Merge branch 'main' into dj

docs/dj.1

@@ -4,7 +4,7 @@
 .\" This work is licensed under CC BY-SA 4.0. To see a copy of this license,
 .\" visit <http://creativecommons.org/licenses/by-sa/4.0/>.
 .\"
-.TH DJ 1 2024-06-29 "Harakit X.X.X"
+.TH DJ 1 2024-07-03 "Harakit X.X.X"
 .SH NAME
 dj \(en disk jockey
 .\"
@@ -34,8 +34,13 @@ respectively. This language is inherited from the
 .BR dd (1p)
 utility and used here to decrease ambiguity.
 
-When seeking or skipping to a byte, writing or reading starts at the byte
-immediately subsequent to the specified byte.
+The offset used when skipping or seeking refers to how many bytes are skipped
+or sought. Running
+.BR dj (1)
+with a skip offset of 1 skips one byte into the input and reads from the second
+byte onwards. A programmer may think of a file as a zero-indexed array of
+bytes; in this analogy, the offset given is the index of the byte at which to
+start reading or writing.
 .\"
 .SH OPTIONS
 
@@ -80,6 +85,54 @@ input file is \(lq-\(rq.
 The standard output shall be used as an output if no inputs are specified or if
 the output file is \(lq-\(rq.
 .\"
+.SH EXAMPLES
+
+The following
+.BR sh (1p)
+line:
+
+.RS
+printf 'Hello, world!\(rsn' | dj -c 1 -b 7 -s 7 2>/dev/null
+.RE
+
+Produces the following output:
+
+.RS
+world!
+.RE
+
+The following
+.BR sh (1p)
+lines run sequentially:
+
+.RS
+tr '\(rs0' 0    </dev/zero | dj -c 1 -b 6 -o hello.txt
+
+tr '\(rs0' H    </dev/zero | dj -c 1 -b 1 -o hello.txt
+
+tr '\(rs0' e    </dev/zero | dj -c 1 -b 1 -o hello.txt -S 1
+
+tr '\(rs0' l    </dev/zero | dj -c 1 -b 2 -o hello.txt -S 2
+
+tr '\(rs0' o    </dev/zero | dj -c 1 -b 1 -o hello.txt -S 4
+
+tr '\(rs0' '\(rsn' </dev/zero | dj -c 1 -b 1 -o hello.txt -S 5
+
+dj -i hello.txt
+.RE
+
+Produce the following output:
+
+.RS
+Hello
+.RE
+
+It may be particularly illuminating to print the contents of the example
+.B hello.txt
+after each
+.BR dj (1)
+invocation.
+.\"
 .SH DIAGNOSTICS
 
 On a partial or empty read, a diagnostic message is printed. Then, the program
@@ -128,9 +181,20 @@ option is specified, this could make written data nonsensical.
 
 Existing files are not truncated on ouput and are instead overwritten.
 
-Many lowercase options have capitalized variants and vice-versa which can be
-confusing. Capitalized options tend to affect output or are more intense
-versions of lowercase options.
+The options
+.B -b
+and
+.B -B
+could be confused for each other, and so could
+.B -s
+and
+.BR -S .
+The lowercase option affects input and the capitalized option affects output.
+
+The skipped or sought bytes while processing irregular files, such as streams,
+are reported in the diagnostic output, because they were actually read or
+written. This is as opposed to bytes skipped while processing regular files,
+which are not reported.
 .\"
 .SH RATIONALE
 
@@ -151,3 +215,4 @@ Copyright \(co 2023 DTB. License AGPLv3+: GNU AGPL version 3 or later
 .\"
 .SH SEE ALSO
 .BR dd (1p)
+.BR lseek (3p)

src/dj.c

@@ -16,12 +16,15 @@
  * along with this program. If not, see https://www.gnu.org/licenses/.
  */
 
