Skip to content
Antoine Le LouëtIndependent senior developer (Home)
Back to the Journal

Advice3 min read

Take over a project without rewriting everything

Where do you start with existing code? A business flow helps choose a first fix and verify its effect.

Three connected steps: understand uses, protect flows and evolve the project.
A progression from context to a first verifiable improvement.

Start with how the application is used

Unfamiliar code often seems more complicated than code you would have written yourself. That difference is insufficient to justify a rewrite. Before choosing a new architecture, I suggest understanding the work the application enables and the constraints it addresses.

Take request approval. Walking through this flow with someone who knows the business can reveal input the interface forces users to repeat or an error blocking their work. These observations give context to the files and components you read next.

  • Identify a user and the task they are trying to complete.
  • Walk through the flow in an appropriate environment using test data.
  • Record dependencies: APIs, deferred processing, external services and access permissions.
  • Identify what is known, what remains to be checked and who can answer.

Protect a flow before changing it

A useful first protection is to describe observed behaviour. What should happen when a request is valid? What does the user see when a service does not respond? A targeted test, verification procedure or reference screenshot can help make this behaviour explicit. The choice depends on risk and the tools already present in the project.

You also need to know how to run the project, produce a release and return to the previous one. If these steps rely on oral knowledge, documenting them is already a contribution. Even a tiny fix is difficult to deliver when you cannot verify its result or identify the live version.

Choose a limited first step

I then favour an intervention whose result is easy to verify. If a network error erases the current request, preserving input and allowing another attempt provides an initial scope. This fix requires understanding the form and checking its failure behaviour, with a change small enough to remain readable.

TypeScript
const premiereEtape = {
  parcours: "Valider une demande",
  probleme: "Une erreur réseau ne permet pas de réessayer",
  changement: "Conserver la saisie et proposer un nouvel essai",
  verification: "Simuler un échec puis une réponse valide",
};

This first delivery may reveal a missing way to simulate a network error. Preparing that case then becomes a next task, with a specific reason to address it. The fix thus helps identify what actually slows work on this project.

Keep rewriting as a decision that needs evidence

A rewrite may become appropriate when specific constraints require it. It then deserves an explicit comparison with gradual evolution: which uses must be preserved, how will data migrate, which parts can coexist and how will the result be verified? Understanding the existing system also helps answer these questions.

To begin, choose a flow with the person who uses it and record the behaviour to preserve before touching code. The first fix will provide evidence for deciding what comes next.

Share on LinkedIn (new tab)

Keep reading
All articles in this category