pyproject.toml

[build-system]
requires = ["setuptools>=58.0.0", "wheel"]
build-backend = "setuptools.build_meta"

[project]
name = "benchbox"
version = "0.2.1"
description = "Embedded benchmark datasets and queries for database evaluation. Implements TPC-H, TPC-DS, TPC-DI and other industry-standard benchmarks."
readme = "README.md"
authors = [{ name = "Joe Harris", email = "joe@benchbox.dev" }]
license = {text = "MIT"}
keywords = ["benchmark", "database", "tpc-h", "tpc-ds", "tpc-di", "analytics", "olap", "performance-testing"]
classifiers = [
    "Development Status :: 4 - Beta",
    "Intended Audience :: Developers",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3.10",
    "Programming Language :: Python :: 3.11",
    "Programming Language :: Python :: 3.12",
    "Programming Language :: Python :: 3.13",
    "Programming Language :: Python :: 3.14",
    "Topic :: Database",
    "Topic :: Software Development :: Testing :: Acceptance",
    "Topic :: System :: Benchmark",
]
requires-python = ">=3.10,<3.15"
dependencies = [
    "textcharts>=0.1.0",
    # sqlglot ships majors ~quarterly with dialect-translation breakage; cap to next major to catch
    # regressions in CI rather than silently in `uv sync`. Bump after validating on the new major.
    # Floor is v25.6.0 — first release containing PR #3756, which preserves DuckDB
    # GROUP/ORDER BY ALL as bare keywords under identify=True (was a BenchBox post-fixup).
    "sqlglot>=25.6.0,<31.0.0",
    # CLI essentials
    # click: capped <9 - historically click majors rename decorators/context APIs.
    "click>=8.3.3,<9.0.0",  # >=8.3.3: PYSEC-2026-2132 fix
    "rich>=13.0.0",
    "keyring>=25.0.0",
    "psutil>=5.9.0",
    # pydantic: v1→v2 migration broke the entire validation surface; capped <3 until pydantic v3 lands
    # and is audited.
    "pydantic>=2.0.0,<3.0.0",
    "pyyaml>=6.0.0",
    "packaging>=24.0",
    # Python 3.10 compatibility (tomllib is stdlib in 3.11+)
    "tomli>=2.0.0; python_version < '3.11'",
    # pyarrow: ABI-sensitive (bundled C++); wide cap to <25 to give runway while catching jumps.
    "pyarrow>=21.0.0,<25.0.0",
    "numpy>=1.24.0",
    # pandas is intentionally NOT a core dependency. It is an optional
    # engine/DataFrame dep (DataFrame mode, TPC-DI, ClickHouse-local) declared
    # in the relevant extras (pandas, dataframe-pandas, tpcdi, modin, dask) and
    # pulled transitively by chdb. All pandas uses on the import path are lazy;
    # the minimal-import-surface CI gate enforces this. See
    # docs/development/dependency-inventory.md (pandas = DataFrame/Benchmark).
    "zstandard>=0.20.0",
    # requests: imported unconditionally by benchbox.core.data_fetch.downloader, which is
    # reached on the canonical JoinOrder benchmark import path. Must be core so a bare
    # `pip install benchbox` (no extras) doesn't raise ModuleNotFoundError. Floor matches
    # the version already pinned in the questdb / cloud-spark extras.
    "requests>=2.31.0",
    # grpcio / grpcio-status are NOT declared at the core level. They are
    # pulled in transitively by optional cloud / connectivity extras
    # (e.g. influxdb3-python, snowflake, databricks-sdk, pyspark[connect]).
    # Users without those extras don't pay the grpc weight.
]

[tool.setuptools]
# Include non-Python files specified in MANIFEST.in
include-package-data = true

[tool.setuptools.packages.find]
include = ["benchbox*"]
exclude = [
    "benchmark_runs",
    "benchmark_runs.*",
    # benchbox.experimental is intentionally INCLUDED in the default wheel (Option A, Beta contract).
    # It is an explicitly unsupported namespace with no CLI/registry integration.
    # See README.md "Beta Software" section for the documented support boundary.
    # To exclude it from the wheel, add "benchbox.experimental" and "benchbox.experimental.*" here.
]

[project.optional-dependencies]
# Development and documentation
dev = [
    "pytest>=8.0.0",
    "pytest-cov>=4.1.0",
    "pytest-benchmark>=4.0.0",
    "pytest-xdist>=3.0.0",
    # duckdb: tight integration with BenchBox SQL runner and browser snapshot writer.
    # Cap <2 to catch the next-major API break in CI; browser results.duckdb files
    # use the default v1.0.0 on-disk format (version 64), which @duckdb/duckdb-wasm
    # attaches regardless of writer version.
    "duckdb>=1.0.0,<2.0.0",
    "tox>=4.13.0",
    "pyspark>=3.5.0",  # Aligned with pyspark extra
]
docs = [
    "sphinx>=7.2.0",
    "roman-numerals<4.0",  # Sphinx 9 needs roman_numerals module; v4.x broke the import
    "furo>=2025.9.25",
    "myst-parser>=2.0.0",
    "sphinxcontrib-mermaid>=0.8.1",
    "pygments>=2.18.0",
    "sphinx-tags>=0.4.0",
    "sphinx-design>=0.6.0",
    "ablog>=0.11.13 ; python_version >= '3.11'",  # uv source override pins to git main for unreleased Sphinx 9 fix
]
viz = []
explorer = ["duckdb>=1.0.0,<2.0.0"]

# Meta-groups for common scenarios
all = [
    # Browser-generated results.duckdb files use the default v1.0.0 on-disk format,
    # readable by @duckdb/duckdb-wasm regardless of writer version; cap <2 to catch
    # the next-major API break in CI.
    "duckdb>=1.0.0,<2.0.0",
    "clickhouse-connect>=0.10.0",
    "clickhouse-driver>=0.2.0",
    "databricks-sql-connector>=2.0.0",
    "databricks-sdk>=0.20.0",
    "databricks-connect>=14.0.0",  # For DataFrame mode
    "datafusion>=50.1.0",  # Updated to match datafusion extra
    "pyarrow>=10.0.0",  # For DataFusion (also in core with higher version)
    "firebolt-sdk>=1.18.0",
    "google-cloud-bigquery>=3.0.0",
    "google-cloud-storage>=2.0.0",
    "redshift-connector>=2.0.0",
    "snowflake-connector-python>=3.0.0",
    "trino>=0.328.0",
    "presto-python-client>=0.8.4",
    "psycopg[binary]>=3.1",
    "pyodbc>=4.0.0",
    "azure-identity>=1.15.0",  # For Azure Synapse and Fabric
    "azure-storage-file-datalake>=12.14.0",  # For Microsoft Fabric OneLake
    "pyathena>=3.0.0",  # For AWS Athena
    "boto3>=1.20.0",
    "cloudpathlib>=0.15.0",
    "pandas>=2.0.0",  # For TPC-DI benchmark
    "chdb>=0.10.0; sys_platform != 'win32'",   # For ClickHouse local mode (not available on Windows)
    "deltalake>=1.2.1",  # For Delta Lake table format
    "pyiceberg>=0.10.0",  # For Iceberg table format
    "influxdb3-python>=0.10.0",  # For InfluxDB 3.0 time-series
    "pymysql>=1.1.0",  # For StarRocks and Doris via MySQL protocol
    "singlestoredb>=1.0.0",  # For SingleStore (Helios / self-managed)
    # DataFrame platforms
    "polars>=1.35.2",
    "dask[distributed]>=2024.1.0",
    "modin[ray]>=0.32.0",
    "pyspark[connect]>=3.5.0",
]
cloud = [
    "databricks-sql-connector>=2.0.0",
    "databricks-sdk>=0.20.0",
    "firebolt-sdk>=1.18.0",
    "google-cloud-bigquery>=3.0.0",
    "google-cloud-storage>=2.0.0",
    "redshift-connector>=2.0.0",
    "snowflake-connector-python>=3.0.0",
    "trino>=0.328.0",
    "pyathena>=3.0.0",  # For AWS Athena
    "boto3>=1.20.0",
    "cloudpathlib>=0.15.0",
]
cloudstorage = ["cloudpathlib>=0.15.0"]

# Individual platforms (for minimal installations)
duckdb = ["duckdb>=1.0.0,<2.0.0"]
clickhouse = ["clickhouse-driver>=0.2.0"]  # TCP binary protocol (clickhouse-server)
clickhouse-server = ["clickhouse-driver>=0.2.0"]  # Canonical extra for clickhouse-server platform
clickhouse-cloud = ["clickhouse-connect>=0.10.0"]  # HTTP protocol (ClickHouse Cloud)
databricks = [
    "databricks-sql-connector>=2.0.0",
    "databricks-sdk>=0.20.0",
    "cloudpathlib>=0.15.0",
]
bigquery = [
    "google-cloud-bigquery>=3.0.0",
    "google-cloud-storage>=2.0.0",
    "cloudpathlib>=0.15.0",
]
redshift = [
    "redshift-connector>=2.0.0",
    "boto3>=1.20.0",
    "cloudpathlib>=0.15.0",
]
snowflake = ["snowflake-connector-python>=3.0.0", "cloudpathlib>=0.15.0"]
trino = ["trino>=0.328.0", "cloudpathlib>=0.15.0"]
presto = ["presto-python-client>=0.8.4", "cloudpathlib>=0.15.0"]
postgresql = ["psycopg[binary]>=3.1"]
questdb = ["psycopg[binary]>=3.1", "requests>=2.28.0"]
synapse = ["pyodbc>=4.0.0"]
fabric = [
    "pyodbc>=4.0.0",
    "azure-identity>=1.15.0",
    "azure-storage-file-datalake>=12.14.0",
]
athena = [
    "pyathena>=3.0.0",
    "boto3>=1.20.0",
]
datafusion = ["datafusion>=50.1.0", "pyarrow>=10.0.0"]
firebolt = [
    "firebolt-sdk>=1.18.0",
    "cloudpathlib>=0.15.0",
]
databend = [
    "databend-driver>=0.28.0",
]
doris = ["pymysql>=1.0.0"]
influxdb = [
    "influxdb3-python>=0.10.0",
    "pyarrow>=10.0.0",
]
starrocks = ["pymysql>=1.1.0"]
singlestore = ["singlestoredb>=1.0.0"]

# Individual benchmark extras
tpcdi = ["pandas>=2.0.0"]
clickhouse-local = ["chdb>=0.10.0; sys_platform != 'win32'"]
table-formats = [
    "deltalake>=1.2.1",
    "pyiceberg>=0.10.0",
    "vortex-data>=0.29.0",
]
ai-primitives = [
    # ML/NLP stack used by benchbox/core/ai_primitives/ (all imports guarded with try/except)
    "sentence-transformers>=2.0.0",
    "torch>=2.0.0",
    "textblob>=0.17.0",
    "spacy>=3.0.0",
]

# DataFrame platform extras (for benchbox.platforms.dataframe)
# Plain-name extras are the preferred install names.
# dataframe-* variants are kept as backward-compatible aliases.

