shithub: mc

ref: bc72f16e90932998ac4a404af25ce72efeada9a3
dir: /lib/bio/bio.myr/

View raw version
use std
use "types"

pkg bio =
	/* creation */
	const mk	: (mode : mode, vt : vtab	-> file#)
	const close	: (f : file# -> bool)
	const free	: (f : file# -> void)

	/* basic i/o. Returns sub-buffer when applicable. */
	const write	: (f : file#, src : byte[:]	-> std.result(std.size, err))
	const read	: (f : file#, dst : byte[:]	-> std.result(byte[:], err))
	const flush	: (f : file# -> bool)

	/* seeking */
	const seek	: (f : file#, off : std.off -> std.result(std.off, err))

	/* single unit operations */
	const putb	: (f : file#, b : byte	-> std.result(std.size, err))
	const putc	: (f : file#, c : char	-> std.result(std.size, err))
	const getb	: (f : file# -> std.result(byte, err))
	const getc	: (f : file# -> std.result(char, err))

	/* peeking */
	const peekb	: (f : file# -> std.result(byte, err))
	const peekc	: (f : file# -> std.result(char, err))

	/* delimited read; returns freshly allocated buffer. */
	const readln	: (f : file#	-> std.result(byte[:], err))
	const readto	: (f : file#, delim : byte[:]	-> std.result(byte[:], err))
	const skipto	: (f : file#, delim : byte[:]	-> bool)
	const skipspace	: (f : file# -> bool)

	/* formatted i/o */
	const put	: (f : file#, fmt : byte[:], args : ... -> std.result(std.size, err))
	const putv	: (f : file#, fmt : byte[:], ap : std.valist# -> std.result(std.size, err))

	/* pkg funcs */
	pkglocal const ensureread	: (f : file#, n : std.size -> std.result(std.size, err))
	pkglocal const ensurewrite	: (f : file#, n : std.size -> std.result(std.size, err))
;;

const Bufsz = 16*std.KiB
const Small = 512

/* Creates a file from an fd, opened in the given mode. */
const mk = {mode, vt
	var f

	f = std.alloc()

	f.mode = mode
	f.read = std.fndup(vt.read)
	f.write = std.fndup(vt.write)
	f.seek = std.fndup(vt.seek)
	f.close = std.fndup(vt.close)

	f.lasterr = 0
	if mode & Rd != 0
		f.rbuf = std.slalloc(Bufsz)
		f.rstart = 0
		f.rend = 0
	;;
	if mode & Wr != 0
		f.wbuf = std.slalloc(Bufsz)
		f.wend = 0
	;;
	-> f
}

/* closes a file, flushing it to the output fd */
const close = {f
	var ok

	ok = flush(f)
	ok = ok && f.close()
	_free(f)
	-> ok
}

const free = {f
	flush(f)
	_free(f)
}

const _free = {f
	if f.mode & Rd != 0
		std.slfree(f.rbuf)
	;;

	if f.mode & Wr != 0
		std.slfree(f.wbuf)
	;;
	std.fnfree(f.read)
	std.fnfree(f.write)
	std.fnfree(f.seek)
	std.fnfree(f.close)
	std.free(f)
}

/* 
writes to as much from `src` as possible to a file,
returning the number of bytes written.
*/
const write = {f, src
	std.assert(f.mode & Wr != 0, "File is not in write mode")
	/*
	Tack small writes onto the buffer end. Big ones
	flush the buffer and then go right to kernel.
	*/
	if src.len <= (f.wbuf.len - f.wend)
		std.slcp(f.wbuf[f.wend:f.wend+src.len], src)
		f.wend += src.len
		-> `std.Ok src.len
	else
		flush(f)
		-> writebuf(f, src)
	;;
}

/*
reads as much into 'dst' as possible, up to the size of 'dst',
returning the number of bytes read.
*/
const read = {f, dst
	var count, cap : std.size
	var d : byte[:]

