Merget vs graphify: more complete call graphs

Merget decides which pull requests can merge together by reading a call graph: which function calls which. A graph that misses calls hides risk, and one that invents calls holds work back for nothing. We scored Merget’s graph and the graph of graphify, an open-source code-graph tool, against the same compiler-grade reference on 12 open-source repositories in 9 languages.

Merget found more reference calls in 11 of 12 repositories and achieved higher verified precision in 10 of 12. That gives the merge queue a more complete map of dependencies across these codebases.

MeasuredOctober 13, 2026
ReferenceSCIP index per repository
Scope12 repositories, 9 languages

A more complete dependency graph across 9 languages. Merget leads recall in 11 of 12 tested repositories and verified precision in 10.

11/12repositories where Merget’s graph holds more of the real calls (higher recall)
10/12repositories where more of Merget’s checked edges are right (higher precision)
9languages represented across the 12 tested repositories
Recall

Of the calls the reference lists, the share a graph also contains. Low recall means missed calls.

Precision

Of a graph’s edges that the reference could check, the share it did not contradict. Low precision means invented or misdirected calls.

Reference

A compiler-grade index of each repository (SCIP) that records where every call goes, built by a third-party indexer that uses the language’s own compiler or type checker. Both graphs were scored against the same one.

Merget finds more reference calls in 11 of 12 repositories. Inspect both coverage and verified precision for your stack.

How to read thisEach pair of bars is one tool’s score from 0% to 100%; longer is better. The diamond marks the higher score.

MergetgraphifyBetter of the two
RepositoryReference factsRecallshare of real calls foundPrecisionshare of checked edges right
TypeScriptTypeScript · 45,011 reference facts45,011Merget: 94.1% (better)graphify: 66.0%Merget: 99.8% (better)graphify: 95.7%
prometheusGo · 26,610 reference facts26,610Merget: 95.4% (better)graphify: 44.0%Merget: 99.4% (better)graphify: 88.0%
erpnextPython · 23,408 reference facts23,408Merget: 81.9% (better)graphify: 78.2%Merget: 99.0% (better)graphify: 95.0%
redisC · 20,497 reference facts20,497Merget: 80.3%graphify: 86.5% (better)Merget: 99.2%graphify: 99.7% (better)
djangoPython · 9,998 reference facts9,998Merget: 87.7% (better)graphify: 59.9%Merget: 98.6% (better)graphify: 79.4%
ripgrepRust · 5,449 reference facts5,449Merget: 89.1% (better)graphify: 35.4%Merget: 97.8% (better)graphify: 62.1%
commons-langJava · 4,699 reference facts4,699Merget: 90.1% (better)graphify: 42.8%Merget: 99.4% (better)graphify: 78.7%
eslintJavaScript · 4,561 reference facts4,561Merget: 86.1% (better)graphify: 52.4%Merget: 98.4% (better)graphify: 95.6%
AutoMapperC# · 1,204 reference facts1,204Merget: 83.6% (better)graphify: 39.5%Merget: 98.7% (better)graphify: 78.1%
catsScala · 345 reference facts345Merget: 71.6% (better)graphify: 4.6%Merget: 97.6% (better)graphify: 11.9%
flaskPython · 253 reference facts253Merget: 94.9% (better)graphify: 64.0%Merget: 96.8% (better)graphify: 92.1%
expressJavaScript · 56 reference facts56Merget: 100% (better)graphify: 55.4%Merget: 91.8%graphify: 93.9% (better)
Full results: edge counts per repository

Merget on every repository, with the counts behind each rate.

RepositoryLanguageReference factsRecallPrecisionJudgedWrongEdges
TypeScriptTypeScript45,01194.1%99.8%42,6229043,325
prometheusGo26,61095.4%99.4%25,52914325,529
erpnextPython23,40881.9%99.0%19,37418922,395
redisC20,49780.3%99.2%17,92315218,983
djangoPython9,99887.7%98.6%8,93112610,436
ripgrepRust5,44989.1%97.8%4,9651115,000
commons-langJava4,69990.1%99.4%4,258264,258
eslintJavaScript4,56186.1%98.4%4,363706,364
AutoMapperC#1,20483.6%98.7%1,640212,683
catsScala34571.6%97.6%2536268
flaskPython25394.9%96.8%2528264
expressJavaScript56100%91.8%615597

graphify on the same repositories and reference facts.

RepositoryRecallPrecisionJudgedWrongEdges scoredLinks not scored
TypeScript66.0%95.7%31,1271,35331,303384
prometheus44.0%88.0%13,3181,60413,3181,103
erpnext78.2%95.0%19,27097119,4921,705
redis86.5%99.7%17,9695718,912436
django59.9%79.4%7,5501,5547,7531,260
ripgrep35.4%62.1%3,1111,1793,142171
commons-lang42.8%78.7%2,5575452,558340
eslint52.4%95.6%2,5301122,664217
AutoMapper39.5%78.1%6351391,4801,677
cats4.6%11.9%1351191501,253
flask64.0%92.1%177141915
express55.4%93.9%332364

When either tool says an edge is certain, Merget’s are contradicted less often in 11 of 12 repositories.

