feat: add sequence command for Mermaid sequence diagrams#345
feat: add sequence command for Mermaid sequence diagrams#345carlos-alm wants to merge 9 commits intomainfrom
Conversation
Sequence diagram (generated by this PR)
sequenceDiagram
participant builder as builder.js
participant ast as ast.js
participant cfg as cfg.js
participant complexity as complexity.js
participant config as config.js
participant constants as constants.js
participant dataflow as dataflow.js
participant db as db.js
participant journal as journal.js
participant logger as logger.js
participant native as native.js
participant parser as parser.js
participant registry as registry.js
participant resolve as resolve.js
participant structure as structure.js
builder->>db: openDb
builder->>db: initSchema
builder->>config: loadConfig
builder->>parser: getActiveEngine
builder->>logger: info
builder->>db: getBuildMeta
builder->>builder: loadPathAliases
builder->>constants: normalizePath
builder->>builder: collectFiles
builder->>builder: getChangedFiles
builder->>parser: parseFilesAuto
builder->>cfg: buildCFGData
builder->>dataflow: buildDataflowEdges
builder->>db: closeDb
builder->>journal: writeJournalHeader
builder->>logger: debug
builder->>builder: purgeFilesFromGraph
builder->>builder: readFileSafe
builder->>builder: fileStat
builder->>builder: fileHash
builder->>resolve: resolveImportsBatch
builder->>builder: getResolved
builder->>builder: isBarrelFile
builder->>builder: resolveBarrelExport
builder->>resolve: computeConfidence
builder->>structure: buildStructure
builder->>structure: classifyNodeRoles
builder->>ast: buildAstNodes
builder->>complexity: buildComplexityMetrics
builder->>parser: ensureWasmTrees
builder->>logger: warn
builder->>db: setBuildMeta
builder->>registry: registerRepo
db->>db: acquireAdvisoryLock
db->>logger: debug
config->>logger: debug
config->>config: mergeConfig
config->>config: resolveSecrets
config->>config: applyEnvOverrides
parser->>parser: resolveEngine
cfg->>parser: createParsers
cfg->>complexity: findFunctionNode
cfg->>cfg: buildFunctionCFG
cfg->>logger: info
dataflow->>parser: createParsers
dataflow->>dataflow: extractDataflow
dataflow->>dataflow: resolveNode
dataflow->>logger: info
db->>db: releaseAdvisoryLock
journal->>logger: warn
resolve->>native: loadNative
resolve->>resolve: convertAliasesForNative
resolve->>constants: normalizePath
structure->>constants: normalizePath
structure->>logger: debug
structure->>structure: median
ast->>logger: debug
ast->>ast: findParentDef
ast->>ast: walkAst
complexity->>parser: createParsers
complexity->>parser: getParser
complexity->>complexity: findFunctionNode
complexity->>complexity: computeAllMetrics
complexity->>logger: info
parser->>parser: createParsers
registry->>registry: loadRegistry
registry->>registry: saveRegistry
registry->>logger: debug
note right of builder: Truncated at depth 3
|
Greptile SummaryThis PR adds a Key finding:
Confidence Score: 2/5
Flowchart%%{init: {'theme': 'neutral'}}%%
flowchart TD
CLI["CLI: sequence <name>"]
MCP["MCP tool: sequence"]
API["Programmatic API\n(src/index.js)"]
CLI -->|calls| SEQ["sequence()"]
MCP -->|calls| SEQD["sequenceData()"]
API -->|exports| SEQD
SEQ --> SEQD
SEQ -->|mermaid output| MERM["sequenceToMermaid()"]
SEQD --> OPEN["openReadonlyOrFail()"]
SEQD --> FBM["findBestMatch()\n(duplicated from flow.js)"]
SEQD --> BFS["BFS forward traversal\n(edges WHERE kind='calls')"]
SEQD --> DF{"withDataflow?"}
DF -->|yes| DFA["Annotate params + return arrows\n(dataflow table)"]
DF -->|no| SORT["Sort messages by depth"]
DFA --> SORT
SORT --> ALIAS["buildAliases()\n⚠️ join('/') produces invalid IDs"]
ALIAS --> PAG["paginateResult()\n+ filter orphaned participants"]
PAG --> RES["{ entry, participants,\nmessages, truncated }"]
MERM -->|escapeMermaid labels| OUT["sequenceDiagram\nparticipant …\nA->>B: label"]
Last reviewed commit: 2f7af10 |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
src/sequence.js
Outdated
| for (let d = 1; d <= maxDepth; d++) { | ||
| const nextFrontier = []; | ||
|
|
||
| for (const fid of frontier) { | ||
| const callees = db | ||
| .prepare( | ||
| `SELECT DISTINCT n.id, n.name, n.kind, n.file, n.line | ||
| FROM edges e JOIN nodes n ON e.target_id = n.id | ||
| WHERE e.source_id = ? AND e.kind = 'calls'`, | ||
| ) | ||
| .all(fid); |
There was a problem hiding this comment.