+#include <assert.h> /* assert(3) */
 #include <errno.h> /* errno */
 #include <fcntl.h> /* open(2) */
 #include <stdio.h> /* fprintf(3), stderr */
-#include <stdlib.h> /* free(3), malloc(3), strtol(3), size_t */
+#include <stdlib.h> /* malloc(3), strtol(3), size_t */
 #include <string.h> /* memcpy(3), memmove(3), memset(3) */
-#include <sysexits.h> /* EX_OK, EX_USAGE */
+#if !defined EX_OK || !defined EX_OSERR || !defined EX_USAGE
+#	include <sysexits.h>
+#endif
 #include <unistd.h> /* close(2), getopt(3), lseek(2), read(2), write(2),
                      * optarg, optind, STDIN_FILENO, STDOUT_FILENO */
 #include <sys/stat.h> /* S_IRGRP, S_IROTH, S_IRUSR, S_IWGRP, S_IWOTH,
@@ -34,198 +37,90 @@ char *program_name = "dj";
  * writing ends of its jockeyed "pipe". User-configurable members are noted
  * with their relevant options. */
 struct Io{
-	int bs;			/* buffer size (-bB) */
+	char *buf;	  	/* buffer */
+	char *fn;	  	/* file name (-io) */
+	size_t bs;    	/* buffer size (-bB) */
 	size_t bufuse;	/* buffer usage */
-	char *buf;		/* buffer */
-	int bytes;		/* bytes processed */
-	int fd;			/* file descriptor */
-	int fl;			/* file opening flags */
-	char *fn;		/* file name (may be stdin_name or stdout_name) (-io) */
-	int prec;		/* partial records processed */
-	int rec;		/* records processed */
-	long seek;		/* bytes to seek/skip (will be 0 after skippage) (-sS) */
-}; 
+	size_t bytes; 	/* bytes processed */
+	size_t prec;  	/* partial records processed */
+	size_t rec;   	/* records processed */
+	long seek;  	/* remaining bytes to seek/skip (-sS) */
+	int error;  	/* errno */
+	int fd;		  	/* file descriptor */
+	int fl;		  	/* file opening flags */
+};
 
-/* To be assigned to main:fmt_output and used with output(). */
+/* To be assigned to main:fmt and used with printio(). */
 static char *fmt_asv   = "%d\037%d\036%d\037%d\035%d\036%d\034";
 static char *fmt_human = "%d+%d > %d+%d; %d > %d\n";
 
 static char *stdin_name = "<stdin>";
 static char *stdout_name = "<stdout>";
 
+static int creat_mode = S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH
+	| S_IWOTH; /* Consistent with touch(1p). */
 static int read_flags = O_RDONLY; /* Consistent with Busybox dd(1). */
 static int write_flags = O_WRONLY | O_CREAT;
 
 #define MIN(a, b) (((a) < (b)) ? (a) : (b))
 
-/* Macro to check if fd is a std* file, e.g. stdin. */
-#define fdisstd(fd) \
-		((fd) == STDIN_FILENO \
-		|| (fd) == STDOUT_FILENO \
-		|| (fd) == STDERR_FILENO)
+/* Macro to check if fd is stdin or stdout */
+#define fdisstd(fd) ((fd) == STDIN_FILENO || (fd) == STDOUT_FILENO)
 