# Pandas family - string-based column access, eager evaluation
pandas = ["pandas>=2.0.0"]
dataframe-pandas = ["pandas>=2.0.0"]
modin = ["modin[ray]>=0.32.0", "pandas>=2.0.0"]
dataframe-modin = [
    "modin[ray]>=0.32.0",
    "pandas>=2.0.0",
]
dask = ["dask[distributed]>=2024.1.0", "pandas>=2.0.0"]
dataframe-dask = [
    "dask[distributed]>=2024.1.0",
    "pandas>=2.0.0",
]
# Note: cudf requires NVIDIA GPU with CUDA 12.x
# cuDF is not on PyPI directly - install via RAPIDS:
#   pip install --extra-index-url=https://pypi.nvidia.com cudf-cu12
# This extra is a placeholder to document the dependency; actual installation
# requires the RAPIDS pip index or conda.
cudf = []
dataframe-cudf = [
    # Placeholder - actual cudf installation requires NVIDIA pip index
]

# Expression family - expression-based API, lazy evaluation
polars = ["polars>=1.35.2"]
dataframe-polars = [
    "polars>=1.35.2",
]
pyspark = ["pyspark>=3.5.0"]
lakesail = ["pyspark[connect]>=3.5.0"]
velox = ["pyspark[connect]>=3.5.0"]
dataframe-pyspark = [
    "pyspark>=3.5.0",
]
dataframe-datafusion = [
    "datafusion>=50.1.0",
    "pyarrow>=10.0.0",
]

# DataFrame family convenience groups
dataframe-pandas-family = [
    "pandas>=2.0.0",
    "modin[ray]>=0.32.0",
    "dask[distributed]>=2024.1.0",
    # cudf not included - requires NVIDIA GPU/CUDA, install separately
]
dataframe-expression-family = [
    "polars>=1.35.2",
    "pyspark>=3.5.0",
    "datafusion>=50.1.0",
]
dataframe-all = [
    "pandas>=2.0.0",
    "modin[ray]>=0.32.0",
    "dask[distributed]>=2024.1.0",
    "polars>=1.35.2",
    "pyspark>=3.5.0",
    "datafusion>=50.1.0",
]

# Quick start - DuckDB + Polars for easy experimentation
quick-start = [
    "polars>=1.35.2",
]

# Cloud Spark Platforms - Managed Spark with DataFrame API
# These extras enable DataFrame mode execution on managed cloud Spark platforms

# Provider-specific cloud Spark groups
cloud-spark-aws = [
    "boto3>=1.34.0",
]
cloud-spark-gcp = [
    "google-cloud-dataproc>=5.0.0",
    "google-cloud-storage>=2.0.0",
]
cloud-spark-azure = [
    "azure-identity>=1.15.0",
    "azure-storage-file-datalake>=12.14.0",
    "requests>=2.31.0",
]
cloud-spark-snowflake = [
    "snowflake-snowpark-python>=1.11.0",
    "pyspark>=3.5.0",
]
cloud-spark-databricks = [
    "databricks-sql-connector>=2.0.0",
    "databricks-sdk>=0.20.0",
    "databricks-connect>=14.0.0",
    "cloudpathlib>=0.15.0",
]

# All cloud Spark platforms combined
cloud-spark = [
    # AWS (Glue, EMR Serverless, Athena Spark)
    "boto3>=1.34.0",
    # GCP (Dataproc, Dataproc Serverless)
    "google-cloud-dataproc>=5.0.0",
    "google-cloud-storage>=2.0.0",
    # Azure (Synapse Spark, Fabric Spark)
    "azure-identity>=1.15.0",
    "azure-storage-file-datalake>=12.14.0",
    "requests>=2.31.0",
    # Snowflake (Snowpark)
    "snowflake-snowpark-python>=1.11.0",
    "pyspark>=3.5.0",
    # Databricks
    "databricks-sql-connector>=2.0.0",
    "databricks-sdk>=0.20.0",
    "databricks-connect>=14.0.0",
    "cloudpathlib>=0.15.0",
]

# MCP (Model Context Protocol) server for AI agent integration
# Note: mcp requires Python 3.10+, so this extra only works on 3.10+
# Includes visualization deps since MCP server exposes chart generation tools
mcp = [
    "mcp>=1.28.1",  # >=1.28.1: CVE-2026-52870/52869/59950 fixes
]
spark = [
    "pyspark>=4.0.1",
]

[project.scripts]
benchbox = "benchbox.cli.app:main"
benchbox-mcp = "benchbox.mcp.cli:main"

[project.urls]
"Homepage" = "https://github.com/joeharris76/benchbox"
"Bug Tracker" = "https://github.com/joeharris76/benchbox/issues"
"Documentation" = "https://benchbox.readthedocs.io/"
"Repository" = "https://github.com/joeharris76/benchbox.git"
"Changelog" = "https://github.com/joeharris76/benchbox/blob/release/CHANGELOG.md"

[tool.ty]
# ty configuration

[tool.ty.src]
# Source code to type check
include = ["benchbox", "tests"]
exclude = ["tests/fixtures", "**/__pycache__"]

[tool.ty.rules]
# Configure rule severities - start lenient, gradually tighten
# Note: This codebase was developed without strict type checking.
# Rules are configured permissively; tighten as code is updated.

# Error: These rules are now enforced (previously ignored, now fixed)
too-many-positional-arguments = "error"  # Fixed: method signature alignment
conflicting-declarations = "error"       # Fixed: optional import patterns
invalid-parameter-default = "error"      # Fixed: Optional[] wrappers added
unresolved-reference = "error"           # Fixed: missing imports/variables
not-iterable = "error"                   # Fixed: type narrowing and assertions

# Ignore: These rules have too many false positives or require major refactoring
unresolved-attribute = "ignore"         # 954 hits - Missing type stubs, dynamic attributes, class method discovery
invalid-argument-type = "ignore"        # 763 hits - Function param type mismatches (str vs Literal, Optional narrowing)
invalid-assignment = "ignore"           # 463 hits - Literal type mismatches, Optional subscript access
unsupported-operator = "ignore"         # 184 hits - Dict value type inference (counter += 1 patterns)
invalid-method-override = "ignore"      # 143 hits - Intentional LSP violations (run_benchmark, execute_query, generate_data, etc.)
not-subscriptable = "ignore"            # ~700 hits - Dynamic DataFrame APIs, Optional dict access patterns
call-non-callable = "ignore"            # ~700 hits - Optional imports, dynamic DataFrame APIs, hasattr patterns

# Warn: Track these but don't fail CI
unresolved-import = "warn"              # Often optional dependencies
invalid-return-type = "warn"            # Should be fixed but not blocking
invalid-type-form = "warn"              # Syntax issues
no-matching-overload = "warn"           # Complex overload patterns
unknown-argument = "warn"               # Keyword argument issues
missing-argument = "warn"               # Tuple unpacking not tracked

