku" 273 }, 274 "data_collection_described": { 275 "applies": true, 276 "answer": true, 277 "justification": "Benchmarks are described with citations and key properties (Defects4J-v1.2: 395 bugs from 6 projects; Defects4J-v2.0: 420 bugs from 17 projects; QuixBugs: 40 single-line algorithm bugs). Subset selection criteria for Defects4J-v2.0 (single-location patches only) are stated.", 278 "source": "haiku" 279 }, 280 "recruitment_methods_described": { 281 "applies": false, 282 "answer": false, 283 "justification": "NA — the study uses standard software engineering benchmarks with no human participant recruitment.", 284 "source": "haiku" 285 }, 286 "data_pipeline_documented": { 287 "applies": true, 288 "answer": true, 289 "justification": "The complete pipeline from input buggy program through fault localization, template selection, mask prediction, and patch validation is documented in Section III with Fig. 1 illustrating the full workflow.", 290 "source": "haiku" 291 } 292 }, 293 "contamination": { 294 "training_cutoff_stated": { 295 "applies": true, 296 "answer": false, 297 "justification": "The paper describes UniXcoder's training data (CodeSearchNet: 2.3M NL-PL pairs + 4.1M unimodal code) but states no training data cutoff date for any of the three PLMs evaluated.", 298 "source": "haiku" 299 }, 300 "train_test_overlap_discussed": { 301 "applies": true, 302 "answer": true, 303 "justification": "Section VI explicitly investigates train-test overlap, identifying three leaked bugs (Closure-73, Closure-126, Time-19) via manual query of CodeSearchNet, and performs perturbation tests to verify robustness.", 304 "source": "haiku" 305 }, 306 "benchmark_contamination_addressed": { 307 "applies": true, 308 "answer": true, 309 "justification": "The paper addresses this directly: even excluding the 3 leaked bugs, GAMMA still outperforms TBar (79 vs. 68) and AlphaRepair (79 vs. 74) on Defects4J-v1.2.", 310 "source": "haiku" 311 } 312 }, 313 "human_studies": { 314 "pre_registered": { 315 "applies": false, 316 "answer": false, 317 "justification": "NA — no human participant study conducted.", 318 "source": "haiku" 319 }, 320 "irb_or_ethics_approval": { 321 "applies": false, 322 "answer": false, 323 "justification": "NA — no human participants.", 324 "source": "haiku" 325 }, 326 "demographics_reported": { 327 "applies": false, 328 "answer": false, 329 "justification": "NA — no human participants.", 330 "source": "haiku" 331 }, 332 "inclusion_exclusion_criteria": { 333 "applies": false, 334 "answer": false, 335 "justification": "NA — no human participants.", 336 "source": "haiku" 337 }, 338 "randomization_described": { 339 "applies": false, 340 "answer": false, 341 "justification": "NA — no human participants.", 342 "source": "haiku" 343 }, 344 "blinding_described": { 345 "applies": false, 346 "answer": false, 347 "justification": "NA — no human participants.", 348 "source": "haiku" 349 }, 350 "attrition_reported": { 351 "applies": false, 352 "answer": false, 353 "justification": "NA — no human participants.", 354 "source": "haiku" 355 } 356 }, 357 "cost_and_practicality": { 358 "inference_cost_reported": { 359 "applies": true, 360 "answer": false, 361 "justification": "A 5-hour per-bug time limit is mentioned and CodeBERT is noted as slower than UniXcoder, but no systematic per-bug inference time, GPU-hours, or monetary cost is reported.", 362 "source": "haiku" 363 }, 364 "compute_budget_stated": { 365 "applies": true, 366 "answer": false, 367 "justification": "Hardware is specified (two Tesla V100-SXM2 GPUs, Ubuntu 18.04.3) but total compute time or GPU-hours for the full experimental runs is not stated.", 368 "source": "haiku" 369 } 370 } 371 } 372 }, 373 "claims": [ 374 { 375 "claim": "GAMMA correctly repairs 82 bugs on Defects4J-v1.2, outperforming all prior APR approaches including TBar (68) and AlphaRepair (74).", 376 "evidence": "Table I shows GAMMA at 82/101 plausible patches, highest among all 10 compared approaches across all six projects.", 377 "supported": "strong" 378 }, 379 { 380 "claim": "GAMMA achieves 81.19% precision (correct/plausible ratio), reducing patch overfitting compared to TBar (71.58%) and Recoder (58.04%).", 381 "evidence": "Table I reports plausible and correct counts for all tools, and precision is computed and explicitly