我今天写了一个实现。唉，R5RS 既没有字节向量也没有二进制 I/O，因此它使用 R7RS API 来处理字节向量和二进制 I/O。将这些 API 桥接到您的 Scheme 实现的本机 API 应该很容易（例如，我实际上在 Racket 和 Guile 上测试了我的实现）。
一些注意事项：
- 此代码假定区分大小写。这是 R7RS 的默认设置，但不是 R5RS，因此如果您使用 R5RS 实现，请小心。
- 它需要 SRFI1, 26, 43, and 60.
- 我强调优雅和清晰而不是速度。事实上，代码相当慢。
- 与我的个人资料所述相反，我仅根据以下许可授权此代码阿帕奇许可证 2.0（除了标准 Stack Overflow 许可证之外抄送-SA 3.0），并且不在 CC0 或类似公共领域的任何内容下。
无论如何，事不宜迟，这就是（也可以作为要点):
;;; Auxiliary definitions to avoid having to use giant tables of constants.
(define primes80 '(2 3 5 7 11 13 17 19 23 29 31 37 41 43 47 53 59 61 67 71 73
79 83 89 97 101 103 107 109 113 127 131 137 139 149 151 157
163 167 173 179 181 191 193 197 199 211 223 227 229 233 239
241 251 257 263 269 271 277 281 283 293 307 311 313 317 331
337 347 349 353 359 367 373 379 383 389 397 401 409))
(define (sqrt x)
(fold (lambda (_ y) (/ (+ (/ x y) y) 2)) 4 (iota 7)))
(define (cbrt x)
(fold (lambda (_ y) (/ (+ (/ x y y) y y) 3)) 4 (iota 8)))
(define (frac x scale base)
(bitwise-and (floor (* x (arithmetic-shift 1 scale)))
(- (arithmetic-shift 1 base) 1)))
;;; The actual initialisation and constant values.
(define sha1-init '(#x67452301 #xefcdab89 #x98badcfe #x10325476 #xc3d2e1f0))
(define sha2-init (map (lambda (x) (frac (sqrt x) 64 64)) (take primes80 16)))
(define-values (sha512-init sha384-init) (split-at sha2-init 8))
(define sha256-init (map (cut arithmetic-shift <> -32) sha512-init))
(define sha224-init (map (cut frac <> 0 32) sha384-init))
(define sha1-const (map (lambda (x) (frac (sqrt x) 30 32)) '(2 3 5 10)))
(define sha512-const (map (lambda (x) (frac (cbrt x) 64 64)) primes80))
(define sha256-const (map (cut arithmetic-shift <> -32) (take sha512-const 64)))
;;; Utility functions used by the compression and driver functions.
(define (u32+ . xs) (bitwise-and (apply + xs) #xffffffff))
(define (u64+ . xs) (bitwise-and (apply + xs) #xffffffffffffffff))
(define (bitwise-majority x y z)
(bitwise-xor (bitwise-and x y) (bitwise-and x z) (bitwise-and y z)))
(define (bytevector-be-ref bv base n)
(let loop ((res 0) (i 0))
(if (< i n)
(loop (+ (arithmetic-shift res 8) (bytevector-u8-ref bv (+ base i)))
(+ i 1))
res)))
(define (bytevector-u64-ref bv i)
(bytevector-be-ref bv (arithmetic-shift i 3) 8))
(define (bytevector-u32-ref bv i)
(bytevector-be-ref bv (arithmetic-shift i 2) 4))
(define (bytevector-be-set! bv base n val)
(let loop ((i n) (val val))
(when (positive? i)
(bytevector-u8-set! bv (+ base i -1) (bitwise-and val 255))
(loop (- i 1) (arithmetic-shift val -8)))))
(define (md-pad! bv offset count counter-size)
(define block-size (bytevector-length bv))
(unless (negative? offset)
(bytevector-u8-set! bv offset #x80))
(let loop ((i (+ offset 1)))
(when (< i block-size)
(bytevector-u8-set! bv i 0)
(loop (+ i 1))))
(when count
(bytevector-be-set! bv (- block-size counter-size) counter-size
(arithmetic-shift count 3))))
(define (hash-state->bytevector hs trunc word-size)
(define result (make-bytevector (* trunc word-size)))
(for-each (lambda (h i)
(bytevector-be-set! result i word-size h))
hs (iota trunc 0 word-size))
result)
;;; The compression functions.
