Coupling (computer programming)

In software engineering, coupling is the degree of interdependence between software modules, a measure of how closely connected two routines or modules are,^[1] and the strength of the relationships between modules.^[2] Coupling is not binary but multi-dimensional. ^[3]

Coupling is usually contrasted with cohesion. Low coupling often correlates with high cohesion, and vice versa. Low coupling is often thought to be a sign of a well-structured computer system and a good design, and when combined with high cohesion, supports the general goals of high readability and maintainability.^{[citation needed]}

History

The software quality metrics of coupling and cohesion were invented by Larry Constantine in the late 1960s as part of a structured design, based on characteristics of “good” programming practices that reduced maintenance and modification costs. Structured design, including cohesion and coupling, were published in the article Stevens, Myers & Constantine (1974)^[4] and the book Yourdon & Constantine (1979),^[5] and the latter subsequently became standard terms.

Types of coupling

Coupling can be "low" (also "loose" and "weak") or "high" (also "tight" and "strong"). Some types of coupling, in order of highest to lowest coupling, are as follows:

Procedural programming

A module here refers to a subroutine of any kind, i.e. a set of one or more statements having a name and preferably its own set of variable names.

Content coupling (high): Content coupling is said to occur when one module uses the code of another module, for instance a branch. This violates information hiding – a basic software design concept.
Common coupling: Common coupling is said to occur when several modules have access to the same global data. But it can lead to uncontrolled error propagation and unforeseen side-effects when changes are made.
External coupling: External coupling occurs when two modules share an externally imposed data format, communication protocol, or device interface. This is basically related to the communication to external tools and devices.
Control coupling: Control coupling is one module controlling the flow of another, by passing it information on what to do (e.g., passing a what-to-do flag).
Stamp coupling (data-structured coupling): Stamp coupling occurs when modules share a composite data structure and use only parts of it, possibly different parts (e.g., passing a whole record to a function that needs only one field of it).; In this situation, a modification in a field that a module does not need may lead to changing the way the module reads the record. To illustrate the concept of stamp coupling, consider a scenario involving a UserProfile component. This component is designed to return the entire user profile information in response to requests, even when consumers only require a specific attribute. This practice exemplifies stamp coupling, which can lead to significant bandwidth issues, especially at scale. When any attribute within the UserProfile component changes, all consumers that interact with it may need to undergo testing, even if they do not utilize the modified attribute. ^[6]
Data coupling: Data coupling occurs when modules share data through, for example, parameters. Each datum is an elementary piece, and these are the only data shared (e.g., passing an integer to a function that computes a square root).

Object-oriented programming

Subclass coupling: Describes the relationship between a child and its parent. The child is connected to its parent, but the parent is not connected to the child.

Temporal coupling: It is when two actions are bundled together into one module just because they happen to occur at the same time.

In recent work various other coupling concepts have been investigated and used as indicators for different modularization principles used in practice.^[7]

Dynamic coupling

The goal of defining and measuring this type of coupling is to provide a run-time evaluation of a software system. It has been argued that static coupling metrics lose precision when dealing with an intensive use of dynamic binding or inheritance.^[8] In the attempt to solve this issue, dynamic coupling measures have been taken into account.

Semantic coupling

This kind of a coupling metric considers the conceptual similarities between software entities using, for example, comments and identifiers and relying on techniques such as latent semantic indexing (LSI).

Logical coupling

Logical coupling (or evolutionary coupling or change coupling) analysis exploits the release history of a software system to find change patterns among modules or classes: e.g., entities that are likely to be changed together or sequences of changes (a change in a class A is always followed by a change in a class B).

Dimensions of coupling

According to Gregor Hohpe, coupling is multi-dimensional:^[3]

Technology Dependency
Location Dependency
Topology Dependency
Data Format & Type Dependency
Semantic Dependency
Conversation Dependency
Order Dependency
Temporal Dependency

Disadvantages of tight coupling

Tightly coupled systems tend to exhibit the following developmental characteristics, which are often seen as disadvantages:

A change in one module usually forces a ripple effect of changes in other modules.
Assembly of modules might require more effort and/or time due to the increased inter-module dependency.
A particular module might be harder to reuse and/or test because dependent modules must be included.