db.prepare() is called inside the nested BFS loop (lines 201–207), recompiling the SQL statement on every frontier node in every depth level. better-sqlite3 does not automatically cache Statement objects — each call allocates a new one. For large graphs, this adds unnecessary overhead.
Hoist the prepared statement above the outer for (let d = …) loop:
| for (let d = 1; d <= maxDepth; d++) { | |
| const nextFrontier = []; | |
| for (const fid of frontier) { | |
| const callees = db | |
| .prepare( | |
| `SELECT DISTINCT n.id, n.name, n.kind, n.file, n.line | |
| FROM edges e JOIN nodes n ON e.target_id = n.id | |
| WHERE e.source_id = ? AND e.kind = 'calls'`, | |
| ) | |
| .all(fid); | |
| const getCallees = db.prepare( | |
| `SELECT DISTINCT n.id, n.name, n.kind, n.file, n.line | |
| FROM edges e JOIN nodes n ON e.target_id = n.id | |
| WHERE e.source_id = ? AND e.kind = 'calls'`, | |
| ); | |
| for (let d = 1; d <= maxDepth; d++) { | |
| const nextFrontier = []; | |
| for (const fid of frontier) { | |
| const callees = getCallees.all(fid); |
There was a problem hiding this comment.
Fixed — hoisted the \ prepared statement above the outer \ loop so it's compiled once and reused across all BFS iterations.
| for (const msg of [...messages]) { | ||
| if (msg.type !== 'call') continue; | ||
| const targetNode = [...idToNode.values()].find( | ||
| (n) => n.name === msg.label && n.file === msg.to, | ||
| ); | ||
| if (!targetNode) continue; | ||
|
|
||
| const returnKey = `${msg.to}->${msg.from}:${msg.label}`; | ||
| if (seenReturns.has(returnKey)) continue; | ||
|
|
||
| const returns = db | ||
| .prepare( | ||
| `SELECT d.expression FROM dataflow d | ||
| WHERE d.source_id = ? AND d.kind = 'returns'`, | ||
| ) | ||
| .all(targetNode.id); | ||
|
|
||
| if (returns.length > 0) { | ||
| seenReturns.add(returnKey); | ||
| const expr = returns[0].expression || 'result'; | ||
| messages.push({ | ||
| from: msg.to, | ||
| to: msg.from, | ||
| label: expr, | ||
| type: 'return', | ||
| depth: msg.depth, | ||
| }); | ||
| } | ||
| } | ||
|
|
||
| // Annotate call messages with parameter names | ||
| for (const msg of messages) { | ||
| if (msg.type !== 'call') continue; | ||
| const targetNode = [...idToNode.values()].find( | ||
| (n) => n.name === msg.label && n.file === msg.to, | ||
| ); | ||
| if (!targetNode) continue; | ||
|
|
||
| const params = db | ||
| .prepare( | ||
| `SELECT d.expression FROM dataflow d | ||
| WHERE d.target_id = ? AND d.kind = 'flows_to' | ||
| ORDER BY d.param_index`, | ||
| ) | ||
| .all(targetNode.id); | ||
|
|
||
| if (params.length > 0) { | ||
| const paramNames = params | ||
| .map((p) => p.expression) | ||
| .filter(Boolean) | ||
| .slice(0, 3); | ||
| if (paramNames.length > 0) { | ||
| msg.label = `${msg.label}(${paramNames.join(', ')})`; | ||
| } | ||
| } | ||
| } |
There was a problem hiding this comment.
