Lectures

We break down how to build a CLI with subcommands in Go using flag.FlagSet: separate flags for each command, isolated parsing, error control via flag.ContinueOnError, output configuration via SetOutput, and a custom Usage. At the end, we cover common mistakes when using FlagSet and how to avoid them.

We break down a convenient CLI approach: “subcommands instead of a mode flag”. The launch format is app <cmd> [flags] [args], the dispatcher’s role (routing), extracting the command from os.Args, minimal routing via switch, and a scalable option via a map[ string] handler table. We also pin down the responsibility boundary: the dispatcher chooses the command, while flags are parsed by the command itself (no global flag.Parse()), and we improve UX with usage and clear errors.

We cover how to correctly read positional arguments in a Go CLI via fs.Parse ( argv ) and the methods fs.Args(), fs.NArg(), fs.Arg(i); why you should not work with the “raw” argv; how to design a strict command contract ( 0, 1, N arguments); how to parse types via strconv.Atoi; and how to return meaningful usage errors with the ErrUsage marker and the %w wrapper.

We break down why CLI apps need a stable exit-code contract 0/ 1/ 2, how to separate “invalid invocation” errors from runtime errors via the sentinel ErrUsage and errors.Is, why os.Exit “breaks” defer, and how to build an architecture with a single exit point: run returns error, and the top level prints diagnostics to stderr and maps errors to an exit code.

We break down a practical approach to errors in a Go CLI: how to separate a short user-facing message from the technical cause in the error chain, print an error exactly once at the application boundary, assign exit codes (

0

/

1

/

2

) consistently via errors.Is and the ErrUsage marker, and how to carefully extract causes for diagnostics (a debug chain) without “wall-of-text” output.