Code quality · Updated 2026-09-17
Code duplication
The same or near-identical logic existing in more than one place, so a change has to be made more than once to be made correctly.
What it is
Detectors report it as a percentage of duplicated lines or blocks, usually with a configurable similarity threshold. Exact copies are easy to find; near-duplicates that drifted apart are harder and more dangerous, because they look intentional.
Why the cost is in the update, not the storage
Duplication is not wasteful because of disk space. It is expensive because the next person to fix a bug will fix one copy, tests will pass, and the other copy will keep the bug — often for years. The failure is silent and shows up far from the change.
Why the metric needs judgement
Not all duplication should be removed. Two services that happen to validate an email address the same way are not obviously better off sharing a library; coupling them creates a new failure mode to buy a saved function. The classic guidance — prefer duplication to the wrong abstraction — exists because premature deduplication produces code that is harder to change than the copies were.
A duplication percentage cannot tell the difference. A reviewer can, and this is one of the places an AI reviewer with real repository context can genuinely help: it can point at the existing abstraction the author did not know about.
Common mistakes
- Setting an organisation-wide duplication threshold.
- Deduplicating test code aggressively, where explicitness usually beats reuse.
- Ignoring near-duplicates because the detector’s threshold was set too strict.