-/* Macro to call the cleanup functions that operate on struct io on the
- * particular io[2] used in main. Error conditions are not checked because this
- * is only used when the program is about to terminate (hence its name). */
-#define terminate(io) do{ \
-	free((io[0]).buf); \
-	free((io[1]).buf); \
-	Io_fdclose(&(io)[0]); \
-	Io_fdclose(&(io)[1]); }while(0)
-
-/* Allocates *io's buffer. Returns NULL if unsuccessful. */
-static void *
-Io_bufalloc(struct Io *io){
-
-	return (io->buf = malloc(io->bs * (sizeof *io->buf)));
-}
-
-/* Fills the unused portion of io's buffer with padding, updating io->bufuse.
- * Returns io. */
-static struct Io *
-Io_bufrpad(struct Io *io, int padding){
-	
-	memset(io->buf + io->bufuse, padding, io->bs - io->bufuse);
-	io->bufuse = io->bs;
-
-	return io;
-}
-
-/* Copies from the buffer in src as much as possible to the free space in the
- * dest buffer, removing the copied units from src and permuting the remaining
- * units in the src buffer to the start of the buffer, modifying both the src
- * and dest bufuse and returning dest. */
-static struct Io*
-Io_bufxapp(struct Io *dest, struct Io *src){
-	int n;
-
-	n = MIN(src->bufuse, dest->bs - dest->bufuse);
-	memcpy(dest->buf + dest->bufuse, src->buf, n);
-	dest->bufuse += n;
-	memmove(src->buf, src->buf + n, src->bs - n);
-	src->bufuse -= n;
-
-	return dest;
-}
-
-/* Copies from the buffer in src to the buffer in dest no more than n units,
- * removing the copied units from src and permuting the remaining units in the
- * src buffer to the start of the buffer, modifying both the src and dest
- * bufuse and returning dest. */
-static struct Io*
-Io_bufxfer(struct Io *dest, struct Io *src, int n){
-
-	memcpy(dest->buf, src->buf, (dest->bufuse = n));
-	memmove(src->buf, src->buf + n, (src->bufuse -= n));
-
-	return dest;
-}
-
-/* Closes io->fn and returns -1 on error, otherwise io->fd. */
-static int
-Io_fdclose(struct Io *io){
-
-	return fdisstd(io->fd)
-		? 0
-		: close(io->fd);
-}
-
-/* Opens io->fn and saves the file descriptor into io->fd. Returns io->fd,
- * which will be -1 if an error occured. */
-static int
-Io_fdopen(struct Io *io, char *fn){
-	int fd;
-	
-	if((fd = open(fn, io->fl,
-					/* these are the flags used by touch(1p) */
-					S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH))
-				!= -1
-			&& Io_fdclose(io) == 0){
-		io->fd = fd;
-		io->fn = fn;
-	}
-
-	return fd;
-}
-
-/* Seeks io->seek bytes through *io's file descriptor, subtracting the number
- * of sought bytes from io->seek. This procedure leaves garbage in io->buf. */
-static void
-Io_fdseek(struct Io *io){
-	
-	if(io->seek != 0
-			|| (!fdisstd(io->fd) && lseek(io->fd, io->seek, SEEK_SET) != -1))
-		return;
-
-	if(io->fl == write_flags)
-		memset(io->buf, '\0', io->bs);
-
-	if(io->fl == write_flags){
-		memset(io->buf, '\0', io->bs);
-		/* We're going to cheat and use bufuse as the retval for write(2),
-		 * which is fine because it'll be zeroed as this function returns
-		 * anyway. */
-		do{
-			if((io->bufuse = write(io->fd, io->buf, MIN(io->bs, io->seek)))
-					== 0)
-				/* second chance */
-				io->bufuse = write(io->fd, io->buf, MIN(io->bs, io->seek));
-		}while((io->seek -= io->bufuse) > 0 && io->bufuse != 0);
-	}else if(io->fl == read_flags){
-		do{
-			if((io->bufuse = read(io->fd, io->buf, MIN(io->bs, io->seek)))
-					== 0)
-				/* second chance */
-				io->bufuse = read(io->fd, io->buf, MIN(io->bs, io->seek));
-		}while((io->seek -= io->bufuse) > 0 && io->bufuse != 0);
-	}
-
-	io->bufuse = 0;
-
-	return;
-}
-
-/* Reads io->bs bytes from *io's file descriptor into io->buf, storing the
- * number of read bytes in io->bufuse and updating io->bytes. If io->bufuse is
- * 0, errno will probably be set. Returns io. */
 static struct Io *
 Io_read(struct Io *io){
+	int t;
 
-	io->bytes += (io->bufuse = read(io->fd, io->buf, io->bs));
+	assert(io->bs > 0);
+	assert(io->bufuse < io->bs);
+
+	if((t = read(io->fd, &(io->buf)[io->bufuse], io->bs - io->bufuse)) < 0){
+		io->error = errno;
+		t = 0;
+	}
+
+	io->bufuse += t;
+	io->bytes  += t;
+	io->prec   += (0 < io->bufuse && io->bufuse < io->bs);
+	io->rec    += (io->bufuse == io->bs);
+
+	assert(io->bufuse <= io->bs);
 
 	return io;
 }
 
-/* Writes io->bufuse units from io->buf to io->fd, permuting any unwritten
- * bytes to the start of io->buf and updating io->bufuse. If io->bufuse doesn't
- * change, errno will probably be set. Returns io. */
 static struct Io *
 Io_write(struct Io *io){
 	int t;
 