[...idToNode.values()].find() is called inside two separate loops over messages (lines 251–253 and lines 282–284), scanning the entire idToNode map linearly for every message. With M messages and N nodes, this becomes O(M × N).
Build a secondary lookup map keyed on ${name}|${file} once before the dataflow annotation block to reduce both lookups to O(1):
| for (const msg of [...messages]) { | |
| if (msg.type !== 'call') continue; | |
| const targetNode = [...idToNode.values()].find( | |
| (n) => n.name === msg.label && n.file === msg.to, | |
| ); | |
| if (!targetNode) continue; | |
| const returnKey = `${msg.to}->${msg.from}:${msg.label}`; | |
| if (seenReturns.has(returnKey)) continue; | |
| const returns = db | |
| .prepare( | |
| `SELECT d.expression FROM dataflow d | |
| WHERE d.source_id = ? AND d.kind = 'returns'`, | |
| ) | |
| .all(targetNode.id); | |
| if (returns.length > 0) { | |
| seenReturns.add(returnKey); | |
| const expr = returns[0].expression || 'result'; | |
| messages.push({ | |
| from: msg.to, | |
| to: msg.from, | |
| label: expr, | |
| type: 'return', | |
| depth: msg.depth, | |
| }); | |
| } | |
| } | |
| // Annotate call messages with parameter names | |
| for (const msg of messages) { | |
| if (msg.type !== 'call') continue; | |
| const targetNode = [...idToNode.values()].find( | |
| (n) => n.name === msg.label && n.file === msg.to, | |
| ); | |
| if (!targetNode) continue; | |
| const params = db | |
| .prepare( | |
| `SELECT d.expression FROM dataflow d | |
| WHERE d.target_id = ? AND d.kind = 'flows_to' | |
| ORDER BY d.param_index`, | |
| ) | |
| .all(targetNode.id); | |
| if (params.length > 0) { | |
| const paramNames = params | |
| .map((p) => p.expression) | |
| .filter(Boolean) | |
| .slice(0, 3); | |
| if (paramNames.length > 0) { | |
| msg.label = `${msg.label}(${paramNames.join(', ')})`; | |
| } | |
| } | |
| } | |
| // Build name|file lookup for O(1) target node access | |
| const nodeByNameFile = new Map(); | |
| for (const n of idToNode.values()) { | |
| nodeByNameFile.set(`${n.name}|${n.file}`, n); | |
| } | |
| for (const msg of [...messages]) { | |
| if (msg.type !== 'call') continue; | |
| const targetNode = nodeByNameFile.get(`${msg.label}|${msg.to}`); | |
| if (!targetNode) continue; |
And similarly in the second loop (line 280–284) replace the find with:
const targetNode = nodeByNameFile.get(`${msg.label}|${msg.to}`);There was a problem hiding this comment.
Fixed — built a nodeByNameFile Map keyed on ${name}|${file} from idToNode.values() before the dataflow annotation block. Both loops now use O(1) .get() instead of O(N) .find().