# Control flow tracking limitations
possibly-unresolved-reference = "warn"
possibly-missing-attribute = "warn"

[tool.ty.terminal]
# Output configuration
error-on-warning = false

[dependency-groups]
dev = [
    "ruff==0.11.13",  # pinned: preview = true is active globally; ruff upgrades may activate new rules
    "ty>=0.0.1a14",
    "ansi2html>=1.8.0",  # scripts/capture_chart_images.py - ANSI→HTML conversion for chart captures
    "pillow>=12.3.0",  # scripts/capture_chart_images.py - PIL.Image for chart PNG output; >=12.3.0: PYSEC-2026-2253..3453 fixes
    # Explorer fixture and deployment snapshots use the default v1.0.0 on-disk format,
    # readable by @duckdb/duckdb-wasm regardless of writer version; cap <2 to catch
    # the next-major API break in CI.
    "duckdb>=1.0.0,<2.0.0",
    "polars>=1.35.2",
    "clickhouse-connect>=0.10.0",
    "psycopg[binary]>=3.1",
    "pytest>=8.0.0", # Test framework
    "pytest-cov>=6.2.1",
    "pytest-benchmark>=5.1.0",
    "pytest-xdist>=3.8.0",
    "pytest-timeout>=2.4.0",
    "pandas>=2.0.0", # DataFrame dependencies (needed for many tests)
    "pyiceberg[sql-sqlite,pyarrow]>=0.10.0", # Table format converters (needed for tests/unit/utils/format_converters/)
    "deltalake>=1.2.1",
    "datafusion>=40.0.0", # DataFusion (needed for tests/integration/test_datafusion_integration.py)
    "cloudpathlib[s3,gs,azure]>=0.15.0", # Cloud storage with S3 support (needed for tests/unit/utils/test_cloud_storage.py)
    "chdb>=0.10.0; sys_platform != 'win32'", # ClickHouse local (chdb is not available on Windows)
    "clickhouse-driver>=0.2.0", # ClickHouse server (needed for tests/unit/platforms/test_clickhouse_local.py server tests)
    "trino>=0.336.0", # Firebolt (needed for tests/unit/platforms/test_firebolt_adapter.py)
    "firebolt-sdk>=1.18.0", # InfluxDB (needed for tests/unit/platforms/test_influxdb_adapter.py)
    "influxdb3-python>=0.10.0", # Databricks SDK (needed for tests/unit/platforms/test_databricks_copy_into.py UC volumes)
    "databricks-sdk>=0.20.0", # PySpark (needed for tests/unit/platforms/dataframe/test_pyspark_df.py and related)
    "pyspark>=3.5.0", # XML parsing (needed for tests/unit/core/tpcdi/test_etl_sources.py)
    "mcp>=1.28.1",  # >=1.28.1: CVE-2026-52870/52869/59950 fixes
    "sphinx>=7.4.7", # Documentation
    "roman-numerals<4.0", # Sphinx 9 needs roman_numerals module; v4.x broke the import
    "myst-parser>=3.0.1",
    "sphinxcontrib-mermaid>=1.2.3",
    "sphinx-tags>=0.4",
    "sphinx-design>=0.6.1",
    "furo>=2025.9.25",
    "ablog>=0.11.13 ; python_version >= '3.11'", # version that includes a required Sphinx 9 fix
    "mutmut>=3.5.0",
    "dask[distributed]>=2026.1.1",
    "pymysql>=1.1.2",
    "delta-spark>=4.1.0",
    "presto-python-client>=0.8.4",
    "codespell>=2.4.2",
    "pre-commit>=4.6.0",
    "import-linter>=2.13",
]

[tool.mutmut]
paths_to_mutate = [
    "benchbox/platforms/duckdb.py",
    "benchbox/platforms/base/adapter.py",
    "benchbox/core/runner/runner.py",
    "benchbox/core/visualization/chart_generator.py",
    "benchbox/cli/commands/run.py",
]
tests_dir = ["tests/unit/"]

[tool.setuptools.package-data]
"benchbox.core" = ["benchmark_registry.yaml"]
"benchbox.core.equivalence" = ["cross_surface_baseline.yaml"]
"benchbox.core.flightdata" = ["query_catalog.yaml"]
"benchbox.core.nyctaxi" = ["query_catalog.yaml"]
"benchbox.core.results" = ["benchmark_specs.yaml"]
"benchbox.core.primitives.catalog" = ["queries.yaml"]
"benchbox.core.tsbs_devops" = ["query_catalog.yaml"]
"benchbox.data.coffeeshop" = ["*.csv"]

[tool.ruff]
line-length = 120
exclude = [
    "migrations",
    "venv",
    ".git",
    "__pycache__",
    ".tox",
    # Non-production directories
    "_project",             # Internal project tooling and scripts
    "examples/notebooks",   # Example notebooks (different code style)
    "examples/visualization", # Visualization notebooks
    "tests/validation",     # One-off validation scripts
]

[tool.ruff.lint]
# preview = true is required to activate PLW1514 (unspecified-encoding).
# Side-effect: surfaces pre-existing F822 violations in __init__.py lazy-import
# modules and one C419/F401 elsewhere. Those are suppressed below.
preview = true
# Enable comprehensive linting rules that replace flake8/pycodestyle
select = [
    "E",       # pycodestyle errors
    "W",       # pycodestyle warnings
    "F",       # pyflakes
    "I",       # isort
    "N",       # pep8-naming
    "UP",      # pyupgrade
    "B",       # flake8-bugbear
    "C4",      # flake8-comprehensions
    "C90",     # mccabe complexity
    "PIE",     # flake8-pie
    "SIM",     # flake8-simplify
    "PLW1514", # unspecified-encoding: open() without explicit encoding= (Windows compat)
]
ignore = [
    "E501",   # line too long (handled by formatter)
    # The following are temporarily ignored to allow CI to pass.
    # These pre-existed and should be addressed in a future cleanup.
    # TODO: Create TODO item to address these lint issues systematically
    "E402",   # module-import-not-at-top-of-file (often intentional for conditional imports)
    "E721",   # type-comparison (use isinstance - valid but requires careful review)
    "E731",   # lambda-assignment (sometimes clearer than def)
    "E741",   # ambiguous-variable-name (context-dependent)
    # Bugbear rules - good practices but not critical
    "B905",   # zip-without-explicit-strict (19 occurrences; TODO: add strict= and remove)
    "B007",   # unused-loop-control-variable (often intentional, e.g., for _ in range())
    "B017",   # assert-raises-exception (test pattern)
    "B023",   # function-uses-loop-variable (often intentional in closures)
    "B024",   # abstract-base-class-without-abstract-method
    "B027",   # empty-method-without-abstract-decorator
    # Simplify rules - style preferences
    "SIM102", # collapsible-if
    "SIM103", # needless-bool
    "SIM105", # suppressible-exception
    "SIM108", # if-else-block-instead-of-if-exp (ternaries can reduce readability)
    "SIM110", # reimplemented-builtin
    "SIM113", # enumerate-for-loop
    "SIM115", # open-file-with-context-handler (sometimes intentional)
    "SIM116", # if-else-block-instead-of-dict-lookup
    "SIM117", # multiple-with-statements
    "SIM118", # in-dict-keys
    "SIM201", # negate-equal-op
    "SIM401", # if-else-block-instead-of-dict-get
    # Comprehensions
    "C401",   # unnecessary-generator-set
    # PIE rules
    "PIE810", # multiple-starts-ends-with
    # Upgrade rules - require newer Python versions
    "UP006",  # non-pep585-annotation (use dict instead of Dict - requires Python 3.9+)
    "UP007",  # non-pep604-annotation-union (use X | Y - requires Python 3.10+)
    "UP035",  # deprecated-import (typing.Dict etc - requires Python 3.9+)
    "UP045",  # non-pep604-annotation-optional
    # Naming conventions - often intentional in specific contexts
    "N802",   # invalid-function-name (test methods, compatibility)
    "N803",   # invalid-argument-name (mathematical notation, API compatibility)
    "N806",   # non-lowercase-variable-in-function (constants, compatibility)
    "N811",   # constant-imported-as-non-constant
    "N818",   # error-suffix-on-exception-name
    # C901 (mccabe complexity) is now a permanent gate - all violations resolved.
]

[tool.ruff.lint.isort]
force-single-line = false
combine-as-imports = true

[tool.ruff.lint.mccabe]
# Cyclomatic complexity gate: no function may exceed McCabe 18.
# All pre-existing violations resolved as of 2026-04-14.
max-complexity = 18

[tool.ruff.lint.per-file-ignores]
# Files that use `from x import *` pattern legitimately
"benchbox/core/tpcds/schema/registry.py" = ["F405"]
# Test files often have unused imports for fixtures or type checking
"tests/**/conftest.py" = ["F401", "F811"]
"tests/**/test_*.py" = ["F401", "F811"]
# __init__.py files often re-export symbols via lazy imports; F822 fires in
# preview mode for names defined in TYPE_CHECKING blocks or deferred imports.
"**/__init__.py" = ["F401", "F811", "F822"]
# Pre-existing violations surfaced by preview mode; out of scope for windows-ci-hardening.
"benchbox/utils/runtime_env.py" = ["F401"]               # unused importlib.machinery import
# PLW1514 (unspecified-encoding): all benchbox/ sites now have explicit encoding=
# (fixed under quality-bulk-fix-encoding-lint-and-ratchet). Per-file-ignores removed.

[tool.ruff.format]
# Use black-compatible formatting
quote-style = "double"
indent-style = "space"
skip-magic-trailing-comma = false
line-ending = "auto"

[tool.interrogate]
# Docstring coverage enforcement
ignore-init-method = false
ignore-init-module = false
ignore-magic = false
ignore-module = false
ignore-nested-functions = true
ignore-nested-classes = true
ignore-private = true
ignore-property-decorators = false
# Keep the blocking gate focused on public API documentation; single-underscore
# helpers churn heavily and are covered by code review/tests instead.
ignore-semiprivate = true
ignore-setters = false
fail-under = 90  # Ratcheted from baseline after local audit reported 93.1% coverage.
exclude = [
    "tests",
    "_build",
    "venv",
    ".venv",
    "benchmark_runs",
    "_project",
    "docs/_build",
    "benchbox/platforms/credentials",
    # Bulk mechanical DataFrame query catalogs are generated-style
    # implementations; keep the gate focused on hand-authored public APIs.
    "benchbox/core/clickbench/dataframe_queries/queries.py",
    "benchbox/core/datavault/dataframe_queries/queries.py",
    "benchbox/core/flightdata/dataframe_queries.py",
    "benchbox/core/nyctaxi/dataframe_queries/queries.py",
    "benchbox/core/tpcds/dataframe_queries/queries.py",
    "benchbox/core/tpcds_obt/dataframe_queries.py",
    "benchbox/core/tpchavoc/dataframe_queries",
]
verbose = 2
quiet = false
whitelist-regex = []
color = true