-	if((t = write(io->fd, io->buf, io->bufuse)) > 0)
-		memmove(io->buf, io->buf + t, (io->bufuse -= t));
+	assert(io->bufuse > 0);
+	assert(io->bufuse <= io->bs);
+
+	if((t = write(io->fd, io->buf, io->bufuse)) < 0){
+		io->error = errno;
+		t = 0;
+	}else if(t > 0)
+		memmove(io->buf, &(io->buf)[t], (io->bufuse -= t));
+
 	io->bytes += t;
+	io->prec  += (t > 0 && io->bufuse > 0);
+	io->rec   += (t > 0 && io->bufuse == 0);
 
 	return io;
 }
 
-/* Prints an error message suitable for the event of an operating system error,
- * with the error itself to be described in the string s. */
 static int
-oserr(char *s){
-
-	fprintf(stderr, "%s: %s: %s\n", program_name, s, strerror(errno));
-
+oserr(char *e, int n){
+	fprintf(stderr, "%s: %s: %s\n", program_name, e, strerror(n));
 	return EX_OSERR;
 }
 
 /* Prints statistics regarding the use of dj, particularly partially and
  * completely read and written records. */
 static void
-output(struct Io io[2], char *fmt){
+fprintio(FILE *stream, char *fmt, struct Io io[2]){
 
-	fprintf(stderr, fmt,
+	fprintf(stream, fmt,
 		io[0].rec, io[0].prec, io[1].rec, io[1].prec,
 		io[0].bytes, io[1].bytes);
 
@@ -258,24 +153,26 @@ usage(char *s){
 }
 
 int main(int argc, char *argv[]){
-	int align; /* low 8b used, negative if no alignment is being done */
-	int count; /* 0 if dj(1) runs until no more reads are possible */
-	char *fmt_output; /* == fmt_asv (default) or fmt_human (-H) */
-	size_t i; /* side of io being modified */
-	struct Io io[2];
+ 	int align;    /* low 8b used, negative if no alignment is being done */
+	int count;    /* 0 if dj(1) runs until no more reads are possible */
+	char *fmt;    /* == fmt_asv (default) or fmt_human (-H) */
+	size_t i;     /* side of io being modified */
 	char noerror; /* 0=exits (default) 1=retries on partial reads or writes */
+	struct Io io[2 /* { in, out } */];
 
 	/* Set defaults. */
 	align = -1;
 	count = 0;
-	fmt_output = fmt_asv;
+	fmt = fmt_asv;
 	noerror = 0;
-	for(i = 0; i < 2; ++i){
+	for(i = 0; i < (sizeof io) / (sizeof *io); ++i){
 		io[i].bs = 1024 /* 1 KiB */; /* GNU dd(1) default; POSIX says 512B */
+		io[i].bufuse = 0;
 		io[i].bytes = 0;
-		io[i].fd = i ? STDIN_FILENO : STDOUT_FILENO;
-		io[i].fn = i ? stdin_name   : stdout_name;
-		io[i].fl = i ? read_flags   : write_flags;
+		io[i].fd = i == 0 ? STDIN_FILENO : STDOUT_FILENO;
+		io[i].fn = i == 0 ? stdin_name   : stdout_name;
+		io[i].fl = i == 0 ? read_flags   : write_flags;
+		io[i].error = 0;
 		io[i].prec = 0;
 		io[i].rec = 0;
 		io[i].seek = 0;
@@ -285,19 +182,26 @@ int main(int argc, char *argv[]){
 		int c;
 