| // Dataflow annotations: add return arrows | ||
| if (withDataflow && messages.length > 0) { | ||
| const hasTable = db | ||
| .prepare("SELECT name FROM sqlite_master WHERE type='table' AND name='dataflow'") | ||
| .get(); | ||
|
|
||
| if (hasTable) { | ||
| // For each called function, check if it has return edges | ||
| const seenReturns = new Set(); | ||
| for (const msg of [...messages]) { | ||
| if (msg.type !== 'call') continue; | ||
| const targetNode = [...idToNode.values()].find( | ||
| (n) => n.name === msg.label && n.file === msg.to, | ||
| ); | ||
| if (!targetNode) continue; | ||
|
|
||
| const returnKey = `${msg.to}->${msg.from}:${msg.label}`; | ||
| if (seenReturns.has(returnKey)) continue; | ||
|
|
||
| const returns = db | ||
| .prepare( | ||
| `SELECT d.expression FROM dataflow d | ||
| WHERE d.source_id = ? AND d.kind = 'returns'`, | ||
| ) | ||
| .all(targetNode.id); | ||
|
|
||
| if (returns.length > 0) { | ||
| seenReturns.add(returnKey); | ||
| const expr = returns[0].expression || 'result'; | ||
| messages.push({ | ||
| from: msg.to, | ||
| to: msg.from, | ||
| label: expr, | ||
| type: 'return', | ||
| depth: msg.depth, | ||
| }); | ||
| } | ||
| } | ||
|
|
||
| // Annotate call messages with parameter names | ||
| for (const msg of messages) { | ||
| if (msg.type !== 'call') continue; | ||
| const targetNode = [...idToNode.values()].find( | ||
| (n) => n.name === msg.label && n.file === msg.to, | ||
| ); | ||
| if (!targetNode) continue; | ||
|
|
||
| const params = db | ||
| .prepare( | ||
| `SELECT d.expression FROM dataflow d | ||
| WHERE d.target_id = ? AND d.kind = 'flows_to' | ||
| ORDER BY d.param_index`, | ||
| ) | ||
| .all(targetNode.id); | ||
|
|
||
| if (params.length > 0) { | ||
| const paramNames = params | ||
| .map((p) => p.expression) | ||
| .filter(Boolean) | ||
| .slice(0, 3); | ||
| if (paramNames.length > 0) { | ||
| msg.label = `${msg.label}(${paramNames.join(', ')})`; | ||
| } | ||
| } | ||
| } | ||
| } | ||
| } |
There was a problem hiding this comment.
Three separate db.prepare() calls (returns query at line 259 and flows_to query at line 287) are compiled inside loops over messages. Like the BFS loop, these should be hoisted outside so the statement is compiled once and reused.
Additionally, the current approach generates 2 × M database round-trips (one returns query + one flows_to query per message). Consider batching with WHERE source_id IN (…) to reduce database calls and improve performance on large message arrays.
There was a problem hiding this comment.
Fixed — hoisted both getReturns and getFlowsTo prepared statements outside the message loops, right after building the nodeByNameFile map. Statements are now compiled once and reused for all messages.
Before calling ensureWasmTrees, check whether native engine already supplies CFG and dataflow data for all files. When it does, skip the WASM pre-parse entirely — avoiding a full WASM parse of every file on native builds where the data is already available. Impact: 1 functions changed, 14 affected
…omplete changelog
Impact: 1 functions changed, 0 affected
d.cfg?.blocks?.length evaluates to 0 (falsy) when the native engine
returns cfg: { blocks: [] } for trivial functions, spuriously triggering
needsWasmTrees. Use Array.isArray(d.cfg?.blocks) instead, and preserve
d.cfg === null for nodes that intentionally have no CFG (e.g. interface
members).
Impact: 1 functions changed, 1 affected
Impact: 1 functions changed, 0 affected
Generate inter-module call sequence diagrams from the graph DB. Participants are files (not functions), keeping diagrams readable. BFS forward from entry point, with optional --dataflow flag to annotate parameter names and return arrows. New: src/sequence.js, tests/integration/sequence.test.js Modified: cli.js, mcp.js, index.js, CLAUDE.md, mcp.test.js Impact: 8 functions changed, 6 affected
Impact: 1 functions changed, 1 affected
carlos-alm
force-pushed
the
feat/sequence-command
branch
from
133453f to
0201d31
Compare
|
Claude finished @carlos-alm's task —— View job PR Review: Sequence Command ImplementationConducting comprehensive review following 4-phase analysis process... Todo List
Phase 1: Root Cause Analysis ✅Problem Being Solved: Adding visualization capability to codegraph via Mermaid sequence diagrams. This addresses a legitimate need for understanding call flows in codebases, particularly for architectural analysis and debugging. Why This Approach: The implementation follows established patterns in the codebase:
Risk Assessment: LOW - This is a pure additive feature with no changes to core parsing/building logic. Phase 2: Configuration Change Detection ✅No High-Risk Configuration Changes Detected:
Phase 3: Test Integrity ✅Test Quality Assessment:
Performance Improvement Note: Recent commit Phase 4: Code Quality ✅Security Assessment:
Implementation Quality:
Minor Areas for Future Enhancement (non-blocking):
Critical Concerns (if any)None identified - This is a well-implemented additive feature following established patterns. Final Assessment |
tests/integration/sequence.test.js
Outdated
| // Should have 4 files as participants (builder, parser, resolve, src/utils/helper) | ||
| // (test file excluded by noTests, lib/utils/helper reachable via helperA) | ||
| expect(data.participants.length).toBe(5); |
There was a problem hiding this comment.