[tool.coverage.run]
# Test coverage configuration
source = ["benchbox"]
branch = true
omit = [
    "benchbox/__main__.py",
    "**/test_*.py",
    "**/conftest.py",
    # Low-ROI bulk DF query sets: mechanical translations untestable without
    # full-scale benchmark infrastructure (TPC-DS requires SF>=1, others need
    # large external datasets). Excluding ~5,742 statements.
    "benchbox/core/tpcds/dataframe_queries/queries.py",
    "benchbox/core/tpcds/dataframe_queries/rollup_helper.py",
    "benchbox/core/tpcds/dataframe_queries/channel_helper.py",
    "benchbox/core/datavault/dataframe_queries/queries.py",
    "benchbox/core/nyctaxi/dataframe_queries/queries.py",
    # TPC-DI ETL: requires full CDC/SCD pipeline infrastructure with staged
    # incremental loads. Cannot be meaningfully unit-tested at unit granularity.
    "benchbox/core/tpcdi/etl/scd_processor.py",
    # Experimental load testing: changes frequently, pool tester requires real
    # concurrent database connections (~284 miss stmts).
    "benchbox/experimental/load_testing/executor.py",
    "benchbox/experimental/load_testing/pool_tester.py",
    # ClickBench DF queries: single wide table with 105 columns and specific
    # data patterns, same rationale as TPC-DS/datavault/nyctaxi DF omissions
    # (~310 miss stmts).
    "benchbox/core/clickbench/dataframe_queries/queries.py",
]

[tool.coverage.report]
# Coverage reporting thresholds
fail_under = 80
show_missing = true
skip_covered = false
exclude_lines = [
    "pragma: no cover",
    "if TYPE_CHECKING:",
    "if __name__ == .__main__.:",
    "raise NotImplementedError",
    "@abstractmethod",
]