Performance issues

Whether loosely or tightly coupled, a system's performance is often reduced by message and parameter creation, transmission, translation (e.g. marshaling) and message interpretation (which might be a reference to a string, array or data structure), which require less overhead than creating a complicated message such as a SOAP message. Longer messages require more CPU and memory to produce. To optimize runtime performance, message length must be minimized and message meaning must be maximized.

Message Transmission Overhead and Performance: Since a message must be transmitted in full to retain its complete meaning, message transmission must be optimized. Longer messages require more CPU and memory to transmit and receive. Also, when necessary, receivers must reassemble a message into its original state to completely receive it. Hence, to optimize runtime performance, message length must be minimized and message meaning must be maximized.

Message Translation Overhead and Performance: Message protocols and messages themselves often contain extra information (i.e., packet, structure, definition and language information). Hence, the receiver often needs to translate a message into a more refined form by removing extra characters and structure information and/or by converting values from one type to another. Any sort of translation increases CPU and/or memory overhead. To optimize runtime performance, message form and content must be reduced and refined to maximize its meaning and reduce translation.

Message Interpretation Overhead and Performance: All messages must be interpreted by the receiver. Simple messages such as integers might not require additional processing to be interpreted. However, complex messages such as SOAP messages require a parser and a string transformer for them to exhibit intended meanings. To optimize runtime performance, messages must be refined and reduced to minimize interpretation overhead.

Solutions

One approach to decreasing coupling is functional design, which seeks to limit the responsibilities of modules along functionality. Coupling increases between two classes A and B if:

A has an attribute that refers to (is of type) B.
A calls on services of an object B.
A has a method that references B (via return type or parameter).
A is a subclass of (or implements) class B.

Low coupling refers to a relationship in which one module interacts with another module through a simple and stable interface and does not need to be concerned with the other module's internal implementation (see Information Hiding).

Systems such as CORBA or COM allow objects to communicate with each other without having to know anything about the other object's implementation. Both of these systems even allow for objects to communicate with objects written in other languages.

Coupling vs Connascence

Coupling describes the degree and nature of dependency between software components, focusing on both what they share (e.g., data, control flow, technology) and how tightly they're bound. It evaluates two key dimensions: strength (how difficult it is to change the dependency) and directness (whether dependencies are direct or indirect). Traditional coupling types, ranked from strongest to weakest, include content coupling, common coupling, external coupling, control coupling, stamp coupling, and data coupling. ^[9]^[10]

Connascence, introduced by Meilir Page-Jones, provides a formal and systematic framework for analyzing and classifying coupling dependencies. Rather than being separate from coupling, Connascence serves as a granular metric for measuring coupling strength and resilience. It evaluates dependencies along two key dimensions: strength, which measures the effort needed to refactor or change the dependency, and locality, which considers how physically close the dependent components are in the codebase. ^[9]^[10]

Connascence can be further divided into static and dynamic forms. Static Connascence refers to dependencies detectable at compile-time, such as method signatures, while dynamic Connascence refers to dependencies visible only at runtime, such as dynamic behavior in polymorphic code. ^[9]^[10]

When two components depend on parameter order in a method signature, they exhibit Connascence of Position, a form of dependency that is fragile and difficult to refactor because reordering breaks the interface. In contrast, Connascence of Name relies on parameter or field names, making it more resilient to changes. Two services using a shared JSON API often exhibit data coupling, and beyond that, semantic coupling, which refers to the shared meaning and interpretation of the data being exchanged. Modern API designs often favor Connascence of Name over Connascence of Position to improve resilience against change. ^[9]^[10]

In essence, coupling maps the dependency landscape, identifying how components are connected. Connascence systematically measures and classifies these dependencies, providing deeper insights into their strength, locality, and resilience to change. Together, they offer complementary perspectives for designing maintainable, robust systems. ^[9]^[10]

Coupling versus cohesion

Coupling and cohesion are terms which occur together very frequently. Coupling refers to the interdependencies between modules, while cohesion describes how related the functions within a single module are. Low cohesion implies that a given module performs tasks which are not very related to each other and hence can create problems as the module becomes large.

