go-data-structures

name	go-data-structures
description	Correct use of Go's built-in data structures: slices (nil vs empty, append semantics, aliasing, preallocation), maps (comma-ok, sets, iteration order), arrays, and choosing between them. Use when: "slice vs array", "nil slice", "empty slice", "preallocate", "map iteration", "use a set in Go", "slice aliasing", "append gotcha", "copy a slice", "sync.Map or mutex". Do NOT use for: protecting structures shared across goroutines (use go-concurrency-review), allocation profiling (use go-performance-review), or generic container design (use go-design-patterns).
license	MIT
metadata	{"version":"1.0.0"}

Go Data Structures

Slices and maps look simple and hide the sharpest edges in the language. These rules prevent the aliasing, nil, and iteration bugs that survive code review.

1. Nil Slice vs Empty Slice

var a []int          // nil slice — len 0, cap 0, no allocation
b := []int{}         // empty slice — len 0, allocated header
c := make([]int, 0)  // empty slice — same as b

• len, cap, range, and append treat all three identically.
• Prefer the nil slice as the "no elements" value; don't allocate just to return "empty".
• Exception: JSON. nil marshals to null, empty marshals to []. If the API contract requires [], return an empty slice explicitly.
• Never distinguish nil from empty in logic — check len(s) == 0.

2. Append Semantics and Aliasing

append MAY return the same backing array or a new one. Both cases bite:

// ❌ Bad — result may alias the input
func addSuffix(base []string) []string {
    return append(base, "suffix") // if cap(base) > len(base),
}                                 // this WRITES INTO base's array

// ✅ Good — force a copy when the input must not be touched
func addSuffix(base []string) []string {
    out := make([]string, len(base), len(base)+1)
    copy(out, base)
    return append(out, "suffix")
}

// ❌ Bad — subslice keeps the whole 64 MB alive
func header(big []byte) []byte {
    return big[:512] // backing array is still the full big
}

// ✅ Good — copy the window you keep
func header(big []byte) []byte {
    return slices.Clone(big[:512]) // Go 1.21+; or copy() manually
}

Rule: a function either owns a slice or copies it. Returning a subslice of a caller's slice, or appending to one, silently shares memory.

3. Preallocation

When the final size is known or bounded, allocate once:

// ✅ Good — one allocation
names := make([]string, 0, len(users))
for _, u := range users {
    names = append(names, u.Name)
}

// ❌ Bad — repeated growth and copying
var names []string
for _, u := range users {
    names = append(names, u.Name)
}

Same for maps: make(map[string]int, len(items)). Don't preallocate when the size is unknown — a wrong large cap wastes memory; append growth is fine for cold paths.

4. Map Essentials

// Comma-ok distinguishes "missing" from "zero value"
count, ok := hits[key]
if !ok { /* key absent */ }

// Zero value reads are safe; writes to a nil map PANIC
var m map[string]int
_ = m["x"]      // 0, fine
m["x"] = 1      // panic: assignment to entry in nil map — make() first

// Iteration order is RANDOM and differs between runs.
// Sort keys when output must be deterministic:
keys := slices.Sorted(maps.Keys(m)) // Go 1.23+
for _, k := range keys {
    fmt.Println(k, m[k])
}

• Map values are not addressable: m[k].Field = v doesn't compile for struct values. Use a map of pointers, or read-modify-write.
• Deleting during range is safe; inserting during range is unspecified (the new key may or may not be visited).

5. Sets

The idiomatic set is a map with empty-struct values:

seen := make(map[string]struct{}, len(items))
for _, it := range items {
    if _, dup := seen[it.ID]; dup {
        continue
    }
    seen[it.ID] = struct{}{}
    process(it)
}

struct{} occupies zero bytes; map[string]bool also works and reads better when you'll test membership with if seen[id].

6. Arrays vs Slices

• Arrays ([4]byte) are values: assignment and passing copy the whole array. Comparable with == when elements are comparable.
• Use arrays for fixed-size data with value semantics: hashes ([32]byte), IPv4 addresses, fixed matrices, map keys.
• Everything else is a slice. A function taking [100]int copies 800 bytes per call — almost always wrong.

7. Choosing a Structure

Need	Use
Ordered collection, growable	[]T
Membership / dedup	map[K]struct{}
Key→value lookup	map[K]V
Fixed size, value semantics, comparable	[N]T array
FIFO queue (single goroutine)	slice with head index, or container/list for heavy churn
Stack	slice + append / s[:len(s)-1]
Concurrent map, write-once read-many keys	sync.Map — otherwise mutex + map

sync.Map is a special-case tool (append-only caches, disjoint key sets). Default to map + sync.RWMutex; see the concurrency skill for locking patterns.

Verification Checklist

1. No logic distinguishes nil slice from empty slice; len() used for emptiness
2. JSON-facing slices explicitly empty (not nil) where the contract requires []
3. No append to a slice the function doesn't own; copies made explicit
4. No long-lived subslices of large arrays without slices.Clone/copy
5. Slices and maps preallocated with capacity when size is known
6. Comma-ok used wherever "missing" differs from zero value
7. No writes to possibly-nil maps
8. Deterministic output paths sort map keys before iteration
9. Sets built as map[K]struct{} (or map[K]bool for readability)
10. Arrays only where value semantics or comparability is the point