Misleading participant count in comment
The inline comment says "4 files" but the assertion is toBe(5). Counting the listed files — builder.js, parser.js, resolve.js, src/utils/helper.js, and lib/utils/helper.js — there are 5. The comment should be updated to match the assertion and the actual fixture data.
| // Should have 4 files as participants (builder, parser, resolve, src/utils/helper) | |
| // (test file excluded by noTests, lib/utils/helper reachable via helperA) | |
| expect(data.participants.length).toBe(5); | |
| // Should have 5 files as participants (builder, parser, resolve, src/utils/helper, lib/utils/helper) | |
| // (test file excluded by noTests) | |
| expect(data.participants.length).toBe(5); |
There was a problem hiding this comment.
Fixed in 19a990a — updated the test comment to correctly say 5 files, matching the assertion.
| frontier = nextFrontier; | ||
| if (frontier.length === 0) break; | ||
|
|
||
| if (d === maxDepth && frontier.length > 0) { | ||
| truncated = true; | ||
| } |
There was a problem hiding this comment.
Possible false-positive truncated flag
truncated is set to true whenever nextFrontier is non-empty at d === maxDepth, but the nodes in nextFrontier at that point are the depth-maxDepth callees — they haven't yet been queried as callers. If all of them are leaf nodes (no outgoing calls edges), the diagram is actually complete, yet the "Truncated at depth N" note will still appear in the Mermaid output.
To avoid spurious notes, the flag should only be set after confirming at least one of those frontier nodes actually has outgoing calls:
frontier = nextFrontier;
if (frontier.length === 0) break;
if (d === maxDepth && frontier.length > 0) {
// Only mark truncated if at least one frontier node has further callees
const hasMoreCalls = frontier.some((fid) => getCallees.all(fid).length > 0);
if (hasMoreCalls) truncated = true;
}This re-uses the already-prepared getCallees statement, so there's no extra statement compilation cost.
There was a problem hiding this comment.
Fixed in 19a990a — changed the truncated check to only set true when nextFrontier contains nodes not already in visited, avoiding false-positives when the frontier at maxDepth has already been fully explored.
| function escapeMermaid(str) { | ||
| return str.replace(/:/g, '#colon;').replace(/"/g, '#quot;'); | ||
| } |
There was a problem hiding this comment.
escapeMermaid doesn't handle < / > characters
Mermaid sequence diagram labels that contain angle brackets — common in TypeScript function signatures stored in the graph (e.g. createMap<K, V>, Promise<Result>) — can break the renderer. Only : and " are currently escaped.
| function escapeMermaid(str) { | |
| return str.replace(/:/g, '#colon;').replace(/"/g, '#quot;'); | |
| } | |
| function escapeMermaid(str) { | |
| return str | |
| .replace(/</g, '<') | |
| .replace(/>/g, '>') | |
| .replace(/:/g, '#colon;') | |
| .replace(/"/g, '#quot;'); | |
| } |
There was a problem hiding this comment.
Fixed in 19a990a — added angle bracket escaping (\ → , \ → ) to escapeMermaid, so TypeScript generic signatures like \ render correctly in Mermaid labels.
src/sequence.js
Outdated
| // Build participant list from files | ||
| const aliases = buildAliases([...fileSet]); | ||
| const participants = [...fileSet].map((file) => ({ | ||
| id: aliases.get(file), | ||
| label: file.split('/').pop(), | ||
| file, | ||
| })); | ||
|
|
||
| // Sort participants: entry file first, then alphabetically | ||
| participants.sort((a, b) => { | ||
| if (a.file === entry.file) return -1; | ||
| if (b.file === entry.file) return 1; | ||
| return a.file.localeCompare(b.file); | ||
| }); | ||
|
|
||
| // Replace file paths with alias IDs in messages | ||
| for (const msg of messages) { | ||
| msg.from = aliases.get(msg.from); | ||
| msg.to = aliases.get(msg.to); | ||
| } | ||
|
|
||
| db.close(); | ||
|
|
||
| const base = { | ||
| entry, | ||
| participants, | ||
| messages, | ||
| depth: maxDepth, | ||
| totalMessages: messages.length, | ||
| truncated, | ||
| }; | ||
| return paginateResult(base, 'messages', { limit: opts.limit, offset: opts.offset }); |
There was a problem hiding this comment.