[tool.benchbox.complexity]
# Cyclomatic complexity check configuration (used by scripts/check_complexity.py).
# max_complexity: hard ceiling - functions above this are CI failures.
# warn_complexity: advisory - functions above this are printed as warnings.
max_complexity = 20
warn_complexity = 12
# Modules temporarily excluded from per-module complexity checks.
exclude_modules = []
# Individual functions excluded from the check (format: "file:function").
# Use sparingly - only for genuinely irreducible complexity (e.g., CLI dispatch).
exclude_functions = [
    # CLI run() is a Click command with many flags; refactoring underway.
    "benchbox/cli/commands/run.py:run",
    # ── Approved ignore_with_rationale entries (2026-02-20) ──
    # Each maps to a row in quality/complexity/classification-matrix.yaml
    # Re-review triggers are documented per-function in the matrix.
    # --- ASCII visualization renderers (inherent grid/layout complexity) ---
    "benchbox/core/visualization/ascii/line_chart.py:render",          # 43 - Bresenham line drawing + grid state
    "benchbox/core/visualization/ascii/box_plot.py:render",            # 30 - three-line-per-series box plot geometry
    "benchbox/core/visualization/ascii/scatter_plot.py:render",        # 26 - Pareto frontier + grid point placement
    "benchbox/core/visualization/ascii/bar_chart.py:render",           # 22 - outlier truncation + sub-character blocks
    "benchbox/core/visualization/ascii/rank_table.py:render",          # 22 - ranking with tie handling
    "benchbox/core/visualization/ascii/cdf_chart.py:render",           # 21 - grid-based CDF plotting
    # --- TPC specification implementations (spec-driven branching) ---
    "benchbox/core/tpcds/power_test.py:run",                           # 33 - TPC-DS power test per spec
    "benchbox/core/tpcds_obt/query_conversion.py:_post_process",       # 32 - per-query OBT conversion edge cases
    "benchbox/core/tpcds/c_tools.py:_generate_with_binary",            # 25 - dsqgen binary wrapper
    "benchbox/core/tpcds/throughput_test.py:_execute_stream",          # 24 - TPC-DS throughput stream execution
    "benchbox/core/tpcds/benchmark/runner.py:run_throughput_test",     # 22 - TPC-DS throughput test per spec
    "benchbox/core/tpcds/benchmark/runner.py:run_power_test",          # 22 - TPC-DS power test per spec
    "benchbox/core/results/display.py:print_dry_run_summary",          # 21 - conditional section display
    "benchbox/core/tpcds/throughput_test.py:run",                      # 18 - TPC-DS spec; outer loop coordinates concurrent streams
    "benchbox/core/tpcds_obt/query_conversion.py:_rewrite_select",     # 18 - AST rewrite with dependency ordering (role_map → column → join)
    "benchbox/core/tpch/maintenance_test.py:_execute_rf1",             # 18 - TPC-H RF1 spec; each SQL step depends on prior result
    "benchbox/core/tpch/power_test.py:run",                            # 18 - TPC-H power test; spec mandates permutation order
    "benchbox/core/tpch/throughput_test.py:run",                       # 18 - TPC-H throughput; spec mandates TTT timing from first to last stream
    "benchbox/core/runner/dataframe_runner.py:_execute_dataframe_queries",  # 18 - TPC power test methodology; warmup + permuted query loop
    "benchbox/core/tpch/throughput_test.py:_execute_stream",           # 17 - TPC-H single stream; 22 queries in spec permutation order
    # --- Platform adapters (genuinely distinct platform semantics) ---
    "benchbox/platforms/base/adapter.py:_summarize_performance_characteristics",  # 34 - well-structured statistical aggregation
    "benchbox/platforms/dataframe/unified_frame.py:join",              # 33 - multi-platform join dispatch (Polars/PySpark/DataFusion)
    "benchbox/platforms/base/adapter.py:_collect_resource_utilization", # 24 - cross-platform metric collection with independent try/except
    "benchbox/platforms/credentials/redshift.py:setup_redshift_credentials",  # 24 - interactive credential wizard
    "benchbox/platforms/redshift.py:get_platform_info",                # 24 - serverless vs provisioned deployment branching
    "benchbox/platforms/bigquery.py:get_platform_info",                # 23 - independent INFORMATION_SCHEMA queries
    "benchbox/platforms/dataframe/unified_frame.py:_datafusion_join_with_exprs",  # 23 - DataFusion join workaround
    "benchbox/platforms/credentials/redshift.py:validate_redshift_credentials",  # 22 - sequential validation pipeline
    "benchbox/platforms/clickhouse/metadata.py:get_platform_info",     # 21 - local vs server mode branching
    "benchbox/platforms/credentials/bigquery.py:validate_bigquery_credentials",  # 21 - multi-service validation
    "benchbox/platforms/databricks/adapter.py:apply_table_tunings",    # 21 - Liquid Clustering vs Z-ORDER
    "benchbox/platforms/dataframe/unified_frame.py:_extract_multi_agg_arithmetic",  # 21 - DataFusion AST workaround
    "benchbox/platforms/dataframe/unified_frame.py:sort",              # 20 - Polars/PySpark/DataFusion each require distinct sort APIs
    "benchbox/platforms/credentials/athena.py:setup_athena_credentials",  # 20 - interactive credential wizard; auto-detect vs manual with dependent branching
    "benchbox/platforms/credentials/bigquery.py:setup_bigquery_credentials",  # 20 - interactive credential wizard; file validation + optional Cloud Storage path
    "benchbox/platforms/dataframe/unified_frame.py:_apply_datafusion_post_ops",  # 19 - literal vs multi-agg arithmetic post-ops have fundamentally different expression semantics
    "benchbox/platforms/base/adapter.py:capture_query_plan",           # 17 - multi-level filter chain (query ID, iteration count, sampling, EXPLAIN)
    "benchbox/platforms/credentials/snowflake.py:setup_snowflake_credentials",  # 16 - interactive credential wizard; account normalization + multi-step auth validation
    # --- MCP tool registration (structural nesting from framework) ---
    "benchbox/mcp/tools/discovery.py:register_discovery_tools",        # 29 - MCP registration pattern
    "benchbox/mcp/tools/analytics.py:register_analytics_tools",        # 22 - MCP registration pattern
    # --- Thread-safety (single-flight cache; locking protocol cannot be simplified) ---
    "benchbox/core/expected_results/registry.py:_get_benchmark_results",  # 17 - single-flight cache with cross-thread coordination
    # --- Query plan parsers (AST node dispatch; each operator has distinct properties) ---
    "benchbox/core/query_plans/parsers/datafusion.py:_convert_to_logical_operator",  # 17 - AST node → LogicalOperator with operator-specific property extraction
    "benchbox/core/query_plans/parsers/datafusion.py:_extract_operator_info",  # 23 - operator-type dispatch (scan/join/filter/sort/projection/aggregate/limit)
    # --- Platform registration (41 independent try/except blocks, one per platform) ---
    "benchbox/core/platform_registry.py:auto_register_platforms",          # 41 - each platform import is isolated; CC reflects platform count, not algorithmic complexity
    # --- Catalog loaders (per-field YAML validation with distinct rules per field) ---
    "benchbox/core/ai_primitives/catalog/loader.py:load_ai_catalog",       # 29 - field-by-field validation: id, sql, category, variants, skip_on, model, tokens, batch_size, cost
    "benchbox/core/write_primitives/catalog/loader.py:load_write_primitives_catalog",  # 25 - field-by-field validation: id, category, write_sql, validation_queries, cleanup_sql, flags
    # --- Manifest validation (multi-layer validation recently expanded for contamination detection) ---
    "benchbox/core/runner/runner.py:_validate_manifest_if_present",        # 23 - entry validation + directory collision detection + cloud/local path dispatch
    # --- Benchmark test setup (spec-required sequential dependency chain) ---
    "benchbox/core/transaction_primitives/benchmark.py:setup",          # 17 - spec-required: validation → lock → cleanup → staging population in strict order
    # --- Deferred complexity reduction (architectural redesign needed) ---
    "benchbox/core/runner/runner.py:run_benchmark_lifecycle",            # 39 - main orchestration with genuinely distinct code paths (data_only, load_only, SQL, DataFrame)
    "benchbox/cli/orchestrator.py:execute_benchmark",                    # 24 - interleaved output/phase/adapter/credential logic; tightly coupled to CLI UX
    # --- CLI/release (linear workflows with safety-critical branching) ---
    "benchbox/cli/commands/compare_plans.py:compare_plans",            # 27 - deprecated, slated for removal
    "benchbox/cli/commands/export.py:export",                          # 22 - clean CLI with distinct source resolution paths
    "benchbox/release/sync.py:cmd_pull",                               # 25 — auditable safety-check workflow
    "benchbox/release/sync.py:cmd_push",                               # 22 — mirror of cmd_pull
    # --- Interactive shell REPL (user-input dispatch loop is state-machine-inherent) ---
    "benchbox/cli/commands/shell.py:_launch_sqlite_shell",             # 16 - SQLite REPL; command dispatch + SQL execution loop cannot be flattened
    "benchbox/cli/commands/shell.py:_launch_duckdb_shell",             # 15 - DuckDB REPL; same pattern as SQLite shell
    # --- Converters (linear pipelines with resource cleanup) ---
    "benchbox/utils/format_converters/iceberg_converter.py:convert",   # 24 - Iceberg conversion pipeline
    # --- Required complexity at CC=14 (2026-02-21 evaluation, batch w15) ---
    # These sit at CC=14 (below warn=15) but are documented for completeness.
    "benchbox/cli/commands/show_plan.py:show_plan",                    # 14 - 4 except branches + 3-way format dispatch; irreducible for CLI command
    "benchbox/cli/commands/tuning_group.py:_display_config_settings_table",  # 14 - one guard per config field; proportional to field count
    "benchbox/core/query_plans/parsers/base.py:_harmonize_operator_type",  # 14 - 12-way keyword→enum mapping; 1:1 with enum members
    "benchbox/core/concurrency/analyzer.py:analyze_contention",        # 14 - mutually-exclusive classification decision tree
    "benchbox/core/metadata_primitives/complexity.py:__post_init__",   # 14 - N independent field validation guards; inherent to dataclass
    # --- Required complexity at CC=14 (2026-03-12 evaluation) ---
    # Inherent complexity: domain-driven branching that cannot be meaningfully reduced.
    "benchbox/core/results/builder.py:to_dict",                        # 14 - conditional serialization of 12 optional fields; each `if` guards one field
    "benchbox/platforms/clickhouse_cloud.py:_create_external_tables_via_s3",  # 14 - 4-way format dispatch (iceberg/parquet × local/cloud) with distinct S3 semantics
    "benchbox/platforms/databricks/adapter.py:get_platform_info",      # 14 - independent try/except probes for SDK, runtime, and warehouse metadata
    "benchbox/platforms/dataframe/benchmark_mixin.py:run_benchmark",   # 14 - orchestration flow: memory check → validation → context → load → execute → result
    "benchbox/platforms/redshift.py:apply_table_tunings",              # 14 - Redshift DDL constraints require individual DISTSTYLE/DISTKEY/SORTKEY inspection
    "benchbox/utils/format_converters/base.py:map_sql_type_to_arrow",  # 14 - type dispatch; DECIMAL requires parameter parsing that breaks dict-based approach
    "benchbox/utils/input_validation.py:validate_query_id",            # 14 - benchmark-specific allowlists with distinct ID format rules per benchmark
    "benchbox/utils/stale_artifact_pruning.py:prune_stale_table_artifacts",  # 14 - compressed/uncompressed × single/sharded × expected-format; branching reflects real combinations
    # --- Required complexity at CC=13 (2026-03-12 evaluation, wave 3) ---
    # Inherent complexity: validator/serializer logic, spec orchestration, and platform-specific semantics.
    "benchbox/cli/benchmarks.py:prompt_table_format",                  # 13 - interactive format/compression wizard with dependent prompts
    "benchbox/cli/composite_params.py:parse",                          # 13 - per-key plan-capture parsing with distinct validation rules
    "benchbox/cli/config.py:create_sample_unified_tuning_config",      # 13 - sample tuning config mirrors real platform capability differences
    "benchbox/cli/platform.py:setup_platforms",                        # 13 - stateful interactive setup wizard with action dispatch
    "benchbox/core/dataframe/capabilities.py:check_sufficient_memory", # 13 - GPU/system memory checks and platform recommendations are capability-driven
    "benchbox/core/dryrun.py:_extract_queries",                        # 13 - query extraction dispatch is benchmark/spec driven across SQL and DataFrame modes
    "benchbox/core/metadata_primitives/dataframe_operations.py:execute_get_schema",  # 13 - per-platform schema APIs are genuinely distinct
    "benchbox/core/query_plans/parsers/duckdb.py:_convert_to_logical_operator",  # 13 - AST-driven operator dispatch with type-specific property extraction
    "benchbox/core/query_plans/parsers/duckdb.py:_harmonize_duckdb_operator",   # 13 - 1:1 DuckDB operator mapping to logical enums
    "benchbox/core/results/query_plan_models.py:get_structural_signature",       # 13 - semantic fingerprint builder conditionally serializes plan fields
    "benchbox/core/results/schema.py:validate",                       # 13 - schema validator applies separate rules per required block
    "benchbox/core/results/schema.py:_build_execution_from_context",  # 13 - execution serialization merges optional flags with special-case compression/force handling
    "benchbox/core/results/schema.py:_build_config_block",            # 13 - config serialization needs per-field precedence and compression merge rules
    "benchbox/core/results/schema.py:_build_phases_block",            # 13 - phase serialization mirrors heterogeneous setup/power/throughput structures
    "benchbox/core/schemas.py:to_cli_args",                           # 13 - CLI reconstruction is proportional to optional flag formatting rules
    "benchbox/core/tpc_validation.py:validate",                       # 13 - validator subclasses apply one rule per completeness/result/schema concern
    "benchbox/core/tpcdi/queries.py:resolve_query_order",             # 13 - Kahn topological sort with query-only dependency filtering is algorithmic
    "benchbox/core/tpcds/generator/streaming.py:_generate_parent_table_chunk_with_children",  # 13 - TPC-DS parent/child chunk generation couples subprocess, compression, and manifest updates
    "benchbox/core/tpch/generator.py:_run_dbgen_native",              # 13 - dbgen orchestration spans streaming/file modes, env setup, and cleanup
    "benchbox/core/tpch/maintenance_test.py:_execute_rf2",            # 13 - TPC-H RF2 requires ordered deletes, transaction handling, and rollback
    "benchbox/core/transaction_primitives/benchmark.py:execute_operation",  # 13 - transaction operation lifecycle coordinates setup, validation, cleanup, and warnings
    "benchbox/core/tuning/generators/doris.py:generate_tuning_clauses",      # 13 - Doris tuning maps distinct tuning types to engine-specific DDL semantics
    "benchbox/core/tuning/generators/redshift.py:generate_tuning_clauses",   # 13 - Redshift tuning maps sort/dist semantics to engine-specific clauses
    "benchbox/core/write_primitives/benchmark.py:setup",              # 13 - staging-table setup is a guarded sequence with lock/release guarantees
    "benchbox/platforms/azure_synapse.py:load_data",                  # 13 - load path must juggle benchmark tables, manifest fallback, and blob vs bulk-insert semantics
    "benchbox/platforms/base/adapter.py:handle_existing_database",    # 13 - safety-critical reuse/recreate decision tree depends on file vs server DB semantics
    "benchbox/platforms/base/adapter.py:_execute_combined_test",      # 13 - combined TPC flow dispatches benchmark-specific phase orchestration
    "benchbox/platforms/base/adapter.py:_execute_tpcds_power_test",   # 13 - wraps official TPC-DS power test with warmups, stream handling, and normalization
    "benchbox/platforms/base/adapter.py:_execute_tpcds_throughput_test",  # 13 - wraps official TPC-DS throughput test with stream/concurrency normalization
    "benchbox/platforms/base/spark_execution_mixin.py:_load_data_spark",    # 13 - Spark load path branches on format, staging, and Spark-specific writer semantics
    "benchbox/platforms/bigquery.py:apply_table_tunings",             # 13 - tuning application depends on BigQuery partition/clustering metadata constraints
    "benchbox/platforms/clickhouse_cloud.py:_create_external_tables_via_gcs",  # 13 - GCS external view creation spans local/cloud Iceberg and Parquet inputs
    "benchbox/platforms/databricks/adapter.py:_upload_to_uc_volume",  # 13 - UC volume uploads handle local vs manifest sources and Databricks path semantics
    "benchbox/platforms/dataframe/polars_maintenance.py:_do_merge",   # 13 - Polars MERGE emulation dispatches inserts/updates/deletes by clause semantics
    "benchbox/platforms/datafusion.py:_convert_and_register_parquet",  # 13 - DataFusion loading coordinates conversion, registration, and schema preservation
    "benchbox/platforms/duckdb.py:execute_query",                     # 13 - DuckDB execution path combines execution, plan capture, and result-shape compatibility
    "benchbox/platforms/firebolt.py:_resolve_data_files",             # 13 - Firebolt file resolution spans local/cloud/manifest staging variants
    "benchbox/platforms/firebolt.py:_load_data_via_insert",           # 13 - Firebolt insert loading coordinates staged files, SQL generation, and fallback handling
    "benchbox/platforms/influxdb/client.py:to_line_protocol",         # 13 - Influx line-protocol encoding branches by tags, fields, types, and escaping rules
    "benchbox/platforms/presto.py:get_platform_info",                 # 13 - platform info comes from independent engine/catalog/session probes
    "benchbox/platforms/redshift.py:configure_for_benchmark",         # 13 - Redshift session tuning and cache validation are safety-oriented engine semantics
    "benchbox/platforms/snowflake.py:get_platform_info",              # 13 - platform info is assembled from independent account/runtime probes
    "benchbox/platforms/snowflake.py:apply_table_tunings",            # 13 - Snowflake tuning compares live clustering state before engine-specific DDL changes
    "benchbox/release/content_validation.py:_is_example_line",        # 13 — heuristic classifier must combine marker, table, bullet, and blockquote context
    "benchbox/utils/execution_manager.py:execute_power_runs",         # 13 - power-run orchestration manages warmups, monitoring, fail-fast, and stream permutations
    # --- Required complexity at CC=12 (2026-03-12 evaluation, wave 4) ---
    # Inherent complexity: sequential workflows, platform/operator dispatch, measurement protocols.
    "benchbox/cli/commands/run_official.py:run_official",             # 12 - deprecated CLI wrapper with TPC spec validation + argument assembly
    "benchbox/cli/commands/convert.py:convert",                      # 12 - sequential conversion workflow with distinct error handling per step
    "benchbox/cli/tuning.py:_run_simple_wizard",                     # 12 - platform-specific tuning dispatch (Databricks/Snowflake/BigQuery/Redshift)
    "benchbox/core/dataframe/benchmark_suite.py:_benchmark_query",   # 12 - interleaved measurement protocol (warmup/timing/memory/lazy evaluation)
    "benchbox/core/dataframe/benchmark_suite.py:_run_df_query",      # 12 - sequential benchmark phases (resolve/load/lookup/warmup/measure)
    "benchbox/core/metadata_primitives/benchmark.py:run_dataframe_benchmark",  # 12 - schema vs catalog operation dispatch with fundamentally different execution models
    "benchbox/core/metadata_primitives/dataframe_operations.py:execute_complex_type_introspection",  # 12 - platform-specific type APIs (Polars/Spark/Pandas)
    "benchbox/core/query_plans/insights.py:detect_antipatterns",     # 12 - semantically distinct antipattern checks (cross-join/conditions/depth/scans)
    "benchbox/core/query_plans/parsers/duckdb.py:_parse_json_node",  # 12 - operator-specific attribute extraction (scan/filter/join/aggregate/sort)
    "benchbox/core/read_primitives/benchmark.py:_load_data",         # 12 - sequential data loading with reuse check, schema creation, and per-table file resolution
    "benchbox/core/runner/runner.py:_execute_load_only_mode",        # 12 - external vs native table mode dispatch with mode-dependent execution paths
    "benchbox/core/tpcdi/etl/customer_mgmt_processor.py:parse_file", # 12 - XML action type dispatch (NEW/UPDCUST/ADDACCT/INACT) with namespace resilience
    # --- Required complexity at CC=12 (2026-03-12 follow-up, wave 4b) ---
    # Inherent complexity: spec workflows, platform-specific loading, runtime resolution, and SQL normalization.
    "benchbox/core/tpcds/c_tools.py:_clean_sql",                     # 12 - TPC-DS SQL normalization handles vendor-specific syntax families with distinct rewrites
    "benchbox/core/tpcds/dataframe_queries/rollup_helper.py:expand_rollup_pandas",  # 12 - rollup expansion must switch between grouped levels, grand totals, and Dask scalar handling
    "benchbox/core/tpcds/generator/streaming.py:_generate_parent_table_with_children",  # 12 - streaming generation coordinates subprocess output, compression, manifest updates, and cleanup
    "benchbox/core/tpcds_obt/query_conversion.py:_rewrite_columns",  # 12 - AST column rewrite must distinguish fact vs dimension mapping and OBT alias semantics
    "benchbox/core/tpch/generator.py:_generate_local",               # 12 - TPC-H data generation orchestrates validation, sample reuse, regeneration, and streaming/file output modes
    "benchbox/core/tpch/maintenance_test.py:run_maintenance_test",   # 12 - maintenance test follows the spec-required RF1/RF2 workflow with timed intervals and result accounting
    "benchbox/core/tpch/power_test.py:__init__",                     # 12 - seed/validation initialization reflects TPC-H answer-set rules and stream-specific validation constraints
    "benchbox/platforms/adapter_factory.py:get_adapter",             # 12 - unified platform factory must validate mode/deployment combinations before dispatch
    "benchbox/platforms/base/adapter.py:_create_enhanced_data_generation_phase",  # 12 - defensive per-table stats extraction handles arbitrary benchmark table containers
    "benchbox/platforms/clickhouse/tuning.py:generate_tuning_clause",  # 12 - ClickHouse tuning composes partition/order semantics with engine-specific precedence rules
    "benchbox/platforms/databend/adapter.py:_insert_files_batched",  # 12 - batch loader interleaves delimiter validation, row escaping, and flush boundaries
    "benchbox/platforms/databricks/credentials.py:setup_databricks_credentials",  # 12 - interactive credential setup mixes autodetect, secure prompting, validation, and optional defaults
    "benchbox/platforms/dataframe/benchmark_mixin.py:_get_tpcds_queries",  # 12 - TPC-DS stream permutation and variant fallback are spec-driven and registry-dependent
    "benchbox/platforms/datafusion.py:create_connection",            # 12 - DataFusion connection setup must branch across runtime APIs, spill config, and session options
    "benchbox/platforms/duckdb.py:validate_platform_capabilities",   # 12 - capability validator applies independent library/version/memory checks with structured warnings
    "benchbox/platforms/fabric_warehouse.py:load_data",              # 12 - Fabric loading coordinates table discovery, OneLake COPY, direct fallback, and platform-specific errors
    "benchbox/platforms/firebolt.py:_load_data_via_s3",              # 12 - Firebolt S3 staging combines upload, credential resolution, ingest SQL, and per-file accounting
    "benchbox/platforms/postgresql.py:validate_platform_capabilities",  # 12 - capability validator applies independent driver/version/memory checks with structured warnings
    "benchbox/platforms/redshift.py:__init__",                       # 12 - Redshift initialization validates dependencies, staging config, and connection defaults in one workflow
    "benchbox/platforms/starburst.py:_configure_starburst_defaults",  # 12 - Galaxy config must merge env defaults, role-qualified usernames, and required-auth validation
    "benchbox/utils/format_converters/delta_converter.py:convert",   # 12 - Delta conversion pipeline validates schema, reads shards, writes table metadata, and reports progress
    "benchbox/utils/format_converters/vortex_converter.py:convert",  # 12 - Vortex conversion pipeline validates schema, writes output, validates row counts, and cleans partial files
    "benchbox/utils/runtime_env.py:discover_isolated_runtime",       # 12 - isolated runtime discovery scans filesystem candidates, site-packages, and ABI compatibility
    "benchbox/utils/runtime_env.py:ensure_driver_version",           # 12 - driver resolution workflow must check current env, isolated runtimes, and optional auto-install fallbacks
]