 		program_name = argv[0];
-		while((c = getopt(argc, argv, "a:b:B:c:i:hHns:S:o:")) != -1)
+		while((c = getopt(argc, argv, ":a:b:B:c:i:hHns:S:o:")) != -1)
 			switch(c){
 			case 'i': case 'o': i = (c == 'o');
 				if(optarg[0] == '-' && optarg[1] == '\0'){ /* optarg == "-" */
-					io[i].fd = i ? STDIN_FILENO : STDOUT_FILENO;
-					io[i].fn = i ? stdin_name   : stdout_name;
+					io[i].fd = i == 0 ? STDIN_FILENO : STDOUT_FILENO;
+					io[i].fn = i == 0 ? stdin_name   : stdout_name;
 					break;
-				}else if(Io_fdopen(&io[i], optarg) != -1)
-					break;
-				terminate(io);
-				return oserr(optarg);
-			case 'n': noerror = 1; break;
-			case 'H': fmt_output = fmt_human; break;
+				}else{
+					int fd;
+
+					if((fd = open(optarg, io[i].fl, creat_mode)) != -1
+							&& (fdisstd(io[i].fd) || close(io[i].fd) == 0)){
+						io[i].fd = fd;
+						io[i].fn = optarg;
+						break;
+					}
+				}
+				return oserr(optarg, errno);
+			case 'n': noerror = 1;     break;
+			case 'H': fmt = fmt_human; break;
 			case 'a':
 				if(optarg[0] == '\0' || optarg[1] == '\0'){
 					align = optarg[0];
@@ -307,87 +211,142 @@ int main(int argc, char *argv[]){
 			case 'c': case 'b': case 's': case 'B': case 'S': /* numbers */
 				if(c == 'c' && (count = parse(optarg)) >= 0)
 					break;
-				i = (c >= 'A' && c <= 'Z'); /* uppercase changes output */
-				c &= 0x20 /* 0b 0010 0000 */; /* (ASCII) make lowercase */
+				i = (c >= 'A' && c <= 'Z');
+				c |= 0x20 /* 0b 0010 0000 */; /* (ASCII) make lowercase */
 				if((c == 'b' && (io[i].bs = parse(optarg)) > 0)
 						|| (c == 's' && (io[i].seek = parse(optarg)) >= 0))
 					break;
 				/* FALLTHROUGH */
 			default:
-				terminate(io);
 				return usage(program_name);
 			}
 	}
 
-	if(argc > optind){
-		terminate(io);
+	assert(io->fd != STDIN_FILENO  || io->fl == read_flags);
+	assert(io->fd != STDOUT_FILENO || io->fl == write_flags);
+
+	if(argc > optind)
 		return usage(program_name);
-	}
 
-	for(i = 0; i < 2; ++i){
-		if(Io_bufalloc(&io[i]) == NULL){
-			fprintf(stderr, "%s: Failed to allocate %d bytes\n",
+	for(i = 0; i < (sizeof io) / (sizeof *io); ++i){
+		/* buffer allocation */
+		if((io[i].buf = malloc(io[i].bs * (sizeof *(io[i].buf)))) == NULL){
+			fprintf(stderr, "%s: Failed to allocate %zd bytes\n",
 				program_name, io[i].bs);
-			terminate(io);
 			return EX_OSERR;
-		}else if(io[i].seek > 0)
-			Io_fdseek(&io[i]);
-			if(io[i].seek > 0){
-				terminate(io);
-				return oserr(io[i].fn);
-			}
+		}
+		/* easy seeking */
+		if(!fdisstd(io[i].fd) && lseek(io[i].fd, io[i].seek, SEEK_SET) != -1)
+			io[i].seek = 0;
 	}
 
