Move ringbuf into 3rdparty

3rdparty/ringbuf/ringbuf.c

@@ -0,0 +1,394 @@
+/*
+ * ringbuf.c - C ring buffer (FIFO) implementation.
+ *
+ * Written in 2011 by Drew Hess <dhess-src@bothan.net>.
+ *
+ * To the extent possible under law, the author(s) have dedicated all
+ * copyright and related and neighboring rights to this software to
+ * the public domain worldwide. This software is distributed without
+ * any warranty.
+ *
+ * You should have received a copy of the CC0 Public Domain Dedication
+ * along with this software. If not, see
+ * <http://creativecommons.org/publicdomain/zero/1.0/>.
+ */
+
+#ifndef KITTY_DEBUG_BUILD
+#define NDEBUG 1
+#endif
+#include "ringbuf.h"
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <sys/param.h>
+#include <assert.h>
+
+static size_t
+size_t_min(size_t x, size_t y) {
+    return x > y ? y : x;
+}
+
+/*
+ * The code is written for clarity, not cleverness or performance, and
+ * contains many assert()s to enforce invariant assumptions and catch
+ * bugs. Feel free to optimize the code and to remove asserts for use
+ * in your own projects, once you're comfortable that it functions as
+ * intended.
+ */
+
+struct ringbuf_t
+{
+    uint8_t *buf;
+    uint8_t *head, *tail;
+    size_t size;
+};
+
+ringbuf_t
+ringbuf_new(size_t capacity)
+{
+    ringbuf_t rb = malloc(sizeof(struct ringbuf_t));
+    if (rb) {
+
+        /* One byte is used for detecting the full condition. */
+        rb->size = capacity + 1;
+        rb->buf = malloc(rb->size);
+        if (rb->buf)
+            ringbuf_reset(rb);
+        else {
+            free(rb);
+            return 0;
+        }
+    }
+    return rb;
+}
+
+size_t
+ringbuf_buffer_size(const struct ringbuf_t *rb)
+{
+    return rb->size;
+}
+
+void
+ringbuf_reset(ringbuf_t rb)
+{
+    rb->head = rb->tail = rb->buf;
+}
+
+void
+ringbuf_free(ringbuf_t *rb)
+{
+    assert(rb && *rb);
+    free((*rb)->buf);
+    free(*rb);
+    *rb = 0;
+}
+
+size_t
+ringbuf_capacity(const struct ringbuf_t *rb)
+{
+    return ringbuf_buffer_size(rb) - 1;
+}
+
+/*
+ * Return a pointer to one-past-the-end of the ring buffer's
+ * contiguous buffer. You shouldn't normally need to use this function
+ * unless you're writing a new ringbuf_* function.
+ */
+static const uint8_t *
+ringbuf_end(const struct ringbuf_t *rb)
+{
+    return rb->buf + ringbuf_buffer_size(rb);
+}
+
+size_t
+ringbuf_bytes_free(const struct ringbuf_t *rb)
+{
+    if (rb->head >= rb->tail)
+        return ringbuf_capacity(rb) - (rb->head - rb->tail);
+    else
+        return rb->tail - rb->head - 1;
+}
+
+size_t
+ringbuf_bytes_used(const struct ringbuf_t *rb)
+{
+    return ringbuf_capacity(rb) - ringbuf_bytes_free(rb);
+}
+
+int
+ringbuf_is_full(const struct ringbuf_t *rb)
+{
+    return ringbuf_bytes_free(rb) == 0;
+}
+
+int
+ringbuf_is_empty(const struct ringbuf_t *rb)
+{
+    return ringbuf_bytes_free(rb) == ringbuf_capacity(rb);
+}
+
+const void *
+ringbuf_tail(const struct ringbuf_t *rb)
+{
+    return rb->tail;
+}
+
+const void *
+ringbuf_head(const struct ringbuf_t *rb)
+{
+    return rb->head;
+}
+
+/*
+ * Given a ring buffer rb and a pointer to a location within its
+ * contiguous buffer, return the a pointer to the next logical
+ * location in the ring buffer.
+ */
+static uint8_t *
+ringbuf_nextp(ringbuf_t rb, const uint8_t *p)
+{
+    /*
+     * The assert guarantees the expression (++p - rb->buf) is
+     * non-negative; therefore, the modulus operation is safe and
+     * portable.