[tool.benchbox.duplicates]
# Duplicate code detection configuration (scripts/check_duplicate_code.py).
# Uses AST structural hashing to find Type-2 clones (same structure, different
# identifiers). The threshold is a ceiling on total duplicated lines; raise it
# only when intentional duplication is added and lower it as clones are removed.
#
# Re-baseline protocol: when `make duplicate-check-verbose` reports actual
# duplicated lines consistently above the threshold, re-run the classification
# triage (see _project/DONE/core-functionality/active/
# duplicate-code-followup-triage-centralization.yaml for methodology) and set
# threshold = ceil(actual * 1.10).  Last re-baseline: 2026-03-12, 7,339 actual.
source_root = "benchbox"
min_lines = 10                   # minimum function body size to analyse
threshold = 8073                 # max duplicated lines before the check fails
exclude_patterns = [
    "tests/",                    # test files often have intentional boilerplate
    "__pycache__",
]
ignore_function_names = [
    # Structural clone patterns that differ by benchmark domain content
    "_load_queries",
    "_generate_default_params",
    # Facade chains / thin delegations
    "get_all_variants",
    # G5: dataclass-to-dict conversion methods on distinct config/artifact classes
    "to_dict",
    "get_etl_config",
    "get_performance_config",
]
ignore_rules = [
    # Platform adapter interface boilerplate - intentionally repeated per-adapter.
    # Lists explicit adapter files (excludes benchbox/platforms/base/ utilities).
    { function_names = ["from_config", "apply_constraint_configuration", "apply_unified_tuning", "supports_tuning_type", "apply_tuning_configuration"], path_contains = [
        "benchbox/platforms/athena.py", "benchbox/platforms/azure_synapse.py",
        "benchbox/platforms/bigquery.py", "benchbox/platforms/clickhouse_cloud.py",
        "benchbox/platforms/cudf.py", "benchbox/platforms/datafusion.py",
        "benchbox/platforms/doris.py", "benchbox/platforms/duckdb.py",
        "benchbox/platforms/fabric_lakehouse.py", "benchbox/platforms/fabric_warehouse.py",
        "benchbox/platforms/firebolt.py", "benchbox/platforms/lakesail.py",
        "benchbox/platforms/motherduck.py", "benchbox/platforms/pg_duckdb.py",
        "benchbox/platforms/pg_mooncake.py", "benchbox/platforms/polars_platform.py",
        "benchbox/platforms/postgresql.py", "benchbox/platforms/presto.py",
        "benchbox/platforms/questdb.py", "benchbox/platforms/redshift.py",
        "benchbox/platforms/snowflake.py", "benchbox/platforms/spark.py",
        "benchbox/platforms/sqlite.py", "benchbox/platforms/starburst.py",
        "benchbox/platforms/timescaledb.py", "benchbox/platforms/trino.py",
        "benchbox/platforms/snowpark_connect.py",
        "benchbox/platforms/aws/athena_spark_adapter.py",
        "benchbox/platforms/aws/emr_serverless_adapter.py",
        "benchbox/platforms/aws/glue_adapter.py",
        "benchbox/platforms/azure/fabric_spark_adapter.py",
        "benchbox/platforms/azure/synapse_spark_adapter.py",
        "benchbox/platforms/gcp/dataproc_adapter.py",
        "benchbox/platforms/gcp/dataproc_serverless_adapter.py",
        "benchbox/platforms/databricks/adapter.py",
        "benchbox/platforms/databricks/dataframe_adapter.py",
        "benchbox/platforms/databend/adapter.py",
        "benchbox/platforms/influxdb/adapter.py",
        "benchbox/platforms/onehouse/quanton_adapter.py",
        "benchbox/platforms/clickhouse/tuning.py",
        "benchbox/platforms/clickhouse/metadata.py",
        "benchbox/platforms/starrocks/tuning.py",
        "benchbox/platforms/starrocks/metadata.py",
        "benchbox/platforms/influxdb/metadata.py",
        "benchbox/platforms/dataframe/pandas_df.py",
        "benchbox/platforms/dataframe/polars_df.py",
        "benchbox/platforms/dataframe/datafusion_df.py",
        "benchbox/platforms/dataframe/cudf_df.py",
        "benchbox/platforms/dataframe/modin_df.py",
        "benchbox/platforms/dataframe/dask_df.py",
        "benchbox/platforms/dataframe/pyspark_df.py",
    ] },
    # DataFrame adapter/protocol casting boilerplate
    # G26: extend to cover context.py (where cast_date/cast_string actually live as protocol stubs)
    { function_names = ["cast_date", "cast_string"], path_contains = ["benchbox/platforms/dataframe/", "benchbox/core/dataframe/protocols.py", "benchbox/core/dataframe/context.py"] },
    # UnifiedFrame per-operation wrappers are intentional API surface methods
    { function_names = ["__call__", "count", "cum_sum", "cum_max", "cum_min", "cum_prod"], path_contains = ["benchbox/platforms/dataframe/unified_frame.py"] },
    # Protocol method signatures are intentionally repeated contracts
    { function_names = ["agg", "group_by", "first", "max", "min", "mean", "sum"], path_contains = ["benchbox/core/dataframe/protocols.py"] },
    # Top-level benchmark facade wrappers are intentional public API surface
    # w3-G8: extend with generate_data, get_all_aliases, get_benchmark_info and platform_registry
    { function_names = ["get_query", "get_queries", "get_create_tables_sql", "generate_data", "get_all_aliases", "get_benchmark_info"], path_contains = [
        "benchbox/amplab.py", "benchbox/clickbench.py", "benchbox/coffeeshop.py",
        "benchbox/datavault.py", "benchbox/h2odb.py", "benchbox/joinorder.py",
        "benchbox/metadata_primitives.py", "benchbox/nyctaxi.py", "benchbox/read_primitives.py",
        "benchbox/ssb.py", "benchbox/tpcdi.py", "benchbox/tpcds.py", "benchbox/tpcds_obt.py",
        "benchbox/tpch.py", "benchbox/tpch_skew.py", "benchbox/tpchavoc.py",
        "benchbox/transaction_primitives.py", "benchbox/tsbs_devops.py",
        "benchbox/write_primitives.py",
        "benchbox/core/platform_registry.py",
    ] },
    # Cloud/Spark adapter platform-info accessors are intentionally provider-specific
    { function_names = ["get_platform_info"], path_contains = [
        "benchbox/platforms/aws/athena_spark_adapter.py",
        "benchbox/platforms/aws/emr_serverless_adapter.py",
        "benchbox/platforms/aws/glue_adapter.py",
        "benchbox/platforms/azure/fabric_spark_adapter.py",
        "benchbox/platforms/azure/synapse_spark_adapter.py",
        "benchbox/platforms/gcp/dataproc_adapter.py",
        "benchbox/platforms/gcp/dataproc_serverless_adapter.py",
        "benchbox/platforms/databricks/adapter.py",
        "benchbox/platforms/databricks/dataframe_adapter.py",
        "benchbox/platforms/onehouse/quanton_adapter.py",
        "benchbox/platforms/snowpark_connect.py",
    ] },
    # CLI tuning defaults display commands are intentionally duplicated UX pathways
    { function_names = ["show_defaults"], path_contains = ["benchbox/cli/commands/df_tuning.py", "benchbox/cli/commands/tuning_group.py"] },

