Java has a lot of different ways to represent a stream of bytes. Depending on the author and age of a library, it might use byte[]
, InputStream
, ByteBuffer
, or ReadableByteChannel
. If the bytes represent strings, there's also String
, Reader
, and CharSequence
to worry about. Remembering how to convert between all of them is a thankless task, made that much worse by libraries which define their own custom representations, or composing them with Clojure's lazy sequences and stream representations.
This library is a Rosetta stone for all the byte representations Java has to offer, and gives you the freedom to forget all the APIs you never wanted to know in the first place. Complete documentation can be found here.
[byte-streams "0.2.5"]
To convert one byte representation to another, use byte-streams/convert
:
byte-streams> (convert "abcd" java.nio.ByteBuffer)
#<HeapByteBuffer java.nio.HeapByteBuffer[pos=0 lim=4 cap=4]>
byte-streams> (convert *1 String)
"abcd"
(convert data to-type options?)
converts, if possible, the data from its current type to the destination type. This destination type can either be a Java class or a Clojure protocol. However, since there's no direct route from a string to a byte-buffer, under the covers convert
is doing whatever it takes to get the desired type:
byte-streams> (conversion-path String java.nio.ByteBuffer)
([java.lang.String [B]
[[B java.nio.ByteBuffer])
While we can't turn a string into a ByteBuffer
, we can turn a string into a byte[]
, and byte[]
into a ByteBuffer
. When invoked, convert
will choose the minimal path along the graph of available conversions. Common conversions are exposed via to-byte-buffer
, to-byte-buffers
, to-byte-array
, to-input-stream
, to-readable-channel
, to-char-sequence
, to-string
, and to-line-seq
.
Every type can exist either by itself, or as a sequence. For instance, we can create an InputStream
representing an infinite number of repeated strings:
byte-stream> (to-input-stream (repeat "hello"))
#<InputStream byte_streams.InputStream@3962a02c>
And then we can turn that into a lazy sequence of ByteBuffers
:
byte-streams> (take 2
(convert *1
(seq-of java.nio.ByteBuffer)
{:chunk-size 128}))
(#<HeapByteBuffer java.nio.HeapByteBuffer[pos=0 lim=128 cap=128]>
#<HeapByteBuffer java.nio.HeapByteBuffer[pos=0 lim=128 cap=128]>)
Notice that we describe a sequence of a type as (seq-of type)
, and that we've passed a map to convert
describing the size of the ByteBuffers
we want to create. Available options include:
:chunk-size
- the size in bytes of each chunk when converting a stream into a lazy seq of discrete chunks, defaults to4096
:direct?
- whether anyByteBuffers
which are created should be direct, defaults tofalse
:encoding
- the character set for any strings we're encoding or decoding, defaults to"UTF-8"
To create a Manifold stream, use (stream-of type)
. To convert a core.async channel, convert it using manifold.stream/->source
.
While there are conversions defined for all common byte types, this can be extended to other libraries via byte-streams/def-conversion
:
;; a conversion from byte-buffers to my-byte-representation
(def-conversion [ByteBuffer MyByteRepresentation]
[buf options]
(buffer->my-representation buf options))
;; everything that can be converted to a ByteBuffer is transitively fair game now
(convert "abc" MyByteRepresentation)
This mechanism can even be used for types unrelated to byte streams, if you're feeling adventurous.
Simple conversions are useful, but sometimes we'll need to do more than just keep the bytes in memory. When you need to write bytes to a file, network socket, or other endpoints, you can use byte-streams/transfer
.
byte-streams> (def f (File. "/tmp/salutations"))
#'byte-streams/f
byte-streams> (transfer "hello" f {:append? false})
nil
byte-streams> (to-string f)
"hello"
(transfer source sink options?)
allows you pipe anything that can produce bytes into anything that can receive bytes, using the most efficient mechanism available. Custom transfer mechanisms can also be defined:
(def-transfer [InputStream MyByteSink]
[stream sink options]
(send-stream-to-my-sink stream sink))
byte-streams/print-bytes
will print both hexadecimal and ascii representations of a collection of bytes:
byte-streams> (print-bytes (-> #'print-bytes meta :doc))
50 72 69 6E 74 73 20 6F 75 74 20 74 68 65 20 62 Prints out the b
79 74 65 73 20 69 6E 20 62 6F 74 68 20 68 65 78 ytes in both hex
20 61 6E 64 20 41 53 43 49 49 20 72 65 70 72 65 and ASCII repre
73 65 6E 74 61 74 69 6F 6E 73 2C 20 31 36 20 62 sentations, 16 b
79 74 65 73 20 70 65 72 20 6C 69 6E 65 2E ytes per line.
(byte-streams/compare-bytes a b)
will return a value which is positive if a
is lexicographically first, zero if they're equal, and negative otherwise:
byte-streams> (compare-bytes "abc" "abd")
-1
bytes-streams/bytes=
will return true if two byte streams are equal, and false otherwise.
byte-streams/conversion-path
returns all the intermediate steps in transforming one type to another, if one exists:
;; each element is a conversion tuple of to/from
byte-streams> (conversion-path java.io.File String)
([java.io.File java.nio.channels.ReadableByteChannel]
[#'byte-streams/ByteSource java.lang.CharSequence]
[java.lang.CharSequence java.lang.String])
;; but if a conversion is impossible...
byte-streams> (conversion-path java.io.OutputStream java.io.InputStream)
nil
byte-streams/possible-conversions
returns a list of possible conversion targets for a type.
byte-streams> (possible-conversions String)
(java.lang.String java.io.InputStream java.nio.DirectByteBuffer java.nio.ByteBuffer (seq-of java.nio.ByteBuffer) java.io.Reader java.nio.channels.ReadableByteChannel [B java.lang.CharSequence)
byte-streams/optimized-transfer?
returns true if there is an optimized transfer method for two types:
byte-streams> (optimized-transfer? String java.io.File)
true
Copyright © 2014 Zachary Tellman
Distributed under the MIT License