Generating code in Go

Encourage good patterns, and improve developers experience through code generation. Telemetry, decorators, ORMs, schema-first approach - automate these chores, without increasing complexity, without runtime errors, without developer having to remember about them.

Place
Golab #2
Length
60 min
When
November 20th, 2023
15:30

Abstract

When I started my adventure with Go, I couldn't stand generated mocks. Coming from a more dynamic language, I thought that I can recapture that "magic" using Go's reflection... Here I am, 5 years later, advocating for the opposite!

One of the first "Aha!" moments was using gowrap to generate decorators. I believe that good observability in the system primarily requires good discipline. This is where generating decorators, that bring traces, logs and metrics shines. Developers can focus on generating business value, while telemetry will come from coding to the interface, with go:generate one line above. We haven't had issues with missing traces ever since.

The second big leap in productivity came from generating code based on schema. We are using GraphQL and gRPC. Especially for GraphQL, writing all required scaffolding was quite a chore. This often lead to some obscure bugs with types, if we didn't check everything during runtime. Once we changed to gqlgen - a schema-first Go generator for GraphQL - all of that changed. Not only we squashed many bugs right after migrating, because compiler showed all the issues to us, but collaboration between backend and frontend improved. When you put schema first, and it is your source of truth, you can discuss it with your coworkers sooner, and iterate on implementation, or improvement, quicker.

The last part (for now) is about database. While schema-first approach also contributed to change - after all, for SQL databases you probably already have your table definitions, and migrations live in your codebase - undetected breaking changes, and high complexity of reflection based ORMs is what pushed us to sqlboiler. This library generates ORM for you, based on a database you already have. If you never had to go through reflection-based ORM, you might be surprised, that understanding what happens under a simple "Find by ID" query can take you a day. This is something we all hit, once the magic of ORM fails us, and we have to dig into what went wrong. Generated code doesn't have that issue. Computers do not mind repeating themselves, and will not make a mistake when writing the same line of code again, and again, and again. The result is a simple "Find" function, that has 10 lines, and has SQL query you would have written yourself in database console. Oh, and breaking changes? They will be plain as day in pull request.

Slides

Video

Made by DEVELER

GoLab is a conference made by Develer.
Develer is a company based in Campi Bisenzio, near Florence. Our motto is : "Technology to give life to your products". We produce hardware and software to create exceptional products and to improve industrial processes and people's well being.
In Develer we have passion for the new technologies and we offer our clients effective solutions that are also efficient, simple and safe for the end users. We also believe in a friendly and welcoming environment where anybody can give their contribution. This passion and this vision are what we've been driven to organize our conference "made by developers for developers".


Subscribe to our newsletter

We hate spam just as much as you do, which is why we promise to only send you relevant communications. We respect your privacy and will never share your information with third parties.
©2024 GoLab | The international conference on Go in Florence-Design & devCantiere Creativo-Made withDatoCMS