Smart Filtering

Generated Code Filtering

art-dupl automatically filters generated code by default. This removes noise from:

• sqlc — SQL-to-Go code generator
• templ — Templ-generated *_templ.go files
• protobuf — .pb.go, _grpc.pb.go files
• mockgen — Mock generation framework
• stringer — String method generation
• generic — Any file with a “Code generated by” comment

Override Filters

Include specific generated categories with --include-generated:

art-dupl --include-generated sqlc           # Include sqlc-generated
art-dupl --include-generated templ,protobuf # Multiple categories
art-dupl --include-generated all            # Everything

Available categories: sqlc, templ, protobuf, mockgen, stringer, generic, all.

Filter Override Behavior

--include-generated sqlc disables FilterSQLC but the generic filter (any “Code generated by” comment) still catches the file unless it was explicitly included. The shouldIncludeFile function in cmd/util.go handles this logic.

File Type Filtering

art-dupl --only go        # .go files only
art-dupl --only templ     # .templ files only

By default, both .go and .templ files are analyzed.

Custom Patterns

art-dupl --include-pattern "internal/*"     # Include specific paths
art-dupl --exclude-pattern "*_test.go"      # Exclude test files
art-dupl --exclude-pattern "vendor/*"       # Exclude vendor (default)

Multiple patterns can be specified by repeating the flag.

Directory Exclusions

By default, these directories are excluded:

• vendor/
• .git/
• node_modules/

Include them with:

art-dupl --vendor                    # Include vendor directory
art-dupl --include-node-modules      # Include node_modules (hash detection)

Actionability Filtering

art-dupl automatically suppresses clone groups that represent Go boilerplate rather than real duplication:

• Signature-only clones
• Interface implementations
• RAII defer patterns
• Error propagation (err := ...; if err != nil { return ... })
• Test data pairs
• Table-driven test scaffolding
• Data-dominated blocks
• Builder callback patterns
• Cobra CLI boilerplate
• Assertion chains
• Single call expressions
• Assign + error check

A group is marked NonActionable only when EVERY clone matches the same boilerplate pattern. These are still counted in statistics but suppressed from default output.

Test File Threshold

art-dupl --test-threshold 20    # Separate threshold for test files (0 = max(30, threshold))

Test files often have legitimate repetition. The test threshold defaults to max(30, threshold) to reduce noise.