+     */
+    assert((p >= rb->buf) && (p < ringbuf_end(rb)));
+    return rb->buf + ((++p - rb->buf) % ringbuf_buffer_size(rb));
+}
+
+size_t
+ringbuf_findchr(const struct ringbuf_t *rb, int c, size_t offset)
+{
+    const uint8_t *bufend = ringbuf_end(rb);
+    size_t bytes_used = ringbuf_bytes_used(rb);
+    if (offset >= bytes_used)
+        return bytes_used;
+
+    const uint8_t *start = rb->buf +
+        (((rb->tail - rb->buf) + offset) % ringbuf_buffer_size(rb));
+    assert(bufend > start);
+    size_t n = size_t_min(bufend - start, bytes_used - offset);
+    const uint8_t *found = memchr(start, c, n);
+    if (found)
+        return offset + (found - start);
+    else
+        return ringbuf_findchr(rb, c, offset + n);
+}
+
+size_t
+ringbuf_memset(ringbuf_t dst, int c, size_t len)
+{
+    const uint8_t *bufend = ringbuf_end(dst);
+    size_t nwritten = 0;
+    size_t count = size_t_min(len, ringbuf_buffer_size(dst));
+    int overflow = count > ringbuf_bytes_free(dst);
+
+    while (nwritten != count) {
+
+        /* don't copy beyond the end of the buffer */
+        assert(bufend > dst->head);
+        size_t n = size_t_min(bufend - dst->head, count - nwritten);
+        memset(dst->head, c, n);
+        dst->head += n;
+        nwritten += n;
+
+        /* wrap? */
+        if (dst->head == bufend)
+            dst->head = dst->buf;
+    }
+
+    if (overflow) {
+        dst->tail = ringbuf_nextp(dst, dst->head);
+        assert(ringbuf_is_full(dst));
+    }
+
+    return nwritten;
+}
+
+void *
+ringbuf_memcpy_into(ringbuf_t dst, const void *src, size_t count)
+{
+    const uint8_t *u8src = src;
+    const uint8_t *bufend = ringbuf_end(dst);
+    int overflow = count > ringbuf_bytes_free(dst);
+    size_t nread = 0;
+
+    while (nread != count) {
+        /* don't copy beyond the end of the buffer */
+        assert(bufend > dst->head);
+        size_t n = size_t_min(bufend - dst->head, count - nread);
+        memcpy(dst->head, u8src + nread, n);
+        dst->head += n;
+        nread += n;
+
+        /* wrap? */
+        if (dst->head == bufend)
+            dst->head = dst->buf;
+    }
+
+    if (overflow) {
+        dst->tail = ringbuf_nextp(dst, dst->head);
+        assert(ringbuf_is_full(dst));
+    }
+
+    return dst->head;
+}
+
+ssize_t
+ringbuf_read(int fd, ringbuf_t rb, size_t count)
+{
+    const uint8_t *bufend = ringbuf_end(rb);
+    size_t nfree = ringbuf_bytes_free(rb);
+
+    /* don't write beyond the end of the buffer */
+    assert(bufend > rb->head);
+    count = size_t_min(bufend - rb->head, count);
+    ssize_t n = read(fd, rb->head, count);
+    if (n > 0) {
+        assert(rb->head + n <= bufend);
+        rb->head += n;
+
+        /* wrap? */
+        if (rb->head == bufend)
+            rb->head = rb->buf;
+
+        /* fix up the tail pointer if an overflow occurred */
+        if ((size_t)n > nfree) {
+            rb->tail = ringbuf_nextp(rb, rb->head);
+            assert(ringbuf_is_full(rb));
+        }
+    }
+
+    return n;
+}
+
+void *
+ringbuf_memmove_from(void *dst, ringbuf_t src, size_t count)
+{
+    size_t bytes_used = ringbuf_bytes_used(src);
+    if (count > bytes_used)
+        return 0;
+
+    uint8_t *u8dst = dst;
+    const uint8_t *bufend = ringbuf_end(src);
+    size_t nwritten = 0;
+    while (nwritten != count) {
+        assert(bufend > src->tail);
+        size_t n = size_t_min(bufend - src->tail, count - nwritten);
+        memcpy(u8dst + nwritten, src->tail, n);
+        src->tail += n;
+        nwritten += n;
+
+        /* wrap ? */
+        if (src->tail == bufend)
+            src->tail = src->buf;
+    }
+
+    assert(count + ringbuf_bytes_used(src) == bytes_used);
+    return src->tail;