    # --- Boilerplate ignore rules added 2026-02-21 (groups G1-G28 from duplicate-check triage) ---

    # G8,G9,G11,G22: Protocol/ABC abstract method stubs across dataframe type hierarchy
    # G17: window function protocol stubs in context.py
    { function_names = ["col", "filter", "limit", "scalar_to_df", "can_provide", "create_schema", "date_add", "date_sub", "get_table", "table_exists", "lit", "list_tables", "concat", "cast_timestamp", "when", "window_dense_rank", "window_rank", "window_row_number"], path_contains = [
        "benchbox/core/dataframe/context.py",
        "benchbox/core/dataframe/protocols.py",
        "benchbox/core/operations/base.py",
        "benchbox/core/tpcdi/etl/sources.py",
        "benchbox/platforms/base/adapter.py",
        "benchbox/platforms/base/data_loading.py",
        "benchbox/utils/format_converters/base.py",
    ] },
    { function_names = ["_do_delete", "delete_rows", "generate_tuning_clauses"], path_contains = [
        "benchbox/core/dataframe/maintenance_interface.py",
        "benchbox/core/tuning/ddl_generator.py",
        "benchbox/platforms/dataframe/expression_family.py",
    ] },

    # G3,G10: DataFrame platform adapter thin wrappers (col/lit/from_pandas/detect_format/read_*)
    # G29: extend with get_row_count and pandas_family.py
    { function_names = ["_detect_format", "get_dataframe_operations", "from_pandas", "lit", "read_parquet", "to_datetime", "read_csv", "concat", "get_row_count"], path_contains = [
        "benchbox/platforms/dataframe/cudf_df.py",
        "benchbox/platforms/dataframe/dask_df.py",
        "benchbox/platforms/dataframe/datafusion_df.py",
        "benchbox/platforms/dataframe/modin_df.py",
        "benchbox/platforms/dataframe/pandas_df.py",
        "benchbox/platforms/dataframe/polars_df.py",
        "benchbox/platforms/dataframe/expression_family.py",
        "benchbox/platforms/dataframe/pandas_family.py",
        "benchbox/core/write_primitives/benchmark.py",
    ] },

    # G14: Cloud platform apply_platform_optimizations stubs (intentional per-platform future hooks)
    { function_names = ["apply_platform_optimizations"], path_contains = [
        "benchbox/platforms/bigquery.py",
        "benchbox/platforms/databricks/adapter.py",
        "benchbox/platforms/fabric_warehouse.py",
        "benchbox/platforms/firebolt.py",
        "benchbox/platforms/presto.py",
        "benchbox/platforms/redshift.py",
        "benchbox/platforms/snowflake.py",
        "benchbox/platforms/trino.py",
    ] },

    # G2,G5: Benchmark facade delegation methods (get_queries_by_category, get_query_info, etc.)
    { function_names = ["get_queries_by_category", "get_query_info", "get_operations_by_category", "supports_operation", "is_setup", "run_data_validation", "sql", "get_data_source_benchmark", "get_tunings", "get_available_tables"], path_contains = [
        "benchbox/core/ai_primitives/benchmark.py",
        "benchbox/core/joinorder/benchmark.py",
        "benchbox/core/metadata_primitives/benchmark.py",
        "benchbox/core/nyctaxi/benchmark.py",
        "benchbox/core/read_primitives/benchmark.py",
        "benchbox/core/transaction_primitives/benchmark.py",
        "benchbox/core/tsbs_devops/benchmark.py",
        "benchbox/core/tuning/interface.py",
        "benchbox/core/write_primitives/benchmark.py",
        "benchbox/nyctaxi.py",
        "benchbox/tpcdi.py",
        "benchbox/tpcds.py",
        "benchbox/tpch.py",
        "benchbox/tsbs_devops.py",
        "benchbox/write_primitives.py",
    ] },
    { function_names = ["execute_rule", "get_operation", "get_table_row_count"], path_contains = [
        "benchbox/core/ai_primitives/dataframe_operations.py",
        "benchbox/core/joinorder/benchmark.py",
        "benchbox/core/metadata_primitives/dataframe_operations.py",
        "benchbox/core/tpcdi/etl/data_quality_monitor.py",
        "benchbox/core/transaction_primitives/benchmark.py",
        "benchbox/core/transaction_primitives/dataframe_operations.py",
        "benchbox/core/write_primitives/benchmark.py",
        "benchbox/core/write_primitives/dataframe_operations.py",
        "benchbox/platforms/dataframe/datafusion_df.py",
        "benchbox/platforms/dataframe/pyspark_df.py",
        "benchbox/tpcdi.py",
        "benchbox/transaction_primitives.py",
        "benchbox/write_primitives.py",
    ] },

    # G16: Benchmark interface conformance (get_schema, get_csv_loading_config)
    { function_names = ["get_schema", "get_csv_loading_config"], path_contains = [
        "benchbox/core/amplab/benchmark.py",
        "benchbox/core/clickbench/benchmark.py",
        "benchbox/core/h2odb/benchmark.py",
        "benchbox/core/read_primitives/benchmark.py",
        "benchbox/core/ssb/benchmark.py",
        "benchbox/core/transaction_primitives/benchmark.py",
        "benchbox/platforms/base/cloud_spark/mixins.py",
    ] },

