Types, zero values, and control flow

Go assigns a static type to every variable. A declaration can name the type explicitly (var retries int) or infer it (retries := 3). Inference does not make the variable dynamically typed: assigning a string later remains a compile error. Constants are more flexible than variables until a context gives them a type. Conversion such as int64(retries) is explicit, and converting a larger value to a smaller integer type can truncate rather than return an error. Choose widths at storage or wire boundaries; int is convenient for indexing but its size follows the target architecture.

Zero values are part of the design

Uninitialized integers are zero, booleans false, strings empty, pointers/maps/slices/channels nil, and structs recursively contain their fields' zero values. A zero value is useful when it represents an ordinary safe state. It cannot always represent absence: zero is a real measurement. Pair a value with a boolean or use an optional pointer when missing and zero differ. Never dereference a nil pointer.

var attempts int
var enabled bool
var name string
var optional *int
fmt.Println(attempts, enabled, name == "", optional == nil)

Conditions describe the domain

Go has no truthiness conversion: if attempts does not compile; write if attempts > 0. An if initializer limits a temporary to its condition and branches, for example if value, ok := values[key]; ok { ... }. A switch selects one branch without implicit fallthrough. Use explicit cases for known states and a default error for unsupported input when the contract requires one. Rejecting impossible states early keeps the remaining loop invariant simple.

for handles counting, condition-controlled loops, and unbounded loops. Range over a slice yields an index and a copied element. Modifying the copied element does not update the original struct in the slice; use the index or a pointer when mutation is intended. break leaves the nearest loop; continue starts the next iteration. Nested loops sometimes need labels, but extracting a small function can make the exit condition clearer.

Summaries need an invariant

A running total summarizes only the valid measurements already read. A first valid value establishes extrema; later improvements update them. Initializing minimum to zero would invent an observation when every valid value is positive. Counting invalid records by accident can corrupt an average even when the total is correct. The empty/all-invalid input contract must therefore be explicit.

These rules matter in request validation, metrics aggregation and configuration handling. Integer overflow is not automatically a checked error, so production code needs bounds or an explicitly checked arithmetic policy. This exercise keeps every intermediate total within int and focuses on static state, zero values, and control flow. A fixed number of passes remains O(n), while one pass keeps the summary local.