+}
+
+unsigned char
+ringbuf_move_char(ringbuf_t src) {
+    assert(!ringbuf_is_empty(src));
+    const uint8_t *bufend = ringbuf_end(src);
+    assert(bufend > src->tail);
+    uint8_t ans = *src->tail;
+    src->tail += 1;
+    if (src->tail == bufend)
+        src->tail = src->buf;
+    return ans;
+}
+
+size_t
+ringbuf_memcpy_from(void *dst, const ringbuf_t src, size_t count)
+{
+    size_t bytes_used = ringbuf_bytes_used(src);
+    if (count > bytes_used) count = bytes_used;
+
+    uint8_t *u8dst = dst;
+    const uint8_t *bufend = ringbuf_end(src);
+    size_t nwritten = 0;
+    const uint8_t* tail = src->tail;
+    while (nwritten != count) {
+        assert(bufend > tail);
+        size_t n = size_t_min(bufend - tail, count - nwritten);
+        memcpy(u8dst + nwritten, tail, n);
+        tail += n;
+        nwritten += n;
+
+        /* wrap ? */
+        if (tail == bufend)
+            tail = src->buf;
+    }
+
+    assert(ringbuf_bytes_used(src) == bytes_used);
+    return count;
+}
+
+
+ssize_t
+ringbuf_write(int fd, ringbuf_t rb, size_t count)
+{
+    size_t bytes_used = ringbuf_bytes_used(rb);
+    if (count > bytes_used)
+        return 0;
+
+    const uint8_t *bufend = ringbuf_end(rb);
+    assert(bufend > rb->head);
+    count = size_t_min(bufend - rb->tail, count);
+    ssize_t n = write(fd, rb->tail, count);
+    if (n > 0) {
+        assert(rb->tail + n <= bufend);
+        rb->tail += n;
+
+        /* wrap? */
+        if (rb->tail == bufend)
+            rb->tail = rb->buf;
+
+        assert(n + ringbuf_bytes_used(rb) == bytes_used);
+    }
+
+    return n;
+}
+
+void *
+ringbuf_copy(ringbuf_t dst, ringbuf_t src, size_t count)
+{
+    size_t src_bytes_used = ringbuf_bytes_used(src);
+    if (count > src_bytes_used)
+        return 0;
+    int overflow = count > ringbuf_bytes_free(dst);
+
+    const uint8_t *src_bufend = ringbuf_end(src);
+    const uint8_t *dst_bufend = ringbuf_end(dst);
+    size_t ncopied = 0;
+    while (ncopied != count) {
+        assert(src_bufend > src->tail);
+        size_t nsrc = size_t_min(src_bufend - src->tail, count - ncopied);
+        assert(dst_bufend > dst->head);
+        size_t n = size_t_min(dst_bufend - dst->head, nsrc);
+        memcpy(dst->head, src->tail, n);
+        src->tail += n;
+        dst->head += n;
+        ncopied += n;
+
+        /* wrap ? */
+        if (src->tail == src_bufend)
+            src->tail = src->buf;
+        if (dst->head == dst_bufend)
+            dst->head = dst->buf;
+    }
+
+    assert(count + ringbuf_bytes_used(src) == src_bytes_used);
+
+    if (overflow) {
+        dst->tail = ringbuf_nextp(dst, dst->head);
+        assert(ringbuf_is_full(dst));
+    }
+
+    return dst->head;
+}

3rdparty/ringbuf/ringbuf.h

@@ -0,0 +1,252 @@
+#pragma once
+/*
+ * ringbuf.h - C ring buffer (FIFO) interface.
+ *
+ * Written in 2011 by Drew Hess <dhess-src@bothan.net>.
+ *
+ * To the extent possible under law, the author(s) have dedicated all
+ * copyright and related and neighboring rights to this software to
+ * the public domain worldwide. This software is distributed without
+ * any warranty.
+ *
+ * You should have received a copy of the CC0 Public Domain Dedication
+ * along with this software. If not, see
+ * <http://creativecommons.org/publicdomain/zero/1.0/>.
+ */
+
+/*
+ * A byte-addressable ring buffer FIFO implementation.
+ *
+ * The ring buffer's head pointer points to the starting location
+ * where data should be written when copying data *into* the buffer
+ * (e.g., with ringbuf_read). The ring buffer's tail pointer points to
+ * the starting location where data should be read when copying data
+ * *from* the buffer (e.g., with ringbuf_write).