-	do{	/* read */
-		Io_read(&io[0]);
-		if(!noerror && io[0].bufuse == 0)
-			Io_read(&io[0]); /* second chance */
-		if(io[0].bufuse == 0) /* that's all she wrote */
-			break;
-		else if(io[0].bufuse < io[0].bs){
-			++io[0].prec;
-			fprintf(stderr, "%s: Partial read:\n\t", program_name);
-			output(io, fmt_output);
-			if(!noerror)
-				count = 1;
-			if(align >= 0)
-				Io_bufrpad(&io[0], align);
-		}else
-			++io[0].rec;
+	/* hard seeking */
+	if(io[1].seek > 0){
+		size_t t;
+		do{
+			memset(io[1].buf, '\0',
+				(t = io[1].bufuse = MIN(io[1].bs, io[1].seek)));
+			if(Io_write(&io[1])->bufuse == t && !noerror && io[1].error == 0)
+				Io_write(&io[1]); /* second chance */
+			if(io[1].error != 0)
+				return oserr(io[1].fn, io[1].error);
+		}while((io[1].seek -= (t - io[1].bufuse)) > 0 && io[1].bufuse != t);
+		io[1].bufuse = 0;
+	}
+
+	if(io[1].seek > 0){
+		fprintio(stderr, fmt, io);
+		return oserr(io[1].fn, errno);
+	}
+
+	do{
+		assert(io[0].bufuse == 0);
+
+		{ /* read */
+			char skipping;
+			size_t t;
+
+			/* hack to intentionally get a partial read from Io_read */
+			if((skipping = (io[0].seek > 0)) && io[0].seek < io[0].bs)
+				io[0].bufuse = io[0].bs - io[0].seek;
+
+			t = io[0].bufuse;
+			if(Io_read(&io[0])->bufuse == t && !noerror && io[0].error == 0)
+				Io_read(&io[0]); /* second chance */
+			assert(io[0].bufuse >= t);
+			if(io[0].bufuse == t) /* that's all she wrote */
+				break;
+
+			if(/* t < io[0].bufuse && */ io[0].bufuse < io[0].bs){
+				fprintf(stderr, "%s: Partial read:\n\t", program_name);
+				fprintio(stderr, fmt, io);
+				if(!noerror)
+					count = 1;
+				if(align >= 0){
+					/* fill the rest of the ibuf with padding */
+					memset(&(io[0].buf)[io[0].bufuse], align,
+						io[0].bs - io[0].bufuse);
+					io->bufuse = io->bs;
+				}
+			}
+
+			if(skipping){
+				io[0].bufuse = 0;
+				count += (count != 0);
+				continue;
+			}
+		}
 
 		/* write */
 		do{
 			int t;
 
-			if(io[1].bs > io[0].bs){
-				Io_bufxapp(&io[1], &io[0]);
+			if(io[0].bs <= io[1].bs){
+				int n;
+
+				/* saturate obuf */
+				memcpy(io[1].buf, io[0].buf,
+				       (io[1].bufuse = (n = MIN(io[0].bufuse, io[1].bs))));
+				/* permute the copied units out of ibuf */
+				memmove(io[0].buf, &(io[0].buf)[n], (io[0].bufuse -= n));
+			}else /* if(io[0].bs < io[1].bs) */ {
+				int n;
+
+				/* drain what we can from ibuf */
+				memcpy(&(io[1].buf)[io[1].bufuse], io[0].buf,
+				       (n = MIN(io[0].bufuse, io[1].bs - io[1].bufuse)));
+				io[1].bufuse += n;
+				/* permute out the copied units */
+				memmove(io[0].buf, &(io[0].buf)[n], io[0].bs - n);
+				io[0].bufuse -= n;
+
 				if(io[0].bs + io[1].bufuse <= io[1].bs && count != 1)
-					continue; /* we could write more */
-			}else
-				Io_bufxfer(&io[1], &io[0], MIN(io[0].bufuse, io[1].bs));
+					continue; /* obuf not saturated - we could write more */
+			}
 
 			t = io[1].bufuse;
-			Io_write(&io[1]);
-			if(!noerror && io[1].bufuse == t)
+			if(Io_write(&io[1])->bufuse == t && !noerror && io[1].error == 0)
 				Io_write(&io[1]); /* second chance */
-			if(t == io[1].bufuse){ /* no more love */
+			assert(io[1].bufuse <= t);
+			if(io[1].bufuse == t){ /* no more love */
 				count = 1;
 				break;
-			}else if(t > io[1].bufuse && io[1].bufuse > 0){
-				io[1].prec += 1;
+			}
+
+			if(0 < io[1].bufuse /* && io[1].bufuse < t */){
 				fprintf(stderr, "%s: Partial write:\n\t", program_name);
-				output(io, fmt_output);
+				fprintio(stderr, fmt, io);
 				if(!noerror)
 					count = 1;
-			}else if(io[1].bufuse == 0 && t < io[1].bs)
-				++io[1].prec;
-			else
-				++io[1].rec;
+			}
 		}while(io[0].bufuse > 0);
 	}while(count == 0 || --count > 0);
 
-	output(io, fmt_output);
-	terminate(io);
+	fprintio(stderr, fmt, io);
+
+	for(i = 0; i < (sizeof io) / (sizeof *io); ++i)
+		if(io[i].error)
+			return oserr(io[i].fn, io[i].error);
 
 	return EX_OK;
 }