Coupling and cohesion in a design

Coupling and cohesion are two measures of how well a design is split into parts. Low coupling and high cohesion make a system easier to change.

Requirements always change. A good design lets you change one part without breaking the rest. This lesson shows how to count the links that decide that.

Split a system into modules, parts with one job each. Coupling is how much one module depends on another.

High coupling means a change in one module forces changes in others. Low coupling means a module only knows a small interface: the few things it lets others use.

An arrow means "uses". Fan-out is how many arrows leave a module. Fan-in is how many arrive.

Here orders has fan-out 2 (users and db) and fan-in 1 (web). The db module has fan-out 0 and fan-in 2. A module with a high fan-in is risky to change, because many parts rely on it.

A design can be kept as a list of lines, one link per line, "from to". The first Run writes the five links to a file and prints nothing.

The next two count lines. Lines starting with orders are its fan-out, so that prints 2. Lines ending with users are its fan-in, so that prints 2.

printf 'web orders\nweb users\norders users\norders db\nusers db\n' > uses.txt

grep -c '^orders' uses.txt

grep -c 'users$' uses.txt

Cohesion is how well the inside of one module fits one job. A module that books rooms and also sends invoices and also resizes photos has low cohesion.

High cohesion is easy to name: "bookings". If you need the word "and" to describe a module, split it.

The lab's API is a narrow interface. A client asks for one column of one row and knows nothing about how the data is stored. The Run asks for the room of the first booking.

It prints one row with only Lab 2 in it. The database could change its other columns and this client would not notice.

export PGPASSWORD=shop

psql -h db -U shop -c "drop table if exists bookings; create table bookings (id serial primary key, room text not null, slot text not null, booked_by text not null, hours int not null, status text not null default 'booked'); insert into bookings (room, slot, booked_by, hours) values ('Lab 2', 'Mon 10:00', 'Asha', 2), ('Lab 4', 'Mon 10:00', 'Ben', 1), ('Lab 2', 'Tue 14:00', 'Asha', 3); notify pgrst, 'reload schema'"

curl -s -w "\n" "http://api:3000/bookings?select=room&order=id&limit=1"