+ */
+
+#include <stddef.h>
+#include <sys/types.h>
+
+typedef struct ringbuf_t *ringbuf_t;
+
+/*
+ * Create a new ring buffer with the given capacity (usable
+ * bytes). Note that the actual internal buffer size may be one or
+ * more bytes larger than the usable capacity, for bookkeeping.
+ *
+ * Returns the new ring buffer object, or 0 if there's not enough
+ * memory to fulfill the request for the given capacity.
+ */
+ringbuf_t
+ringbuf_new(size_t capacity);
+
+/*
+ * The size of the internal buffer, in bytes. One or more bytes may be
+ * unusable in order to distinguish the "buffer full" state from the
+ * "buffer empty" state.
+ *
+ * For the usable capacity of the ring buffer, use the
+ * ringbuf_capacity function.
+ */
+size_t
+ringbuf_buffer_size(const struct ringbuf_t *rb);
+
+/*
+ * Deallocate a ring buffer, and, as a side effect, set the pointer to
+ * 0.
+ */
+void
+ringbuf_free(ringbuf_t *rb);
+
+/*
+ * Reset a ring buffer to its initial state (empty).
+ */
+void
+ringbuf_reset(ringbuf_t rb);
+
+/*
+ * The usable capacity of the ring buffer, in bytes. Note that this
+ * value may be less than the ring buffer's internal buffer size, as
+ * returned by ringbuf_buffer_size.
+ */
+size_t
+ringbuf_capacity(const struct ringbuf_t *rb);
+
+/*
+ * The number of free/available bytes in the ring buffer. This value
+ * is never larger than the ring buffer's usable capacity.
+ */
+size_t
+ringbuf_bytes_free(const struct ringbuf_t *rb);
+
+/*
+ * The number of bytes currently being used in the ring buffer. This
+ * value is never larger than the ring buffer's usable capacity.
+ */
+size_t
+ringbuf_bytes_used(const struct ringbuf_t *rb);
+
+int
+ringbuf_is_full(const struct ringbuf_t *rb);
+
+int
+ringbuf_is_empty(const struct ringbuf_t *rb);
+
+/*
+ * Const access to the head and tail pointers of the ring buffer.
+ */
+const void *
+ringbuf_tail(const struct ringbuf_t *rb);
+
+const void *
+ringbuf_head(const struct ringbuf_t *rb);
+
+/*
+ * Locate the first occurrence of character c (converted to an
+ * unsigned char) in ring buffer rb, beginning the search at offset
+ * bytes from the ring buffer's tail pointer. The function returns the
+ * offset of the character from the ring buffer's tail pointer, if
+ * found. If c does not occur in the ring buffer, the function returns
+ * the number of bytes used in the ring buffer.
+ *
+ * Note that the offset parameter and the returned offset are logical
+ * offsets from the tail pointer, not necessarily linear offsets.
+ */
+size_t
+ringbuf_findchr(const struct ringbuf_t *rb, int c, size_t offset);
+
+/*
+ * Beginning at ring buffer dst's head pointer, fill the ring buffer
+ * with a repeating sequence of len bytes, each of value c (converted
+ * to an unsigned char). len can be as large as you like, but the
+ * function will never write more than ringbuf_buffer_size(dst) bytes
+ * in a single invocation, since that size will cause all bytes in the
+ * ring buffer to be written exactly once each.
+ *
+ * Note that if len is greater than the number of free bytes in dst,
+ * the ring buffer will overflow. When an overflow occurs, the state
+ * of the ring buffer is guaranteed to be consistent, including the
+ * head and tail pointers; old data will simply be overwritten in FIFO
+ * fashion, as needed. However, note that, if calling the function
+ * results in an overflow, the value of the ring buffer's tail pointer
+ * may be different than it was before the function was called.
+ *
+ * Returns the actual number of bytes written to dst: len, if
+ * len < ringbuf_buffer_size(dst), else ringbuf_buffer_size(dst).
+ */
+size_t
+ringbuf_memset(ringbuf_t dst, int c, size_t len);
+
+/*
+ * Copy n bytes from a contiguous memory area src into the ring buffer
+ * dst. Returns the ring buffer's new head pointer.
+ *
+ * It is possible to copy more data from src than is available in the
+ * buffer; i.e., it's possible to overflow the ring buffer using this
+ * function. When an overflow occurs, the state of the ring buffer is
+ * guaranteed to be consistent, including the head and tail pointers;
+ * old data will simply be overwritten in FIFO fashion, as
+ * needed. However, note that, if calling the function results in an
+ * overflow, the value of the ring buffer's tail pointer may be
+ * different than it was before the function was called.