    # G15: Benchmark base empty_results and table name accessors
    { function_names = ["_empty_results", "_get_incremental_tables", "_known_table_names"], path_contains = [
        "benchbox/base.py",
        "benchbox/core/base_benchmark.py",
        "benchbox/core/joinorder/schema.py",
        "benchbox/core/metadata_primitives/benchmark.py",
        "benchbox/core/tpc_patterns.py",
        "benchbox/core/tpcdi/etl/incremental_loader.py",
        "benchbox/core/tpcds/benchmark/runner.py",
        "benchbox/core/tpcds/c_tools.py",
        "benchbox/core/tpcds/generator/manager.py",
    ] },

    # G1: TPC-DS OBT channel-dispatch configuration tables (same structure, different domain data)
    { function_names = ["_fact_map", "_income_band_aliases", "_role_aliases"], path_contains = ["benchbox/core/tpcds_obt/etl/transformer.py"] },

    # G6,G7: TPC-DI query catalog builders and selector methods
    # w3-G9: extend with get_queries_by_category
    { function_names = ["_load_base_queries", "_load_etl_queries", "get_queries_by_complexity", "get_queries_by_frequency", "get_queries_by_severity", "_load_base_query_metadata", "_load_etl_metadata", "get_queries_by_category"], path_contains = [
        "benchbox/core/tpcdi/queries.py",
        "benchbox/core/tpcdi/query_etl.py",
        "benchbox/core/tpcdi/query_analytics.py",
        "benchbox/core/tpcdi/query_validation.py",
    ] },

    # G26,G28: TPC-DS spec-mandated query families (Q33/Q60 three-channel pattern)
    { function_names = ["q33_expression_impl", "q60_expression_impl", "q33_pandas_impl", "q60_pandas_impl"], path_contains = ["benchbox/core/tpcds/dataframe_queries/queries.py"] },

    # Top-level benchmark module __init__ and execute_query boilerplate
    # G22: extend with facade __init__ on datavault.py, tpcds.py, tpch.py
    # w3-G21: extend with __init__ constructors sharing attribute-assignment AST shape
    { function_names = ["__init__", "execute_query"], path_contains = [
        "benchbox/amplab.py", "benchbox/clickbench.py", "benchbox/coffeeshop.py",
        "benchbox/datavault.py", "benchbox/h2odb.py", "benchbox/ssb.py",
        "benchbox/tpcds.py", "benchbox/tpch.py",
        "benchbox/core/amplab/benchmark.py",
        "benchbox/core/clickbench/benchmark.py",
        "benchbox/core/ssb/benchmark.py",
        "benchbox/platforms/base/data_loading.py",
        "benchbox/platforms/dataframe/unified_frame.py",
    ] },

    # Schema model get_create_table_sql - different DDL per benchmark, same structural pattern
    { function_names = ["get_create_table_sql"], path_contains = [
        "benchbox/core/datavault/schema.py",
        "benchbox/core/tpcds/schema/models.py",
        "benchbox/core/tpch/schema.py",
    ] },

    # Spark/Starburst config builders - platform-specific configuration, same structural pattern
    { function_names = ["_build_spark_config", "_build_starburst_config"], path_contains = [
        "benchbox/platforms/spark.py",
        "benchbox/platforms/starburst.py",
    ] },

    # G18: UnifiedFrame multi-backend dispatch methods
    { function_names = ["first", "floor", "last", "abs_", "round_", "ceil", "sqrt", "log", "quantile"], path_contains = ["benchbox/platforms/dataframe/unified_frame.py"] },

    # G20: ExpressionFamily multi-backend aggregation dispatch and interface methods
    { function_names = ["max_", "std", "var", "date_sub", "date_add", "calculate_official_metrics"], path_contains = [
        "benchbox/platforms/dataframe/expression_family.py",
        "benchbox/core/tpcds/dataframe_queries/parameters.py",
        "benchbox/core/tpchavoc/benchmark.py",
    ] },

    # Data loader column name accessors (TPC-H/TPC-DS spec-defined column lists)
    { function_names = ["get_tpcds_column_names", "get_tpch_column_names"], path_contains = ["benchbox/core/dataframe/data_loader.py"] },

    # Query catalog/generator init boilerplate (different domain data, same dict-building structure)
    { function_names = ["_initialize_variant_generators", "get_query_categories", "get_file_format_info"], path_contains = [
        "benchbox/core/clickbench/queries.py",
        "benchbox/core/tpcdi/source_generators.py",
        "benchbox/core/tpchavoc/queries.py",
    ] },

    # --- Boilerplate ignore rules added 2026-03-12 (wave-2 triage) ---

    # G2: Protocol operation support methods across primitives dataframe_operations and platform adapters
    { function_names = ["supports_operation", "is_setup", "sql"], path_contains = [
        "benchbox/core/ai_primitives/dataframe_operations.py",
        "benchbox/core/metadata_primitives/dataframe_operations.py",
        "benchbox/core/transaction_primitives/dataframe_operations.py",
        "benchbox/core/write_primitives/dataframe_operations.py",
        "benchbox/platforms/dataframe/datafusion_df.py",
        "benchbox/platforms/dataframe/pyspark_df.py",
        "benchbox/transaction_primitives.py",
    ] },

    # G6,G19,G30: TPC-DS spec-mandated structurally similar query pairs (Q62/Q99, Q38/Q87)
    # w3-G27: extend with Q7/Q26 spec pair (promotional item-joined fact table aggregation)
    { function_names = ["q62_expression_impl", "q99_expression_impl", "q38_expression_impl", "q87_expression_impl", "q62_pandas_impl", "q99_pandas_impl", "q7_expression_impl", "q26_expression_impl"], path_contains = ["benchbox/core/tpcds/dataframe_queries/"] },

    # G14: Compression protocol stubs matching canonical implementations
    { function_names = ["compress_file", "decompress_file", "open_for_read"], path_contains = [
        "benchbox/core/dataframe/context.py",
        "benchbox/core/dataframe/protocols.py",
        "benchbox/utils/compression.py",
    ] },

    # G21: TPC-DS channel union operations in channel_helper.py (domain-specific, spec-driven)
    { function_names = ["union_sales_channels_pandas", "union_returns_channels_pandas"], path_contains = ["benchbox/core/tpcds/dataframe_queries/channel_helper.py"] },

    # G25: AST false positive - different purposes but similar nested dict/list structure
    { function_names = ["create_sample_test_results", "get_extended_table_order"], path_contains = [
        "benchbox/core/tpc_validation.py",
        "benchbox/core/tpcdi/schema_extensions.py",
    ] },

    # G27: Thin delegation methods on facade/tuning classes
    { function_names = ["get_stream_info", "get_table_info", "get_table_tuning"], path_contains = [
        "benchbox/core/joinorder/benchmark.py",
        "benchbox/core/tuning/interface.py",
        "benchbox/tpcdi.py",
        "benchbox/tpcds.py",
        "benchbox/tpch.py",
    ] },

    # --- Wave-3 ignore rules (2026-03-12): wave-2 extraction residuals ---

    # w3-G6: Schema get_all_create_table_sql wrappers (residual after collect_create_table_sql extraction)
    { function_names = ["get_all_create_table_sql"], path_contains = [
        "benchbox/core/amplab/schema.py",
        "benchbox/core/read_primitives/schema.py",
        "benchbox/core/ssb/schema.py",
    ] },

    # w3-G19: DataFrame manager factory wrappers (residual after _get_dataframe_manager extraction)
    { function_names = ["get_dataframe_metadata_manager", "get_dataframe_transaction_manager", "get_dataframe_write_manager"], path_contains = [
        "benchbox/core/metadata_primitives/dataframe_operations.py",
        "benchbox/core/transaction_primitives/dataframe_operations.py",
        "benchbox/core/write_primitives/dataframe_operations.py",
    ] },

    # --- Wave-3 ignore rules: AST false positives ---

    # w3-G3: AST false positive - 5 different function names sharing facade/delegation AST shape
    { function_names = ["classify_error", "date_add", "date_sub", "run_full_benchmark", "setup"], path_contains = [
        "benchbox/core/tpcdi/etl/error_recovery.py",
        "benchbox/platforms/dataframe/pandas_family.py",
        "benchbox/tpcdi.py",
        "benchbox/transaction_primitives.py",
        "benchbox/write_primitives.py",
    ] },

    # w3-G22: AST false positive - 5 different function names sharing accessor/delegation AST shape
    { function_names = ["get", "get_variant_description", "run_etl_benchmark", "calculate_official_metrics", "execute_operation"], path_contains = [
        "benchbox/core/tpcds/dataframe_queries/parameters.py",
        "benchbox/core/tpchavoc/benchmark.py",
        "benchbox/tpcdi.py",
        "benchbox/transaction_primitives.py",
    ] },

    # --- Wave-3 ignore rules: intentional boilerplate patterns ---

    # w3-G2: H2O.db spec-driven aggregation query variants (column names/cardinality differ)
    { function_names = ["aggregation_distinct_groupby_expression_impl", "aggregation_distinct_groupby_pandas_impl", "aggregation_groupby_large_expression_impl", "aggregation_groupby_large_pandas_impl"], path_contains = ["benchbox/core/read_primitives/dataframe_queries.py"] },

    # w3-G5: Primitives query/operation catalog lookup methods (type-variant: dict[str, str] vs dict[str, Operation])
    { function_names = ["get_operations_by_category", "get_queries_by_category"], path_contains = [
        "benchbox/core/ai_primitives/queries.py",
        "benchbox/core/metadata_primitives/queries.py",
        "benchbox/core/read_primitives/queries.py",
        "benchbox/core/transaction_primitives/operations.py",
        "benchbox/core/write_primitives/operations.py",
    ] },

    # w3-G7: Capability-matrix methods in metadata_primitives (each defines different platform set)
    { function_names = ["supports_acl_introspection", "supports_column_grants", "supports_role_hierarchy"], path_contains = ["benchbox/core/metadata_primitives/ddl.py"] },
]