	/* Clear the error state so we can retry */
	if f.lasterr != 0
		-> `std.Err geterr(f)
	;;

	/*
	a zero byte read always succeeds, reading 0 bytes; since
	there are an infinite number of zero byte reads you can do
	from anywhere in the file, including the end, this is not
	an EOF condition.
	*/
	if dst.len == 0
		-> `std.Ok dst
	;;
	std.assert(f.mode & Rd != 0, "File is not in read mode")
	/* 
	small reads should try to fill, so we don't have to make a
	syscall for every read
	*/
	cap = f.rend - f.rstart
	if dst.len < Small && cap < dst.len
		fill(f, dst.len)
	;;
	/* Read as much as we can from the buffer */
	count = std.min(dst.len, f.rend - f.rstart)
	std.slcp(dst[:count], f.rbuf[f.rstart:f.rstart+count])
	f.rstart += count

	/* if we drained the buffer, reset it */
	if f.rstart == f.rend
		f.rstart = 0
		f.rend = 0
	;;

	/* Read the rest directly from the fd */
	d = dst[count:]
	while d.len > 0
		match f.read(d)
		| `std.Ok 0:
			break
		| `std.Ok n:
			count += n
			d = d[n:]
		| `std.Err err:
			if count == 0
				-> `std.Err errtype(err)
			else
				f.lasterr = err
			;;
			break
		;;
	;;
	if count == 0
		-> `std.Err `Eof
	else
		-> `std.Ok dst[:count]
	;;
}

/* flushes f out to the backing fd */
const flush = {f
	var ret

	ret = true
	if f.mode & Wr != 0
		match writebuf(f, f.wbuf[:f.wend])
		| `std.Ok n: ret = (n == f.wend)
		| _:	ret = false
		;;
	;;
	f.wend = 0
	-> ret
}

const seek = {f, off
	flush(f)
	f.rstart = f.rend = 0
	match f.seek(off)
	| `std.Ok ret:	-> `std.Ok ret
	| `std.Err e:	-> `std.Err errtype(e)
	;;
}

/* writes a single byte to the output stream */
const putb = {f, b
	match ensurewrite(f, 1)
	| `std.Err e:	-> `std.Err e
	| `std.Ok n:
		f.wbuf[f.wend++] = b
		-> `std.Ok 1
	;;
}

/* writes a single character to the output stream, encoded in utf8 */
const putc = {f, c
	var sz
	
	sz = std.charlen(c)
	match ensurewrite(f, sz)
	| `std.Err e:	-> `std.Err e
	| `std.Ok n:
		std.encode(f.wbuf[f.wend:], c)
		f.wend += sz
		-> `std.Ok sz
	;;
}

/* reads a single byte from the input stream */
const getb = {f
	match ensureread(f, 1)
	| `std.Err e:	-> `std.Err e
	| `std.Ok n:
		-> `std.Ok f.rbuf[f.rstart++]
	;;
}

/* reads a single character from the input stream, encoded in utf8 */
const getc = {f
	var c

	match ensurecodepoint(f)
	| `std.Err e:	-> `std.Err e
	| `std.Ok n:
		c = std.decode(f.rbuf[f.rstart:f.rend])
		f.rstart += std.charlen(c)
		-> `std.Ok c
	;;
}

/* ensures we have enough to read a single codepoint in the buffer */
const ensurecodepoint = {f
	var b
	var len

	match ensureread(f, 1)
	| `std.Err e:	-> `std.Err e
	| `std.Ok n:
		b = f.rbuf[f.rstart]
		if b & 0x80 == 0	/* 0b0xxx_xxxx */
			len = 1
		elif b & 0xe0 == 0xc0	/* 0b110x_xxxx */
			len = 2
		elif b & 0xf0 == 0xe0 	/* 0b1110_xxxx */
			len = 3
		elif b & 0xf8 == 0xf0 	/* 0b1111_0xxx */
			len = 4
		else
			len = 1		/* invalid unicode char */
		;;
		-> ensureread(f, len)
	;;
}

