Cohesion and coupling judged, RFC 0002 step 3 (0.20.0) #38

2026-10-04T12:14:19+02:00

mi commented

2026-10-04 12:14:19 +02:00

The Problem

Since PR#36 cohesion and coupling are measured and since PR#37 shown, but nothing kept a class from falling further apart or depending on ever more classes.
The first measurement of hsadmin-ng also showed that LCOM4 as published counts every method without state as a group of its own: of 66 classes with LCOM4 of 2 or more, most owed it to a toString through a static helper, a bean factory or the methods of an anonymous class.

The Scenarios

@MC-CK-010 a group of methods that uses no instance field at all does not count; a class with only such methods has no LCOM4.
@MC-CK-011 the gate judges LCOM4 and outgoing coupling against maxLcom4 and maxEfferentCoupling; a violation names class and dimension; a class exactly at a limit passes.
@MC-CK-012 a dimension without a limit is reported, not judged; without any limit the gate is inactive, as before.

Verified by

DesignAnalysisTest methodsWithoutStateDoNotCount
GateTest cohesionAndCouplingHaveLimitsOfTheirOwn, aDimensionWithoutALimitIsOnlyReported, classLimitsNeedTheSectionClasses
GateFunctionalTest cohesionAndCouplingLimitsReachTheGate

The Solution

Cohesion keeps only the groups that use an instance field.
GateConfig has the optional limits maxLcom4 and maxEfferentCoupling; the gate judges the methods only with maxMutationCrap and the classes only with one of the new limits.
The failure line counts "limit(s) exceeded", as it now covers classes too.
The Gradle DSL has maxLcom4 and maxEfferentCoupling; a limit below its minimum fails the check with a message.
werkcrappit-pit applies the Gradle plugin 0.19.0, so its own city shows the metrics of step 2.
Version 0.20.0.

Follow-up PRs

Ratchets for werkcrappit itself, once 0.20.0 is released: maxLcom4 = 3, maxEfferentCoupling = 21.
hsadmin-ng, in its pilot branch: maxLcom4 = 4, maxEfferentCoupling = 26.

Attachments

LCOM4 and Ce, measured from the class files on 2026-10-04

	LCOM4 max, as published → without stateless groups	Classes with LCOM4 of 2 or more	Ce max
hsadmin-ng (267 classes, generated code left out)	12 → 4	66 → 19	26 (`HsAccountController`)
werkcrappit-pit (58 classes)	3 → 3	9 → 4	21 (`MutationCrapListener`)

The top of hsadmin-ng now shows real candidates: HsOfficeContact (4) with its independent e-mail, phone and postal address, HsOfficeRelationSearchCriteria (4) with four independent search fields.

🤖 Generated with Claude Code

## Related Links - [RFC 0002](../rfcs/0002-cohesion-and-coupling.md), step 3; [PR#36](2026-10-04-PR%2336-cohesion-coupling-measured.md), [PR#37](2026-10-04-PR%2337-cohesion-coupling-city.md) ## The Problem Since PR#36 cohesion and coupling are measured and since PR#37 shown, but nothing kept a class from falling further apart or depending on ever more classes. The first measurement of hsadmin-ng also showed that LCOM4 as published counts every method without state as a group of its own: of 66 classes with LCOM4 of 2 or more, most owed it to a `toString` through a static helper, a bean factory or the methods of an anonymous class. ## The Scenarios - `@MC-CK-010` a group of methods that uses no instance field at all does not count; a class with only such methods has no LCOM4. - `@MC-CK-011` the gate judges LCOM4 and outgoing coupling against `maxLcom4` and `maxEfferentCoupling`; a violation names class and dimension; a class exactly at a limit passes. - `@MC-CK-012` a dimension without a limit is reported, not judged; without any limit the gate is inactive, as before. ##### Verified by - [DesignAnalysisTest](../../werkcrappit-pit/src/test/java/de/javagil/werkcrappit/pit/design/DesignAnalysisTest.java) `methodsWithoutStateDoNotCount` - [GateTest](../../werkcrappit-pit/src/test/java/de/javagil/werkcrappit/pit/gate/GateTest.java) `cohesionAndCouplingHaveLimitsOfTheirOwn`, `aDimensionWithoutALimitIsOnlyReported`, `classLimitsNeedTheSectionClasses` - [GateFunctionalTest](../../werkcrappit-gradle-plugin/src/functionalTest/java/de/javagil/werkcrappit/gradle/GateFunctionalTest.java) `cohesionAndCouplingLimitsReachTheGate` ## The Solution - `Cohesion` keeps only the groups that use an instance field. - `GateConfig` has the optional limits `maxLcom4` and `maxEfferentCoupling`; the gate judges the methods only with `maxMutationCrap` and the classes only with one of the new limits. The failure line counts "limit(s) exceeded", as it now covers classes too. - The Gradle DSL has `maxLcom4` and `maxEfferentCoupling`; a limit below its minimum fails the check with a message. - werkcrappit-pit applies the Gradle plugin 0.19.0, so its own city shows the metrics of step 2. - Version 0.20.0. ## Follow-up PRs - Ratchets for werkcrappit itself, once 0.20.0 is released: `maxLcom4 = 3`, `maxEfferentCoupling = 21`. - hsadmin-ng, in its pilot branch: `maxLcom4 = 4`, `maxEfferentCoupling = 26`. ## Attachments ### LCOM4 and Ce, measured from the class files on 2026-10-04 | | LCOM4 max, as published → without stateless groups | Classes with LCOM4 of 2 or more | Ce max | |---|---|---|---| | hsadmin-ng (267 classes, generated code left out) | 12 → 4 | 66 → 19 | 26 (`HsAccountController`) | | werkcrappit-pit (58 classes) | 3 → 3 | 9 → 4 | 21 (`MutationCrapListener`) | The top of hsadmin-ng now shows real candidates: `HsOfficeContact` (4) with its independent e-mail, phone and postal address, `HsOfficeRelationSearchCriteria` (4) with four independent search fields. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

mi added 1 commit 2026-10-04 12:14:20 +02:00

Cohesion and coupling judged, RFC 0002 step 3 (0.20.0) (PR#38)

werkator build succeeded after 01:43

Details

6327eb43ca

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

mi merged commit 8a4568e34d into main

2026-10-04 12:25:18 +02:00

mi deleted branch mihoe/cohesion-coupling-gate

2026-10-04 12:25:18 +02:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: mi/werkcrappit#38