compared in Section V-A.", 382 "supported": "strong" 383 }, 384 { 385 "claim": "Using pre-trained LMs for donor code prediction resolves the core limitation of template-based APR (incorrect donor code selection).", 386 "evidence": "Case studies in Listings 3-4 demonstrate specific bugs where TBar fails due to donor code limitations but GAMMA succeeds; comparison to TBar baseline supports the mechanism.", 387 "supported": "moderate" 388 }, 389 { 390 "claim": "GAMMA generalizes to unseen benchmarks: 45 bugs fixed on Defects4J-v2.0 and 22 on QuixBugs, outperforming all baselines.", 391 "evidence": "Table II confirms 45 (vs. 36 for AlphaRepair) and 22 (vs. 28 for AlphaRepair, 26 for CURE) results directly.", 392 "supported": "strong" 393 }, 394 { 395 "claim": "GAMMA is scalable to multiple pre-trained models: CodeBERT achieves 80 correct fixes, ChatGPT achieves 67 on Defects4J-v1.2.", 396 "evidence": "Figure 3 shows per-model results with overlap analysis; results are lower but substantial for all three models.", 397 "supported": "strong" 398 }, 399 { 400 "claim": "Data leakage from pre-training datasets does not drive GAMMA's performance advantage.", 401 "evidence": "Only 3 leaked bugs found in CodeSearchNet; excluding them yields 79 correct fixes, still exceeding TBar (68) and AlphaRepair (74).", 402 "supported": "moderate" 403 } 404 ], 405 "methodology_tags": [ 406 "benchmark-eval" 407 ], 408 "key_findings": "GAMMA integrates pre-trained language models (UniXcoder) into template-based APR by replacing the donor code retrieval step with a mask prediction task, fixing 82 bugs on Defects4J-v1.2 and outperforming all nine baseline APR approaches. The approach achieves 81.19% patch precision, substantially reducing overfitting compared to prior template-based and learning-based techniques. GAMMA generalizes across three benchmarks (Defects4J-v1.2/v2.0, QuixBugs) and three pre-trained models (UniXcoder, CodeBERT, ChatGPT), with UniXcoder performing best due to its MLM pre-training objective. The core finding is that LM-based donor code prediction can directly address the repair ingredient bottleneck in template-based APR without requiring fine-tuning on bug-fixing data.", 409 "red_flags": [ 410 { 411 "flag": "Perfect fault localization only", 412 "detail": "All evaluations assume the exact buggy line is known in advance — an unrealistic setting that inflates all APR results. The paper acknowledges this but does not report any real fault localization results." 413 }, 414 { 415 "flag": "No statistical testing", 416 "detail": "All comparisons between GAMMA and nine baselines are raw counts with no significance tests, confidence intervals, or effect size analysis, making it impossible to determine whether differences are statistically meaningful." 417 }, 418 { 419 "flag": "Subjective correctness labeling", 420 "detail": "Correct patch determination relies on three co-authors manually judging semantic equivalence to developer patches — not blinded, not using independent annotators, and no inter-rater reliability reported." 421 }, 422 { 423 "flag": "Reused baseline results without re-running", 424 "detail": "Table I note explicitly states: 'we reuse the released results from the most recent work [18] instead of directly running the APR tools,' creating potential inconsistency in fault localization or environment conditions." 425 }, 426 { 427 "flag": "No variance across runs", 428 "detail": "No repeated runs or variance statistics are reported for GAMMA or any baseline, despite the stochastic nature of beam search — single-run results are presented as definitive counts." 429 } 430 ], 431 "cited_papers": [ 432 { 433 "title": "TBar: Revisiting Template-Based Automated Program Repair", 434 "relevance": "Primary baseline and template source for GAMMA; the paper directly builds on and compares against TBar's fix patterns." 435 }, 436 { 437 "title": "Less Training, More Repairing Please: Revisiting Automated Program Repair via Zero-Shot Learning (AlphaRepair)", 438 "relevance": "Closest prior work combining pre-trained models with cloze-style APR; key baseline and conceptual comparison point." 439 }, 440 { 441 "title": "UniXcoder: Unified Cross-Modal Pre-Training for Code Representation", 442 "relevance": "The primary pre-trained model used in GAMMA's implementation for mask prediction." 