/*
  writes a single integer-like value to the output stream, in
  little endian format
*/
generic putle = {f, v : @a::(numeric,integral)
	for var i = 0; i < sizeof(@a); i++
		putb(f, (v & 0xff : byte))
		v >>= 8
	;;
	-> sizeof(@a)
}

/*
  writes a single integer-like value to the output stream, in
  big endian format
*/
generic putbe = {f, v : @a::(numeric,integral)
	for var i = sizeof(@a); i != 0; i--
		putb(f, ((v >> ((i-1)*8)) & 0xff : byte))
	;;
	-> sizeof(@a)
}


/* peeks a single byte from an input stream */
const peekb = {f
	match ensureread(f, 1)
	| `std.Err e:	-> `std.Err e
	| `std.Ok n:
		-> `std.Ok f.rbuf[f.rstart]
	;;
}

/* peeks a single character from a utf8 encoded input stream */
const peekc = {f
	match ensurecodepoint(f)
	| `std.Err e:	-> `std.Err e
	| `std.Ok n:
		-> `std.Ok std.decode(f.rbuf[f.rstart:f.rend])
	;;
}

/*
  reads up to a single character delimiter. drops the delimiter
  from the input stream. EOF always counts as a delimiter.
  
  Eg, with the input "foo,bar\n"

  	bio.readto(f, ',')	-> "foo"
  	bio.readto(f, ',')	-> "bar\n"
*/
const readto = {f, delim
	-> readdelim(f, delim, false)
}

/* same as readto, but drops the read data. */
const skipto = {f, delim
	match readdelim(f, delim, true)
	| `std.Ok ign:	-> true
	| `std.Err _:		-> false
	;;
}

const skipspace = {f
	while true
		match bio.peekc(f)
		| `std.Ok c:
			if !std.isspace(c)
				break
			;;
			bio.getc(f)
		| `std.Err e:		-> false
		;;
	;;
	-> true
}

/* Same as delim, but with special handling for '\n', '\r', and '\r\n' */
const readln = {f
	var ret, c

	ret = [][:]
	while true
		/* get at least delimiter count of characters */
		match ensureread(f, 1)
		| `std.Err `Eof:
			ret = readinto(f, ret, f.rend - f.rstart)
			if ret.len > 0
				-> `std.Ok ret
			else
				-> `std.Err `Eof
			;;
		| `std.Err e:	-> `std.Err e
		| `std.Ok _:
		;;
		/* scan for delimiter */
		for var i = f.rstart; i < f.rend; i++
			c = (f.rbuf[i] : char)
			if c == '\r' || c == '\n'
				ret = readinto(f, ret, i - f.rstart)
				f.rstart++
				/* if we have '\r', we can get '\r\n'. */
				if c == '\r' && unwrapc(peekc(f), -1) == '\n'
					f.rstart++
				;;
				-> `std.Ok ret
			;;
:nextitergetln
		;;
		ret = readinto(f, ret, f.rend - f.rstart)
	;;
	std.die("unreachable")
}

const unwrapc = {cc, v
	match cc
	| `std.Ok c:	-> c
	| _:	-> v
	;;
}

const readdelim = {f, delim, drop
	var ret

	ret = [][:]
	while true
		/* get at least delimiter count of characters */
		match ensureread(f, 1)
		| `std.Err `Eof:
			if !drop
				ret = readinto(f, ret, f.rend - f.rstart)
			;;
			if ret.len > 0
				-> `std.Ok ret
			else
				-> `std.Err `Eof
			;;
		| `std.Err e:	-> `std.Err e
		| `std.Ok _:
		;;
		for var i = f.rstart; i < f.rend; i++
			if f.rbuf[i] == delim[0]
				for var j = 0; j < delim.len; j++
					if f.rbuf[i + j] != delim[j]
						goto nextiterread
					;;
				;;
				if !drop
					ret = readinto(f, ret, i - f.rstart)
				;;
				f.rstart += delim.len
				-> `std.Ok ret
			;;
:nextiterread
		;;
		if !drop
			ret = readinto(f, ret, f.rend - f.rstart)
		;;
	;;
	std.die("unreachable")
}

/*
Same as std.put, but buffered. Returns the number of bytes written.

