mizchi MoonBit Code Similarity Detection
What to do
Run similarity-mbt on the target MoonBit project to detect duplicate functions, then analyze the results and propose a refactoring plan.
If similarity-mbt is not installed:
cargo install similarity-mbt
Step 1: Run similarity analysis
Run with the arguments provided, or use sensible defaults:
similarity-mbt $ARGUMENTS
If no arguments are given, scan the current directory:
similarity-mbt . --threshold 0.85 --min-lines 5
Step 2: Analyze the results
Read the output and categorize duplicate pairs:
High-priority refactoring targets
- 100% similarity: Exact structural duplicates. Extract into a shared function with parameters.
- 95-100% similarity: Near-identical. Usually differ by 1-2 identifiers or values. Parameterize the difference.
- X/Y axis pairs (e.g.
get_widthvsget_height,offset_xvsoffset_y): Common in layout/graphics code. Abstract with an axis parameter or generic.
Medium-priority
- 85-95% similarity: Same algorithm with minor structural differences. Consider extracting the common pattern.
- async/sync variants: Same logic duplicated for async and sync versions. Consider a shared core function.
_with_optionsvariants (e.g.parse_rulevsparse_rule_with_diagnostics): Use optional parameters or builder pattern.
Low-priority / acceptable
- Protocol handler boilerplate: Large sets of nearly identical dispatch functions. Better solved with code generation or macros if available.
- Small accessors (< 5 lines, 85-90%): Short functions naturally look similar. Usually not worth refactoring.
Step 3: Propose refactoring
For each high-priority pair, suggest a concrete refactoring:
- Show the two duplicate functions
- Propose the unified version
- Explain what callers need to change
Key CLI Options
| Option | Description |
|---|---|
--threshold <0-1> |
Similarity threshold (default: 0.85) |
--min-lines <n> |
Skip functions shorter than n lines (default: 3) |
--min-tokens <n> |
Skip functions with fewer than n AST nodes |
--print |
Show actual code snippets in output |
--filter-function <name> |
Only show pairs matching function name |
--filter-function-body <text> |
Only show pairs whose body contains text |
--fail-on-duplicates |
Exit code 1 if duplicates found (CI use) |
--rename-cost <0-1> |
APTED rename cost (default: 0.3, lower = more tolerant of renames) |
--no-size-penalty |
Disable penalty for differently-sized functions |
Effective Thresholds
0.95+: Nearly identical (variable renames only)0.85-0.95: Same algorithm, minor differences0.75-0.85: Similar structure, different details0.70-0.75: Related logic, worth investigating
Workflow tips
- Start with
--threshold 0.9to find obvious duplicates, then lower to 0.85 - Use
--printto see actual code for top pairs - Use
--min-lines 5to skip trivial accessors - Use
--filter-function <name>to focus on a specific function - For CI:
similarity-mbt . --threshold 0.95 --fail-on-duplicates --min-lines 5
Common MoonBit refactoring patterns
Axis parameterization
// Before: two functions
fn get_width(self) -> Double { ... }
fn get_height(self) -> Double { ... }
// After: one function with axis parameter
fn get_dimension(self, axis : Axis) -> Double { ... }
Operation abstraction
// Before: set_attribute / remove_attribute with identical structure
// After: shared body with operation parameter or closure
fn modify_attribute(self, node, name, op : (Map[String, String], String) -> Unit) -> Result[Unit, Error] { ... }
Optional parameter consolidation
// Before: parse_rule / parse_rule_with_diagnostics
// After: single function with optional diagnostics collector
fn parse_rule(self, diagnostics~ : Array[Diagnostic]? = None) -> Rule { ... }