(define (sha2-compress K Σ0 Σ1 σ0 σ1 mod+ getter hs)
(define W (vector->list (apply vector-unfold
(lambda (_ a b c d e f g h i j k l m n o p)
(values a b c d e f g h i j k l m n o p
(mod+ a (σ0 b) j (σ1 o))))
(length K)
(list-tabulate 16 getter))))
(define (loop k w a b c d e f g h)
(if (null? k)
(map mod+ hs (list a b c d e f g h))
(let ((T1 (mod+ h (Σ1 e) (bitwise-if e f g) (car k) (car w)))
(T2 (mod+ (Σ0 a) (bitwise-majority a b c))))
(loop (cdr k) (cdr w) (mod+ T1 T2) a b c (mod+ d T1) e f g))))
(apply loop K W hs))
(define (sha512-compress bv hs)
(define (rotr x y) (rotate-bit-field x (- y) 0 64))
(define (shr x y) (arithmetic-shift x (- y)))
(sha2-compress sha512-const
(lambda (x) (bitwise-xor (rotr x 28) (rotr x 34) (rotr x 39)))
(lambda (x) (bitwise-xor (rotr x 14) (rotr x 18) (rotr x 41)))
(lambda (x) (bitwise-xor (rotr x 1) (rotr x 8) (shr x 7)))
(lambda (x) (bitwise-xor (rotr x 19) (rotr x 61) (shr x 6)))
u64+ (cut bytevector-u64-ref bv <>) hs))
(define (sha256-compress bv hs)
(define (rotr x y) (rotate-bit-field x (- y) 0 32))
(define (shr x y) (arithmetic-shift x (- y)))
(sha2-compress sha256-const
(lambda (x) (bitwise-xor (rotr x 2) (rotr x 13) (rotr x 22)))
(lambda (x) (bitwise-xor (rotr x 6) (rotr x 11) (rotr x 25)))
(lambda (x) (bitwise-xor (rotr x 7) (rotr x 18) (shr x 3)))
(lambda (x) (bitwise-xor (rotr x 17) (rotr x 19) (shr x 10)))
u32+ (cut bytevector-u32-ref bv <>) hs))
(define (sha1-compress bv hs)
(define (getter x) (bytevector-u32-ref bv x))
(define (rotl x y) (rotate-bit-field x y 0 32))
(define W (vector->list (apply vector-unfold
(lambda (_ a b c d e f g h i j k l m n o p)
(values a b c d e f g h i j k l m n o p
(rotl (bitwise-xor a c i n) 1)))
80
(list-tabulate 16 getter))))
(define (outer f k w a b c d e)
(if (null? k)
(map u32+ hs (list a b c d e))
(let inner ((i 0) (w w) (a a) (b b) (c c) (d d) (e e))
(if (< i 20)
(let ((T (u32+ (rotl a 5) ((car f) b c d) e (car k) (car w))))
(inner (+ i 1) (cdr w) T a (rotl b 30) c d))
(outer (cdr f) (cdr k) w a b c d e)))))
(apply outer (list bitwise-if bitwise-xor bitwise-majority bitwise-xor)
sha1-const W hs))
;;; The Merkle-Damgård "driver" function.
(define (md-loop init compress block-size trunc word-size counter-size in)
(define leftover (- block-size counter-size))
(define bv (make-bytevector block-size))
(define pad! (cut md-pad! bv <> <> counter-size))
(define hs->bv (cut hash-state->bytevector <> trunc word-size))
(let loop ((count 0) (hs init))
(define read-size (read-bytevector! bv in))
(cond ((eof-object? read-size)
(pad! 0 count)
(hs->bv (compress bv hs)))
((= read-size block-size)
(loop (+ count read-size) (compress bv hs)))
((< read-size leftover)
(pad! read-size (+ count read-size))
(hs->bv (compress bv hs)))
(else
(pad! read-size #f)
(let ((pen (compress bv hs)))
(pad! -1 (+ count read-size))
(hs->bv (compress bv pen)))))))
;;; SHA-512/t stuff.