Module coupling

Coupling in Software Engineering^[11] describes a version of metrics associated with this concept.

For data and control flow coupling:

d_i: number of input data parameters
c_i: number of input control parameters
d_o: number of output data parameters
c_o: number of output control parameters

For global coupling:

g_d: number of global variables used as data
g_c: number of global variables used as control

For environmental coupling:

w: number of modules called (fan-out)
r: number of modules calling the module under consideration (fan-in)

$\mathrm {Coupling} (C)=1-{\frac {1}{d_{i}+2\times c_{i}+d_{o}+2\times c_{o}+g_{d}+2\times g_{c}+w+r}}$

Coupling(C) makes the value larger the more coupled the module is. This number ranges from approximately 0.67 (low coupling) to 1.0 (highly coupled)

For example, if a module has only a single input and output data parameter

$C=1-{\frac {1}{1+0+1+0+0+0+1+0}}=1-{\frac {1}{3}}=0.67$

If a module has 5 input and output data parameters, an equal number of control parameters, and accesses 10 items of global data, with a fan-in of 3 and a fan-out of 4,

$C=1-{\frac {1}{5+2\times 5+5+2\times 5+10+0+3+4}}=0.98$

References

^ ISO/IEC/IEEE 24765:2010 Systems and software engineering — Vocabulary
^ ISO/IEC TR 19759:2005, Software Engineering — Guide to the Software Engineering Body of Knowledge (SWEBOK)
^ ^a ^b Hohpe, Gregor. Enterprise Integration Patterns: Designing, Building, and Deploying Messaging Solutions. Addison-Wesley Professional. ISBN 978-0321200686.
^ Stevens, Wayne P.; Myers, Glenford J.; Constantine, Larry LeRoy (June 1974). "Structured design". IBM Systems Journal. 13 (2): 115–139. doi:10.1147/sj.132.0115.
^ Yourdon, Edward; Constantine, Larry LeRoy (1979) [1975]. Structured Design: Fundamentals of a Discipline of Computer Program and Systems Design. Yourdon Press. Bibcode:1979sdfd.book.....Y. ISBN 978-0-13-854471-3.
^ Richards, Mark. Fundamentals of Software Architecture: An Engineering Approach. O'Reilly Media. ISBN 978-1492043454.
^ Beck, Fabian; Diehl, Stephan (September 2011). "On the Congruence of Modularity and Code Coupling". In Proceedings of the 19th ACM SIGSOFT Symposium and the 13th European Conference on Foundations of Software Engineering (SIGSOFT/FSE '11). Szeged, Hungary. p. 354. doi:10.1145/2025113.2025162. ISBN 9781450304436. S2CID 2413103.{{cite book}}: CS1 maint: location missing publisher (link)
^ Arisholm, Erik; Briand, Lionel C.; Føyen, Audun (August 2004). "Dynamic coupling measurement for object-oriented software". IEEE Transactions on Software Engineering. 30 (8). IEEE: 491–506. doi:10.1109/TSE.2004.41. hdl:10852/9090. S2CID 3074827.
^ ^a ^b ^c ^d ^e Practical Guide to Structured Systems Design. ISBN 978-0136907695.
^ ^a ^b ^c ^d ^e Designing Data-Intensive Applications: The Big Ideas Behind Reliable, Scalable, and Maintainable Systems. ISBN 978-1449373320.
^ Pressman, Roger S. (1982). Software Engineering - A Practitioner's Approach (4 ed.). McGraw-Hill. ISBN 0-07-052182-4.