goccy/tobari/tobari-duplicated-tests-remover

Remove Duplicated Tests

This skill analyzes test coverage data from tobari to find and remove duplicate test cases that cover nearly identical code paths.

Prerequisites

Ensure you have run tests with tobari enabled:

GOFLAGS="$(tobari flags)" go test ./...

How to Read Coverage Data

1. Check if TOBARI_COVERDIR environment variable is set:

``bash echo $TOBARI_COVERDIR ``

1. If set, read $TOBARI_COVERDIR/tobari/tobari.toon
2. If not set or empty, read ./tobari/tobari.toon from current directory

TOON Format Parsing

The tobari.toon file uses Token-Oriented Object Notation format:

TestName[N]{FileName,StartLine,StartCol,EndLine,EndCol,StatementCount,Count}:
	/path/to/file.go,7,24,9,2,1,4
	/path/to/file.go,11,29,12,22,1,0

• Top-level key is the test name (e.g., Add corresponds to TestAdd function)
• [N] indicates the number of entries
• Each indented line (starting with tab) is a coverage entry with comma-separated values:
• FileName: source file path
• StartLine, StartCol: start position
• EndLine, EndCol: end position
• StatementCount: number of statements in this block
• Count: execution count (0 = not covered)

Reports with passedBlocksOnly

If tobari.toon ends with a passedBlocksOnly[N]: section, the tests listed there come from a binary built with --passed-blocks-only: their per-test lists hold only blocks that were actually passed, so none of their entries has Count = 0. Do not conclude that everything is covered. Derive the full set of blocks from the all[N]: list in the metadata: section instead (each line is FileName,StartLine,StartCol,EndLine,EndCol,StatementCount), and treat a block of all that appears in no test as not covered. If the metadata: section also has a sources[N]{Source,Files}: list and the file ends with a source[N]{Name,Source}: section, the report merges several programs: restrict all to the files of the test's own source.

Detect Duplicate Test Cases

Compare coverage entries between all test cases to find tests that cover nearly identical code paths.

Calculate Match Rate

For each pair of tests (TestA, TestB):

1. Create a set of "coverage signatures" for each test:

• Signature = FileName:StartLine:StartCol:EndLine:EndCol:IsCovered
• IsCovered = 1 if Count > 0, else 0

1. Calculate match rate:

`` CommonSignatures = signatures in both TestA and TestB with same IsCovered value AllSignatures = union of all signatures from both tests MatchRate = len(CommonSignatures) / len(AllSignatures) * 100 ``

1. If MatchRate > 95%, these tests are considered duplicates

Action for Duplicates

If duplicate test pairs are found:

1. List all test pairs with >95% match rate, showing:

• Test names
• Match percentage
• Number of shared coverage entries

1. Use AskUserQuestion to ask the user:

• Question: "The following test pairs cover almost identical code paths. Which test(s) would you like to remove?"
• Options for each duplicate pair, plus "Keep all tests"

1. If user selects tests to remove:

• Find the test file containing those tests
• Delete the selected test functions
• Re-run tests with tobari to verify:

``bash GOFLAGS="$(tobari flags)" go test ./... ``

1. If no duplicates found, inform the user that all tests have unique coverage patterns

Guidelines

• Before removing tests, ensure they don't have different assertions even if coverage is similar
• Consider if tests document different use cases or edge cases
• Table-driven tests with multiple cases may show as duplicates - review the test logic before removal