443 }, 444 { 445 "title": "A Syntax-Guided Edit Decoder for Neural Program Repair (Recoder)", 446 "relevance": "State-of-the-art learning-based APR baseline against which GAMMA shows 26.15% improvement." 447 }, 448 { 449 "title": "CURE: Code-Aware Neural Machine Translation for Automatic Program Repair", 450 "relevance": "Recent NMT-based APR baseline included in the comparison study." 451 }, 452 { 453 "title": "Defects4J: A Database of Existing Faults to Enable Controlled Testing Studies for Java Programs", 454 "relevance": "Primary evaluation benchmark; 395 bugs used for all main comparisons." 455 }, 456 { 457 "title": "Automated Program Repair in the Era of Large Pre-Trained Language Models", 458 "relevance": "Contemporary work evaluating large LLMs for APR; excluded as baseline due to training data uncertainty but contextualizes GAMMA's results." 459 }, 460 { 461 "title": "Where Were the Repair Ingredients for Defects4J Bugs? Exploring the Impact of Repair Ingredient Retrieval", 462 "relevance": "Motivates the donor code limitation that GAMMA addresses, showing ~50% of Defects4J bugs lack local donor code." 463 }, 464 { 465 "title": "CoCoNuT: Combining Context-Aware Neural Translation Models Using Ensemble for Program Repair", 466 "relevance": "Learning-based APR baseline; also uses beam size 1000 as comparison point for GAMMA's 250." 467 } 468 ], 469 "engagement_factors": { 470 "practical_relevance": { 471 "score": 2, 472 "justification": "APR tools are directly usable by developers and the GitHub release enables practical adoption, but Java-only and perfect fault localization limits real-world applicability." 473 }, 474 "surprise_contrarian": { 475 "score": 1, 476 "justification": "The insight (use PLMs for donor code prediction instead of local retrieval) is intuitive in retrospect and represents incremental extension rather than paradigm challenge." 477 }, 478 "fear_safety": { 479 "score": 0, 480 "justification": "No AI safety or risk angle; paper is focused on software engineering tooling for bug fixing." 481 }, 482 "drama_conflict": { 483 "score": 0, 484 "justification": "Standard academic tool comparison with no controversy or competing claims from named parties." 485 }, 486 "demo_ability": { 487 "score": 2, 488 "justification": "GitHub repository with source code and examples is publicly available; the tool can be run on Defects4J with some GPU setup effort." 489 }, 490 "brand_recognition": { 491 "score": 0, 492 "justification": "Nanjing University academic group with no major industry affiliation; UniXcoder (Microsoft Research) is used but GAMMA itself is not associated with a prominent lab." 493 } 494 }, 495 "hn_data": { 496 "threads": [ 497 { 498 "hn_id": "34890787", 499 "title": "The usability of advanced type systems: Rust as a case study", 500 "points": 141, 501 "comments": 162, 502 "url": "https://news.ycombinator.com/item?id=34890787" 503 }, 504 { 505 "hn_id": "36962148", 506 "title": "WizMap - A tool for the visualization of latent spaces", 507 "points": 17, 508 "comments": 1, 509 "url": "https://news.ycombinator.com/item?id=36962148" 510 }, 511 { 512 "hn_id": "34310167", 513 "title": "The Usability of Advanced Type Systems: Rust as a Case Study", 514 "points": 2, 515 "comments": 0, 516 "url": "https://news.ycombinator.com/item?id=34310167" 517 }, 518 { 519 "hn_id": "24598805", 520 "title": "Michelson Holography: Dual-SLM Holography with Camera-in-the-Loop Optimization", 521 "points": 2, 522 "comments": 0, 523 "url": "https://news.ycombinator.com/item?id=24598805" 524 }, 525 { 526 "hn_id": "41688042", 527 "title": "Efficient Streaming Inference of Multimodal Large Language Models on 1 GPU", 528 "points": 1, 529 "comments": 0, 530 "url": "https://news.ycombinator.com/item?id=41688042" 531 }, 532 { 533 "hn_id": "35669201", 534 "title": "Data-Centric Parallelism in the Chiplet Age", 535 "points": 1, 536 "comments": 0, 537 "url": "https://news.ycombinator.com/item?id=35669201" 538 }, 539 { 540 "hn_id": "34879761", 541 "title": "The Usability of Advanced Type Systems: Rust as a Case Study", 542 "points": 1, 543 "comments": 0, 544 "url": "https://news.ycombinator.com/item?id=34879761" 545 } 546 ], 547 "top_points": 141, 548 "total_points": 165, 549 "total_comments": 163 550 } 551 }