Certain means each tool’s most-confident label: precise for Merget; EXTRACTED for graphify, its label for a link found directly in source.

12/12repositories where Merget’s certain edges, all of them, find more of the real calls
11/12fewer certain edges contradicted, among those the reference could check
Per repository: certain edges the reference contradicted

Merget’s certain edges are contradicted less often than graphify’s in 11 of 12 repositories.

How to read thisEach bar is the share of a tool’s certain edges that the reference contradicted, out of those it could check; shorter is better. The track ends at 40%; graphify’s cats bar (88.1%) runs past it.

MergetgraphifyBetter of the two
RepositoryCertain edges contradictedout of those checked; fewer is better
TypeScriptTypeScriptMerget: 0.03% · 11 of 39,920 (better)graphify: 4.5% · 1,345 of 30,205
prometheusGoMerget: 0.4% · 100 of 23,332 (better)graphify: 16.2% · 1,522 of 9,376
erpnextPythonMerget: 0.4% · 77 of 18,453 (better)graphify: 2.8% · 366 of 13,287
redisCMerget: 0.8% · 141 of 17,153graphify: 0.4% · 27 of 6,360 (better)
djangoPythonMerget: 1.1% · 98 of 8,873 (better)graphify: 24.7% · 1,498 of 6,060
ripgrepRustMerget: 0.5% · 24 of 4,628 (better)graphify: 39.5% · 1,176 of 2,980
commons-langJavaMerget: 0.2% · 9 of 3,844 (better)graphify: 21.8% · 516 of 2,369
eslintJavaScriptMerget: 0.7% · 27 of 4,105 (better)graphify: 4.4% · 112 of 2,522
AutoMapperC#Merget: 0.8% · 11 of 1,447 (better)graphify: 30.4% · 95 of 312
catsScalaMerget: 0% · 0 of 222 (better)graphify: 88.1% · 118 of 134
flaskPythonMerget: 2.8% · 7 of 251 (better)graphify: 7.4% · 13 of 176
expressJavaScriptMerget: 0% · 0 of 56 (better)graphify: 6.1% · 2 of 33
Full results: certain edges

Recall and precision counted over each tool’s certain edges only, with how many of the checked certain edges were contradicted.

Merget, certain (precise) edgesgraphify, certain (EXTRACTED) edges
RepositoryRecallPrecisionEdgesContradictedRecallPrecisionEdgesContradicted
TypeScript88.4%100%40,55311 of 39,92063.9%95.5%30,3651,345 of 30,205
prometheus87.3%99.6%23,332100 of 23,33229.5%83.8%9,3761,522 of 9,376
erpnext78.5%99.6%21,22477 of 18,45355.2%97.2%13,347366 of 13,287
redis80.2%99.2%17,465141 of 17,15330.5%99.6%6,69027 of 6,360
django87.5%98.9%10,12098 of 8,87345.6%75.3%6,2201,498 of 6,060
ripgrep84.5%99.5%4,63724 of 4,62833.1%60.5%3,0111,176 of 2,980
commons-lang81.6%99.8%3,8449 of 3,84439.4%78.2%2,370516 of 2,369
eslint81.5%99.3%4,38427 of 4,10552.2%95.6%2,656112 of 2,522
AutoMapper68.1%99.2%2,26911 of 1,44716.6%69.6%1,05695 of 312
cats64.3%100%2300 of 2224.6%11.9%138118 of 134
flask94.9%97.2%2607 of 25164.0%92.6%19013 of 176
express100%100%820 of 5655.4%93.9%362 of 33

Merget reaches 76% of later fix targets, versus 49% for graphify. Two calls out from changed code, on 30 scorable Cargo merges.

A graph helps only if walking it leads somewhere. We took 38 merges in cargo, 30 of which a later human fix can score, started from the code both sides changed, and followed each tool’s call edges outward. No model is in the loop.

Merget’s graph reaches more of the fix at every depth.

How to read thisEach bar is the share of the later fixes’ targets a graph reaches 1, 2 and 3 calls out, pooled over the 30 merges.

Mergetgraphify
1 call out
Merget: 52%
graphify: 15%
2 calls out
Merget: 76%
graphify: 49%
3 calls out
Merget: 90%
graphify: 61%
Full results: reaching the fix

All 30 scorable merges: the share of fix targets each graph reaches, and the merges where it reaches at least one.

DepthShare reached, MergetShare reached, graphifyReached any, MergetReached any, graphify
1 call out52%15%21 of 3017 of 30
2 calls out76%49%28 of 3025 of 30
3 calls out90%61%30 of 3028 of 30

More of the dependency graph helps a merge queue see where PRs collide. These results support Merget’s analysis foundation across the tested stacks.

Merget’s full graph had higher recall on every tested repository in TypeScript, Go, Python, Rust, Java, JavaScript, C# and Scala.

The merge queue uses this graph to see which open pull requests touch the same definitions and callers, and to give conflict resolution the context it needs.

Next step

See it on your code. Start in Advisory mode and read Merget’s findings on your own pull requests.

Sign up free
How we ran it

Measured October 13, 2026 on 12 open-source repositories at pinned commits. Both graphs were scored against the same SCIP reference index for each repository; graphify ran in its code-only mode.