Support Memory{UInt8} and AbstractVector{UInt8} as input. (#12)

* Support Memory and AbstractVector as input.

* fix CI

* Update src/CRC32.jl

Co-authored-by: Steven G. Johnson <stevenj@alum.mit.edu>

* Update readme

* Update src/CRC32.jl

Co-authored-by: Steven G. Johnson <stevenj@alum.mit.edu>

* remove unneeded type checks

---------

Co-authored-by: Steven G. Johnson <stevenj@alum.mit.edu>

Author: nhz2

.github/workflows/CI.yml

@@ -4,7 +4,7 @@ on:
   schedule:
     - cron: '0 2 * * *'  # Daily at 2 AM UTC (8 PM CST)
   push:
-    branches: [master]
+    branches: [main]
     tags: ["*"]
   pull_request:
 concurrency:

Project.toml

@@ -10,7 +10,8 @@ Zlib_jll = "83775a58-1f1d-513f-b197-d71354ab007a"
 julia = "1.3"
 
 [extras]
+OffsetArrays = "6fe1bfb0-de20-5000-8ca7-80f57d26f881"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [targets]
-test = ["Test"]
+test = ["OffsetArrays", "Test"]

README.md

@@ -26,10 +26,10 @@ CRC-32 checksum.
 crc32(data, crc::UInt32=0x00000000)
 ```
 
-Computes the CRC-32 checksum (ISO 3309, ITU-T V.42, CRC-32-IEEE) of the given `data`, which can be
-an `Array{UInt8}`, a contiguous subarray thereof, or a `String`.  Optionally, you can pass
-a starting `crc` integer to be mixed in with the checksum.  The `crc` parameter
-can be used to compute a checksum on data divided into chunks: performing
+Compute the CRC-32 checksum (ISO 3309, ITU-T V.42, CRC-32-IEEE) of the given `data`, which can be
+an `Array{UInt8}`, a contiguous subarray thereof, an `AbstractVector{UInt8}`, or a `String`.
+Optionally, you can pass a starting `crc` integer to be mixed in with the checksum.
+The `crc` parameter can be used to compute a checksum on data divided into chunks: performing
 `crc32(data2, crc32(data1))` is equivalent to the checksum of `[data1; data2]`.
 
 There is also a method `crc32(io, nb, crc)` to checksum `nb` bytes from

src/CRC32.jl

@@ -14,17 +14,25 @@ module CRC32
 export crc32
 
 # contiguous byte arrays compatible with C `unsigned char *` API of zlib
-const ByteArray = Union{Array{UInt8},
-                        Base.FastContiguousSubArray{UInt8,N,<:Array{UInt8}} where N,
-                        Base.CodeUnits{UInt8, String}, Base.CodeUnits{UInt8, SubString{String}}}
+if VERSION ≥ v"1.11"
+    const ByteArray = Union{Array{UInt8},
+                            Memory{UInt8},
+                            Base.FastContiguousSubArray{UInt8,N,<:Array{UInt8}} where N,
+                            Base.FastContiguousSubArray{UInt8,1,Memory{UInt8}},
+                            Base.CodeUnits{UInt8, String}, Base.CodeUnits{UInt8, SubString{String}}}
+else
+    const ByteArray = Union{Array{UInt8},
+                            Base.FastContiguousSubArray{UInt8,N,<:Array{UInt8}} where N,
+                            Base.CodeUnits{UInt8, String}, Base.CodeUnits{UInt8, SubString{String}}}
+end
 
 """
     crc32(data, crc::UInt32=0x00000000)
 
 Compute the CRC-32 checksum (ISO 3309, ITU-T V.42, CRC-32-IEEE) of the given `data`, which can be
-an `Array{UInt8}`, a contiguous subarray thereof, or a `String`.  Optionally, you can pass
-a starting `crc` integer to be mixed in with the checksum.  The `crc` parameter
-can be used to compute a checksum on data divided into chunks: performing
+an `Array{UInt8}`, a contiguous subarray thereof, an `AbstractVector{UInt8}`, or a `String`.
+Optionally, you can pass a starting `crc` integer to be mixed in with the checksum.
+The `crc` parameter can be used to compute a checksum on data divided into chunks: performing
 `crc32(data2, crc32(data1))` is equivalent to the checksum of `[data1; data2]`.
 
 There is also a method `crc32(io, nb, crc)` to checksum `nb` bytes from
@@ -39,6 +47,7 @@ For a `String`, note that the result is specific to the UTF-8 encoding
 function crc32 end
 
 crc32(a::ByteArray, crc::UInt32=0x00000000) = _crc32(a, crc)
+crc32(a::AbstractVector{UInt8}, crc::UInt32=0x00000000) = _crc32(a, crc)
 crc32(s::Union{String, SubString{String}}, crc::UInt32=0x00000000) = _crc32(s, crc)
 
 """
@@ -66,6 +75,21 @@ function _crc32(s::Union{String, SubString{String}}, crc::UInt32=0x00000000)
     unsafe_crc32(s, sizeof(s) % Csize_t, crc)
 end
 
+function _crc32(a::AbstractVector{UInt8}, crc::UInt32=0x00000000)
+    # use block size 24576=8192*3, since that is the threshold for
+    # 3-way parallel SIMD code in the underlying jl_crc32 C function.
+    last = lastindex(a)
+    nb = length(a)
+    buf = Vector{UInt8}(undef, min(nb, 24576))
+    while nb > 0
+        n = min(nb, 24576)
+        copyto!(buf, 1, a, last - nb + 1, n)
+        crc = unsafe_crc32(buf, n % Csize_t, crc)
+        nb -= n
+    end
+    return crc
+end
+
 function _crc32(io::IO, nb::Integer, crc::UInt32=0x00000000)
     nb < 0 && throw(ArgumentError("number of bytes to checksum must be ≥ 0, got $nb"))
     # use block size 24576=8192*3, since that is the threshold for

test/runtests.jl

@@ -1,5 +1,6 @@
 using CRC32
 using Test
+using OffsetArrays: Origin
 
 # based on julia/stdlib/CRC32/test/runtests.jl
 
@@ -9,6 +10,12 @@ using Test
         s = String(UInt8[1:n;])
         ss = SubString(String(UInt8[0:(n+1);]), 2:(n+1))
         @test crc32(UInt8[1:n;]) == crc == crc32(s) == crc32(ss) == crc32(codeunits(s)) == crc32(codeunits(ss))
+        @test crc == crc32(UInt8(1):UInt8(n))
+        if VERSION ≥ v"1.11"
+            m = Memory{UInt8}(undef, n)
+            m .= 1:n
+            @test crc == crc32(m)
+        end
     end
 
     # test that crc parameter is equivalent to checksum of concatenated data,
@@ -59,4 +66,9 @@ using Test
     if Sys.WORD_SIZE == 64
         @test crc32(zeros(UInt8, 2^32)) == 0xd202ef8d #crc32(zeros(UInt8, 2^31), crc32(zeros(UInt8, 2^31)))
     end
+
+    # Test crc of AbstractVector{UInt8}
+    a = view(rand(UInt8, 300000), 1:2:300000)
+    @test crc32(a) == crc32(collect(a))
+    @test crc32(Origin(0)(b"hello")) == crc32(b"hello")
 end