1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
|
package hashstructure
import (
"encoding/binary"
"fmt"
"hash"
"hash/fnv"
"reflect"
)
// HashOptions are options that are available for hashing.
type HashOptions struct {
// Hasher is the hash function to use. If this isn't set, it will
// default to FNV.
Hasher hash.Hash64
// TagName is the struct tag to look at when hashing the structure.
// By default this is "hash".
TagName string
}
// Hash returns the hash value of an arbitrary value.
//
// If opts is nil, then default options will be used. See HashOptions
// for the default values.
//
// Notes on the value:
//
// * Unexported fields on structs are ignored and do not affect the
// hash value.
//
// * Adding an exported field to a struct with the zero value will change
// the hash value.
//
// For structs, the hashing can be controlled using tags. For example:
//
// struct {
// Name string
// UUID string `hash:"ignore"`
// }
//
// The available tag values are:
//
// * "ignore" - The field will be ignored and not affect the hash code.
//
// * "set" - The field will be treated as a set, where ordering doesn't
// affect the hash code. This only works for slices.
//
func Hash(v interface{}, opts *HashOptions) (uint64, error) {
// Create default options
if opts == nil {
opts = &HashOptions{}
}
if opts.Hasher == nil {
opts.Hasher = fnv.New64()
}
if opts.TagName == "" {
opts.TagName = "hash"
}
// Reset the hash
opts.Hasher.Reset()
// Create our walker and walk the structure
w := &walker{
h: opts.Hasher,
tag: opts.TagName,
}
return w.visit(reflect.ValueOf(v), nil)
}
type walker struct {
h hash.Hash64
tag string
}
type visitOpts struct {
// Flags are a bitmask of flags to affect behavior of this visit
Flags visitFlag
// Information about the struct containing this field
Struct interface{}
StructField string
}
func (w *walker) visit(v reflect.Value, opts *visitOpts) (uint64, error) {
// Loop since these can be wrapped in multiple layers of pointers
// and interfaces.
for {
// If we have an interface, dereference it. We have to do this up
// here because it might be a nil in there and the check below must
// catch that.
if v.Kind() == reflect.Interface {
v = v.Elem()
continue
}
if v.Kind() == reflect.Ptr {
v = reflect.Indirect(v)
continue
}
break
}
// If it is nil, treat it like a zero.
if !v.IsValid() {
var tmp int8
v = reflect.ValueOf(tmp)
}
// Binary writing can use raw ints, we have to convert to
// a sized-int, we'll choose the largest...
switch v.Kind() {
case reflect.Int:
v = reflect.ValueOf(int64(v.Int()))
case reflect.Uint:
v = reflect.ValueOf(uint64(v.Uint()))
case reflect.Bool:
var tmp int8
if v.Bool() {
tmp = 1
}
v = reflect.ValueOf(tmp)
}
k := v.Kind()
// We can shortcut numeric values by directly binary writing them
if k >= reflect.Int && k <= reflect.Complex64 {
// A direct hash calculation
w.h.Reset()
err := binary.Write(w.h, binary.LittleEndian, v.Interface())
return w.h.Sum64(), err
}
switch k {
case reflect.Array:
var h uint64
l := v.Len()
for i := 0; i < l; i++ {
current, err := w.visit(v.Index(i), nil)
if err != nil {
return 0, err
}
h = hashUpdateOrdered(w.h, h, current)
}
return h, nil
case reflect.Map:
var includeMap IncludableMap
if opts != nil && opts.Struct != nil {
if v, ok := opts.Struct.(IncludableMap); ok {
includeMap = v
}
}
// Build the hash for the map. We do this by XOR-ing all the key
// and value hashes. This makes it deterministic despite ordering.
var h uint64
for _, k := range v.MapKeys() {
v := v.MapIndex(k)
if includeMap != nil {
incl, err := includeMap.HashIncludeMap(
opts.StructField, k.Interface(), v.Interface())
if err != nil {
return 0, err
}
if !incl {
continue
}
}
kh, err := w.visit(k, nil)
if err != nil {
return 0, err
}
vh, err := w.visit(v, nil)
if err != nil {
return 0, err
}
fieldHash := hashUpdateOrdered(w.h, kh, vh)
h = hashUpdateUnordered(h, fieldHash)
}
return h, nil
case reflect.Struct:
var include Includable
parent := v.Interface()
if impl, ok := parent.(Includable); ok {
include = impl
}
t := v.Type()
h, err := w.visit(reflect.ValueOf(t.Name()), nil)
if err != nil {
return 0, err
}
l := v.NumField()
for i := 0; i < l; i++ {
if v := v.Field(i); v.CanSet() || t.Field(i).Name != "_" {
var f visitFlag
fieldType := t.Field(i)
if fieldType.PkgPath != "" {
// Unexported
continue
}
tag := fieldType.Tag.Get(w.tag)
if tag == "ignore" {
// Ignore this field
continue
}
// Check if we implement includable and check it
if include != nil {
incl, err := include.HashInclude(fieldType.Name, v)
if err != nil {
return 0, err
}
if !incl {
continue
}
}
switch tag {
case "set":
f |= visitFlagSet
}
kh, err := w.visit(reflect.ValueOf(fieldType.Name), nil)
if err != nil {
return 0, err
}
vh, err := w.visit(v, &visitOpts{
Flags: f,
Struct: parent,
StructField: fieldType.Name,
})
if err != nil {
return 0, err
}
fieldHash := hashUpdateOrdered(w.h, kh, vh)
h = hashUpdateUnordered(h, fieldHash)
}
}
return h, nil
case reflect.Slice:
// We have two behaviors here. If it isn't a set, then we just
// visit all the elements. If it is a set, then we do a deterministic
// hash code.
var h uint64
var set bool
if opts != nil {
set = (opts.Flags & visitFlagSet) != 0
}
l := v.Len()
for i := 0; i < l; i++ {
current, err := w.visit(v.Index(i), nil)
if err != nil {
return 0, err
}
if set {
h = hashUpdateUnordered(h, current)
} else {
h = hashUpdateOrdered(w.h, h, current)
}
}
return h, nil
case reflect.String:
// Directly hash
w.h.Reset()
_, err := w.h.Write([]byte(v.String()))
return w.h.Sum64(), err
default:
return 0, fmt.Errorf("unknown kind to hash: %s", k)
}
return 0, nil
}
func hashUpdateOrdered(h hash.Hash64, a, b uint64) uint64 {
// For ordered updates, use a real hash function
h.Reset()
// We just panic if the binary writes fail because we are writing
// an int64 which should never be fail-able.
e1 := binary.Write(h, binary.LittleEndian, a)
e2 := binary.Write(h, binary.LittleEndian, b)
if e1 != nil {
panic(e1)
}
if e2 != nil {
panic(e2)
}
return h.Sum64()
}
func hashUpdateUnordered(a, b uint64) uint64 {
return a ^ b
}
// visitFlag is used as a bitmask for affecting visit behavior
type visitFlag uint
const (
visitFlagInvalid visitFlag = iota
visitFlagSet = iota << 1
)
|