dj.1: elaborate on skip/seek behavior, provide another example

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 +34,17 @@ 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.
+
+Seeks and skips aren\(cqt counted in the output statistics because they're
+guaranteed to succeed (or the utility will exit unsuccessfully, before it has
+written any data).
 .\"
 .SH OPTIONS
 
@@ -80,6 +89,54 @@ 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

src/dj.c

@@ -16,10 +16,11 @@
  * 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 */
 #include <unistd.h> /* close(2), getopt(3), lseek(2), read(2), write(2),
@@ -37,16 +38,16 @@ struct Io{
 	int bs;       	/* buffer size (-bB) */
 	size_t bufuse;	/* buffer usage */
 	char *buf;		/* buffer */
-	int bytes;		/* bytes processed */
+	size_t 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 */
+	size_t prec;	/* partial records processed */
+	size_t rec; 	/* records processed */
 	long seek;		/* bytes to seek/skip (will be 0 after skippage) (-sS) */
 }; 
 
-/* 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";
 
@@ -58,56 +59,22 @@ 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 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)));
-}
+/* Macro to check if fd is stdin or stdout */
+#define fdisstd(fd) ((fd) == STDIN_FILENO || (fd) == STDOUT_FILENO)
 
 /* 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){
 
+	assert(io != NULL);
+
 	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
@@ -115,32 +82,27 @@ Io_bufxapp(struct Io *dest, struct Io *src){
 static struct Io*
 Io_bufxfer(struct Io *dest, struct Io *src, int n){
 
+	assert(dest != NULL && src != NULL);
+
 	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;
 
+	assert(io != NULL);
+
 	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){
+			&& (fdisstd(io->fd) || close(io->fd) == 0)){
 		io->fd = fd;
 		io->fn = fn;
 	}
@@ -148,43 +110,6 @@ Io_fdopen(struct Io *io, char *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. */
@@ -223,7 +148,7 @@ oserr(char *s){
 /* Prints statistics regarding the use of dj, particularly partially and
  * completely read and written records. */
 static void
-output(struct Io io[2], char *fmt){
+printio(char *fmt, struct Io io[2]){
 
 	fprintf(stderr, fmt,
 		io[0].rec, io[0].prec, io[1].rec, io[1].prec,
@@ -260,7 +185,7 @@ 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) */
+	char *fmt; /* == fmt_asv (default) or fmt_human (-H) */
 	size_t i; /* side of io being modified */
 	struct Io io[2];
 	char noerror; /* 0=exits (default) 1=retries on partial reads or writes */
@@ -268,14 +193,15 @@ int main(int argc, char *argv[]){
 	/* Set defaults. */
 	align = -1;
 	count = 0;
-	fmt_output = fmt_asv;
+	fmt = fmt_asv;
 	noerror = 0;
 	for(i = 0; i < 2; ++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].prec = 0;
 		io[i].rec = 0;
 		io[i].seek = 0;
@@ -285,19 +211,18 @@ 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;
+			case 'H': fmt = fmt_human; break;
 			case 'a':
 				if(optarg[0] == '\0' || optarg[1] == '\0'){
 					align = optarg[0];
@@ -307,36 +232,71 @@ 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);
 			}
 	}
 
+	assert(io->fd != STDIN_FILENO  || io->fl == read_flags);
+	assert(io->fd != STDOUT_FILENO || io->fl == write_flags);
+
 	if(argc > optind){
-		terminate(io);
 		return usage(program_name);
 	}
 
 	for(i = 0; i < 2; ++i){
-		if(Io_bufalloc(&io[i]) == NULL){
+		/* buffer allocation */
+		if((io[i].buf = malloc(io[i].bs * (sizeof *(io[i].buf)))) == NULL){
 			fprintf(stderr, "%s: Failed to allocate %d 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);
+		}
+		/* easy seeking */
+		if(!fdisstd(io[i].fd) && lseek(io[i].fd, io[i].seek, SEEK_SET) != -1)
+			io[i].seek = 0;
+	}
+
+	/* hard skipping */
+	if(io[0].seek > 0){
+		do{
+			if((io[0].bufuse = read(
+						io[0].fd, io[0].buf, MIN(io[0].bs, io[0].seek)))
+					== 0)
+				/* second chance */
+				io->bufuse = read(
+					io[0].fd, io[0].buf, MIN(io[0].bs, io[0].seek));
+		}while((io[0].seek -= io[0].bufuse) > 0 && io[0].bufuse != 0);
+		io[0].bufuse = 0;
+	}
+
+	/* hard seeking */
+	if(io[1].seek > 0){
+		memset(io[1].buf, '\0', io[1].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[1].bufuse = write(
+						io[1].fd, io[1].buf, MIN(io[1].bs, io[1].seek)))
+					== 0)
+				/* second chance */
+				io[1].bufuse = write(
+					io[1].fd, io[1].buf, MIN(io[1].bs, io[1].seek));
+		}while((io[1].seek -= io[1].bufuse) > 0 && io[1].bufuse != 0);
+		io[1].bufuse = 0;
+	}
+
+	/* Sought bytes aren't counted in the statistics because successful seeking
+	 * is guaranteed here. */
+	for(i = 0; i < 2; ++i)
+		if(io[i].seek > 0)
 			return oserr(io[i].fn);
-			}
-	}
 
 	do{	/* read */
 		Io_read(&io[0]);
@@ -347,7 +307,7 @@ int main(int argc, char *argv[]){
 		else if(io[0].bufuse < io[0].bs){
 			++io[0].prec;
 			fprintf(stderr, "%s: Partial read:\n\t", program_name);
-			output(io, fmt_output);
+			printio(fmt, io);
 			if(!noerror)
 				count = 1;
 			if(align >= 0)
@@ -360,7 +320,16 @@ int main(int argc, char *argv[]){
 			int t;
 
 			if(io[1].bs > io[0].bs){
-				Io_bufxapp(&io[1], &io[0]);
+				int n;
+
+				/* copy from ibuf as much as possible to the obuf */
+				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
@@ -376,7 +345,7 @@ int main(int argc, char *argv[]){
 			}else if(t > io[1].bufuse && io[1].bufuse > 0){
 				io[1].prec += 1;
 				fprintf(stderr, "%s: Partial write:\n\t", program_name);
-				output(io, fmt_output);
+				printio(fmt, io);
 				if(!noerror)
 					count = 1;
 			}else if(io[1].bufuse == 0 && t < io[1].bs)
@@ -386,8 +355,7 @@ int main(int argc, char *argv[]){
 		}while(io[0].bufuse > 0);
 	}while(count == 0 || --count > 0);
 
-	output(io, fmt_output);
-	terminate(io);
+	printio(fmt, io);
 
 	return EX_OK;
 }