FIXME: depends on std.fmt() having a flush buffer API. Until then,
we're stuck with a small static buffer.
*/
const put = {f, fmt, args
	var sl, ap, n

	ap = std.vastart(&args)
	sl = std.fmtv(fmt, &ap)
	n = write(f, sl)
	std.slfree(sl)
	-> n
}

const putv = {f, fmt, ap
	var sl, n

	sl = std.fmtv(fmt, ap)
	n = write(f, sl)
	std.slfree(sl)
	-> n
}

/* 
reads n bytes from the read buffer onto the heap-allocated slice
provided.
*/
const readinto = {f, buf, n
	var ret

	std.assert(f.rstart + n <= f.rend, "Reading too much from buffer")
	ret = std.sljoin(&buf, f.rbuf[f.rstart:f.rstart + n])
	f.rstart += n
	-> ret
}

/* makes sure we can bufferedly write at least n bytes */
const ensurewrite = {f, n
	std.assert(n < f.wbuf.len, "ensured write capacity > buffer size")
	if n > f.wbuf.len - f.wend
		match writebuf(f, f.wbuf[:f.wend])
		| `std.Ok len:
			f.wend = 0
			-> `std.Ok len
		| `std.Err e: -> `std.Err e
		;;
	;;
	-> `std.Ok n
}

/*
makes sure we have at least n bytes buffered. returns true if we succeed
in buffering n bytes, false if we fail.
*/
const ensureread = {f, n
	var held

	std.assert(n < f.rbuf.len, "ensured read capacity > buffer size")
	held = f.rend - f.rstart
	if n > held
		match fill(f, n)
		| `std.Err e:	-> `std.Err e
		| `std.Ok len:
			if len >= n
				-> `std.Ok len
			else
				-> `std.Err `Eof
			;;
		;;
	else
		-> `std.Ok n
	;;
}

/* blats a buffer to an fd */
const writebuf = {f, src
	var count

	count = 0
	while src.len != 0
		match f.write(src)
		| `std.Ok 0:
			-> `std.Err `Eof
		| `std.Ok n:
			count += n
			src = src[n:]
		| `std.Err e:
			-> `std.Err errtype(e)
		;;
	;;
:writedone
	-> `std.Ok count
}



/*
Reads as many bytes as possible from the file into
the read buffer.
*/
const fill = {f, min
	var count, cap

	count = 0
	/* Clear the error state so we can retry */
	if f.lasterr != 0
		-> `std.Err geterr(f)
	;;

	/* if we need to shift the slice down to the start, do it */
	cap = f.rend - f.rstart
	if min > cap
		std.slcp(f.rbuf[:cap], f.rbuf[f.rstart:f.rend])
		f.rstart = 0
		f.rend = cap
	;;
	while count < min
		/*
		If we've already read data, we don't want to
		throw it away, so we report a successful short
		read, and then error on the next read.
		*/
		match f.read(f.rbuf[f.rend:])
		| `std.Ok 0:
			break
		| `std.Ok n:
			count += n
			f.rend += n
		| `std.Err e:
			if count > 0
				f.lasterr = e
			else
				-> `std.Err errtype(e)
			;;
			break
		;;
	;;

	if count == 0
		-> `std.Err `Eof
	else
		-> `std.Ok count
	;;
}

const geterr = {f
	var e

	e = f.lasterr
	f.lasterr = 0
	-> errtype(e)
}

const errtype = {e : std.errno -> err
	var errno

	errno = (e : std.errno)
	if errno == std.Ebadf
		-> `Ebadf
	elif errno == std.Einval
		-> `Ebadf
	elif errno == std.Efault
		-> `Eio
	else
		-> `Eio
	;;
}