+ */
+void *
+ringbuf_memcpy_into(ringbuf_t dst, const void *src, size_t count);
+
+/*
+ * This convenience function calls read(2) on the file descriptor fd,
+ * using the ring buffer rb as the destination buffer for the read,
+ * and returns the value returned by read(2). It will only call
+ * read(2) once, and may return a short count.
+ *
+ * It is possible to read more data from the file descriptor than is
+ * available in the buffer; i.e., it's possible to overflow the ring
+ * buffer using this function. When an overflow occurs, the state of
+ * the ring buffer is guaranteed to be consistent, including the head
+ * and tail pointers: old data will simply be overwritten in FIFO
+ * fashion, as needed. However, note that, if calling the function
+ * results in an overflow, the value of the ring buffer's tail pointer
+ * may be different than it was before the function was called.
+ */
+ssize_t
+ringbuf_read(int fd, ringbuf_t rb, size_t count);
+
+/*
+ * Copy n bytes from the ring buffer src, starting from its tail
+ * pointer, into a contiguous memory area dst. Returns the value of
+ * src's tail pointer after the copy is finished.
+ *
+ * Note that this copy is destructive with respect to the ring buffer:
+ * the n bytes copied from the ring buffer are no longer available in
+ * the ring buffer after the copy is complete, and the ring buffer
+ * will have n more free bytes than it did before the function was
+ * called.
+ *
+ * This function will *not* allow the ring buffer to underflow. If
+ * count is greater than the number of bytes used in the ring buffer,
+ * no bytes are copied, and the function will return 0.
+ */
+void *
+ringbuf_memmove_from(void *dst, ringbuf_t src, size_t count);
+
+/* ringbuf_memmove_from() optimized for a single character.
+ * Must only be called if the ringbuf is not empty */
+unsigned char
+ringbuf_move_char(ringbuf_t src);
+
+/*
+ * Same as ringbuf_memmove_from() except that it does not change the ringbuffer
+ * and returns the actual number of bytes copied, which is the minimum of ringbuf_bytes_used
+ * and count.
+ */
+size_t
+ringbuf_memcpy_from(void *dst, const ringbuf_t src, size_t count);
+
+/*
+ * This convenience function calls write(2) on the file descriptor fd,
+ * using the ring buffer rb as the source buffer for writing (starting
+ * at the ring buffer's tail pointer), and returns the value returned
+ * by write(2). It will only call write(2) once, and may return a
+ * short count.
+ *
+ * Note that this copy is destructive with respect to the ring buffer:
+ * any bytes written from the ring buffer to the file descriptor are
+ * no longer available in the ring buffer after the copy is complete,
+ * and the ring buffer will have N more free bytes than it did before
+ * the function was called, where N is the value returned by the
+ * function (unless N is < 0, in which case an error occurred and no
+ * bytes were written).
+ *
+ * This function will *not* allow the ring buffer to underflow. If
+ * count is greater than the number of bytes used in the ring buffer,
+ * no bytes are written to the file descriptor, and the function will
+ * return 0.
+ */
+ssize_t
+ringbuf_write(int fd, ringbuf_t rb, size_t count);
+
+/*
+ * Copy count bytes from ring buffer src, starting from its tail
+ * pointer, into ring buffer dst. Returns dst's new head pointer after
+ * the copy is finished.
+ *
+ * Note that this copy is destructive with respect to the ring buffer
+ * src: any bytes copied from src into dst are no longer available in
+ * src after the copy is complete, and src will have 'count' more free
+ * bytes than it did before the function was called.
+ *
+ * It is possible to copy more data from src than is available in dst;
+ * i.e., it's possible to overflow dst using this function. When an
+ * overflow occurs, the state of dst is guaranteed to be consistent,
+ * including the head and tail pointers; old data will simply be
+ * overwritten in FIFO fashion, as needed. However, note that, if
+ * calling the function results in an overflow, the value dst's tail
+ * pointer may be different than it was before the function was
+ * called.
+ *
+ * It is *not* possible to underflow src; if count is greater than the
+ * number of bytes used in src, no bytes are copied, and the function
+ * returns 0.
+ */
+void *
+ringbuf_copy(ringbuf_t dst, ringbuf_t src, size_t count);