Lectures

We break down streaming work with JSON in Go: how the json.Marshal/ json.Unmarshal approach differs from json.Encoder/ json.Decoder, how to write JSON directly to an io.Writer and read from an io.Reader, how “value-by-value” reading works up to io.EOF, and how to choose a data protocol: an array or a stream of objects.

We break down how to make JSON parsing in Go “strict”: we enable Decoder.DisallowUnknownFields(), understand why this works only when decoding into a struct (and not into a map), and add a check that after the first Decode the stream really ends ( io.EOF). In the end, we get a reusable helper for “one JSON with no trailing data” and a neat error wrapper via fmt.Errorf with %w.

We’ll break down two practical techniques for safe JSON parsing in Go: how to avoid the “silent” conversion of numbers into float64 during dynamic decoding via any/ map[string]any using Decoder.UseNumber() and json.Number, and how to build a two-step “envelope → payload” parse using json.RawMessage, command dispatch via switch, and proper error wrapping ( %w).

We break down a practical JSON versioning pattern via an “envelope” { version , data }: how to read the version as the first step, how to lazily parse the payload via json.RawMessage, how to keep migrations in DTO structs (v1/v2) and convert them to the domain model. We’ll show strict envelope decoding with json.Decoder and DisallowUnknownFields(), as well as ReadTasks/ WriteTasks functions (read all versions, write only the current one).

We break down how, in Go, to turn “raw” JSON error output into user-friendly messages: add operation context via fmt.Errorf and "%w", preserve the root cause for errors.Is/ errors.As, recognize typed errors from encoding/json (for example, * json.SyntaxError and * json.UnmarshalTypeError), convert Offset into line/column, and define an application boundary (CLI) where a “friendly message” is separated from the “technical cause.”