(define sha512/t-init (map (cut bitwise-xor <> #xa5a5a5a5a5a5a5a5) sha512-init))
(define (make-sha512/t-init t)
(define key (string->utf8 (string-append "SHA-512/" (number->string t))))
(define size (bytevector-length key))
(define bv (make-bytevector 128))
(bytevector-copy! bv 0 key)
(md-pad! bv size size 16)
(sha512-compress bv sha512/t-init))
(define (make-sha512/t t)
(define init (make-sha512/t-init t))
(define words (arithmetic-shift t -6))
(if (zero? (bitwise-and t 63))
(cut md-loop init sha512-compress 128 words 8 16 <>)
(lambda (in)
(bytevector-copy
(md-loop init sha512-compress 128 (ceiling words) 8 16 in)
0 (arithmetic-shift t -3)))))
;;; Public entry points.
(define sha1 (cut md-loop sha1-init sha1-compress 64 5 4 8 <>))
(define sha224 (cut md-loop sha224-init sha256-compress 64 7 4 8 <>))
(define sha256 (cut md-loop sha256-init sha256-compress 64 8 4 8 <>))
(define sha384 (cut md-loop sha384-init sha512-compress 128 6 8 16 <>))
(define sha512 (cut md-loop sha512-init sha512-compress 128 8 8 16 <>))
(define sha512/256 (make-sha512/t 256))
(define sha512/224 (make-sha512/t 224))
我实现了 FIPS 180-4 中的所有算法，但您可以删除不需要的任何算法。
如前所述，我在 Racket 上对此进行了测试；我添加的用于桥接到 Racket 的 API 的定义如下：
#lang racket
(require (only-in srfi/1 iota)
(only-in srfi/26 cut)
(only-in srfi/43 vector-unfold)
(only-in srfi/60 bitwise-if rotate-bit-field)
(rename-in racket/base [build-list list-tabulate]
[bytes-copy! bytevector-copy!]
[bytes-length bytevector-length]
[bytes-ref bytevector-u8-ref]
[bytes-set! bytevector-u8-set!]
[foldl fold]
[make-bytes make-bytevector]
[read-bytes! read-bytevector!]
[string->bytes/utf-8 string->utf8]
[subbytes bytevector-copy]))
以下是 Guile 的定义（需要 2.0.11 或更高版本）：
(use-modules (srfi srfi-1) (srfi srfi-26) (srfi srfi-43) (srfi srfi-60)
(rnrs bytevectors) (ice-9 binary-ports))
(define* (bytevector-copy bv #:optional (start 0) (end (bytevector-length bv)))
(define copy (make-bytevector (- end start)))
(bytevector-copy! copy 0 bv start end)
copy)
(define* (bytevector-copy! to at from #:optional (start 0)
(end (bytevector-length from)))
((@ (rnrs bytevectors) bytevector-copy!) from start to at (- end start)))
(define* (read-bytevector! bv #:optional (port (current-input-port)) (start 0)
(end (bytevector-length bv)))
(get-bytevector-n! port bv start (- end start)))
为您选择的实现制作类似的东西应该很容易。
我还有一个函数，可以将输出打印为十六进制字符串，以便与各种命令行 SHA-1 和 SHA-2 实用程序（例如，sha1sum, sha256sum, sha512sum, etc.):
(define (hex bv)
(define out (open-output-string))
(do ((i 0 (+ i 1)))
((>= i (bytevector-length bv)) (get-output-string out))
(let-values (((q r) (truncate/ (bytevector-u8-ref bv i) 16)))
(display (number->string q 16) out)
(display (number->string r 16) out))))