Orphaned participants when pagination is used
participants is derived from fileSet, which is built across the full BFS traversal. paginateResult then slices only the messages array. When limit/offset are applied, the paginated messages may only reference a subset of files, but all participant declarations are still emitted in the Mermaid output. This produces orphaned participant boxes in the diagram that have no arrows.
Consider filtering participants to only those files that actually appear in the (already-paginated) messages before returning:
const result = paginateResult(base, 'messages', { limit: opts.limit, offset: opts.offset });
if (opts.limit !== undefined || opts.offset !== undefined) {
const activeFiles = new Set(result.messages.flatMap((m) => [m.from, m.to]));
result.participants = result.participants.filter((p) => activeFiles.has(p.id));
}
return result;There was a problem hiding this comment.
Fixed in 19a990a — participants are now filtered after pagination to only include files that appear in the sliced messages array, preventing orphaned participants when pagination is active.
|
Claude finished @carlos-alm's task —— View job PR Review: Sequence Command ImplementationConducting comprehensive review following 4-phase analysis process... Todo List
Phase 1: Root Cause Analysis ✅Problem Being Solved: Adding visualization capability to codegraph via Mermaid sequence diagrams. This addresses a legitimate need for understanding call flows in codebases, particularly for architectural analysis and debugging. Why This Approach: The implementation follows established patterns in the codebase:
Risk Assessment: LOW - This is a pure additive feature with no changes to core parsing/building logic. Phase 2: Configuration Change Detection ✅No High-Risk Configuration Changes Detected:
Phase 3: Test Integrity ✅Test Quality Assessment:
Performance Improvement Note: Recent commit Phase 4: Code Quality ✅Security Assessment:
Implementation Quality:
Minor Areas for Future Enhancement (non-blocking):
Critical Concerns (if any)None identified - This is a well-implemented additive feature following established patterns. Final Assessment |
src/sequence.js
Outdated
| const alias = parts | ||
| .slice(-depth) | ||
| .join('/') | ||
| .replace(/[^a-zA-Z0-9_/-]/g, '_'); |
There was a problem hiding this comment.
Collision aliases contain / — invalid Mermaid participant identifiers
When two files share a basename (e.g. src/utils/helper.js and lib/utils/helper.js), the collision-resolution loop at depth 3 produces src/utils/helper and lib/utils/helper. These are then used literally as Mermaid participant identifiers:
participant src/utils/helper as helper.js
src/utils/helper->>lib/utils/helper: helperB
Mermaid sequence diagram identifiers do not support / — the renderer will fail to parse the diagram. The regex [^a-zA-Z0-9_/-] explicitly preserves /, which is the root cause.
Replace join('/') with join('_') and remove / from the allowed character set:
| const alias = parts | |
| .slice(-depth) | |
| .join('/') | |
| .replace(/[^a-zA-Z0-9_/-]/g, '_'); | |
| const alias = parts | |
| .slice(-depth) | |
| .join('_') | |
| .replace(/[^a-zA-Z0-9_-]/g, '_'); |
With this change, disambiguated aliases use underscores instead of slashes (e.g. src_utils_helper and lib_utils_helper), both valid Mermaid identifiers.
There was a problem hiding this comment.
Fixed in 19a990a — changed \ to \ and removed \ from the allowed regex character class. Participant IDs now only contain , which are valid Mermaid identifiers.
| test('participant alias collision — two helper.js files get distinct IDs', () => { | ||
| const data = sequenceData('buildGraph', dbPath, { noTests: true }); | ||
| const helperParticipants = data.participants.filter((p) => p.label === 'helper.js'); | ||
| expect(helperParticipants.length).toBe(2); | ||
|
|
||
| // IDs should be distinct | ||
| const ids = helperParticipants.map((p) => p.id); | ||
| expect(ids[0]).not.toBe(ids[1]); | ||
| }); |
There was a problem hiding this comment.
Alias-collision test only validates distinctness, not Mermaid format
The test confirms that two helper.js files receive distinct ids, but does not verify that those IDs are valid Mermaid participant identifiers (alphanumeric, hyphen, or underscore only — no slashes). Adding a complementary assertion would catch rendering bugs:
| test('participant alias collision — two helper.js files get distinct IDs', () => { | |
| const data = sequenceData('buildGraph', dbPath, { noTests: true }); | |
| const helperParticipants = data.participants.filter((p) => p.label === 'helper.js'); | |
| expect(helperParticipants.length).toBe(2); | |
| // IDs should be distinct | |
| const ids = helperParticipants.map((p) => p.id); | |
| expect(ids[0]).not.toBe(ids[1]); | |
| }); | |
| test('participant alias collision — two helper.js files get distinct IDs', () => { | |
| const data = sequenceData('buildGraph', dbPath, { noTests: true }); | |
| const helperParticipants = data.participants.filter((p) => p.label === 'helper.js'); | |
| expect(helperParticipants.length).toBe(2); | |
| // IDs should be distinct | |
| const ids = helperParticipants.map((p) => p.id); | |
| expect(ids[0]).not.toBe(ids[1]); | |
| // Validate participant IDs are valid Mermaid identifiers (no slashes) | |
| for (const p of data.participants) { | |
| expect(p.id).toMatch(/^[a-zA-Z0-9_-]+$/); | |
| } | |
| // Validate the IDs appear correctly in Mermaid output | |
| const mermaid = sequenceToMermaid(data); | |
| for (const p of data.participants) { | |
| expect(mermaid).toContain(`participant ${p.id} as`); | |
| } | |
| }); |
There was a problem hiding this comment.
Fixed in 19a990a — added a \ assertion on all participant IDs in the alias collision test, ensuring they conform to valid Mermaid format.
…escape, pagination
- Fix buildAliases join('/') producing Mermaid-invalid participant IDs;
use join('_') and strip '/' from allowed regex chars
- Add Mermaid format validation to alias collision test
- Fix truncated false-positive when nextFrontier has unvisited nodes at
maxDepth boundary
- Add angle bracket escaping to escapeMermaid (#gt;/#lt; entities)
- Filter orphaned participants after pagination slices messages
- Fix misleading test comment (4 files → 5 files)
Impact: 1 functions changed, 2 affected
carlos-alm
force-pushed
the
feat/sequence-command
branch
from
2f7af10 to
19a990a
Compare
carlos-alm
force-pushed
the
feat/sequence-command
branch
from
0197a65 to
d7319cf
Compare
carlos-alm
force-pushed
the
feat/sequence-command
branch
from
d7319cf to
0db9017
Compare
1 participant
Summary
codegraph sequence <name>command that generates MermaidsequenceDiagramoutput from the call graph--dataflowflag to annotate parameter names and return arrowsExample:
codegraph sequence flowData -T --depth 2sequenceDiagram participant flow as flow.js participant db as db.js participant paginate as paginate.js participant queries as queries.js flow->>db: openReadonlyOrFail flow->>flow: findBestMatch flow->>flow: entryPointType flow->>queries: isTestFile flow->>paginate: paginateResult db->>db: findDbPath flow->>queries: isTestFile paginate->>paginate: paginate note right of flow: Truncated at depth 2Files changed
src/sequence.jssequenceData(),sequenceToMermaid(),sequence()tests/integration/sequence.test.jssrc/cli.jssequence <name>commandsrc/mcp.jssequenceMCP tool definition + handlersrc/index.jstests/unit/mcp.test.jssequenceto expected tool namesCLAUDE.mdDeferred (follow-up PR)
sequencetoBATCH_COMMANDSinsrc/batch.jssequence: 50toMCP_DEFAULTSinsrc/paginate.jsTest plan
codegraph sequence buildGraph -Tproduces valid Mermaid (15 participants, 218 messages)codegraph sequence buildGraph -T -joutputs valid JSONcodegraph sequence flowData -T --depth 2produces compact diagram