Change the semantics of -i (in-place editing) so that it treats

each file independently from other files.  The new semantics are
desired in the most of practical cases, e.g.: delete lines 5-9
from each file.

Keep the previous semantics of -i under a new option, -I, which
uses a single continuous address space covering all files to edit
in-place -- they are too cool to just drop them.

Add regression tests for -i and -I.

Approved by:	dds
Compared with:	GNU sed
Discussed on:	-hackers
MFC after:	2 weeks

tools/regression/usr.bin/sed/regress.sh

@@ -2,7 +2,7 @@
 
 REGRESSION_START($1)
 
-echo '1..16'
+echo '1..20'
 
 REGRESSION_TEST(`G', `sed G < regress.in')
 REGRESSION_TEST(`P', `sed P < regress.in')
@@ -28,6 +28,40 @@ foo
 REGRESSION_TEST(`b2a', `sed ''`2,3b
 1,2d''` < regress.in')
 
+`
+inplace_test()
+{
+	expr="$1"
+	rc=0
+	ns=$(jot 5)
+	ins= outs= _ins=
+	for n in $ns; do
+		jot -w "l${n}_%d" 9 | tee lines.in.$n lines._in.$n | \
+		    sed "$expr" > lines.out.$n
+		ins="$ins lines.in.$n"
+		outs="$outs lines.out.$n"
+		_ins="$_ins lines._in.$n"
+	done
+	sed "$expr" $_ins > lines.out
+
+	sed -i "" "$expr" $ins
+	sed -I "" "$expr" $_ins
+
+	for n in $ns; do
+		diff -u lines.out.$n lines.in.$n || rc=1
+	done
+	cat $_ins | diff -u lines.out - || rc=1
+	rm -f $ins $outs $_ins lines.out
+
+	return $rc
+}
+'
+
+REGRESSION_TEST_FREEFORM(`inplace1', `inplace_test 3,6d')
+REGRESSION_TEST_FREEFORM(`inplace2', `inplace_test 8,30d')
+REGRESSION_TEST_FREEFORM(`inplace3', `inplace_test 20,99d')
+REGRESSION_TEST_FREEFORM(`inplace4', `inplace_test "{;{;8,30d;};}"')
+
 REGRESSION_TEST(`hanoi', `echo ":abcd: : :" | sed -f hanoi.sed')
 REGRESSION_TEST(`math', `echo "4+7*3+2^7/3" | sed -f math.sed')
 

usr.bin/sed/extern.h

@@ -52,4 +52,5 @@ char	*cu_fgets(char *, int, int *);
 int	 mf_fgets(SPACE *, enum e_spflag);
 int	 lastline(void);
 void	 process(void);
+void	 resetranges(void);
 char	*strregerror(int, regex_t *);

usr.bin/sed/main.c

@@ -101,9 +101,12 @@ int aflag, eflag, nflag;
 int rflags = 0;
 static int rval;		/* Exit status */
 
+static int ispan;		/* Whether inplace editing spans across files */
+
 /*
  * Current file and line number; line numbers restart across compilation
- * units, but span across input files.
+ * units, but span across input files.  The latter is optional if editing
+ * in place.
  */
 const char *fname;		/* File name. */
 const char *outfname;		/* Output file name */
@@ -127,11 +130,15 @@ main(int argc, char *argv[])
 	fflag = 0;
 	inplace = NULL;
 
-	while ((c = getopt(argc, argv, "Eae:f:i:ln")) != -1)
+	while ((c = getopt(argc, argv, "EI:ae:f:i:ln")) != -1)
 		switch (c) {
 		case 'E':
 			rflags = REG_EXTENDED;
 			break;
+		case 'I':
+			inplace = optarg;
+			ispan = 1;	/* span across input files */
+			break;
 		case 'a':
 			aflag = 1;
 			break;
@@ -149,6 +156,7 @@ main(int argc, char *argv[])
 			break;
 		case 'i':
 			inplace = optarg;
+			ispan = 0;	/* don't span across input files */
 			break;
 		case 'l':
 			if(setlinebuf(stdout) != 0)
@@ -307,7 +315,7 @@ mf_fgets(SPACE *sp, enum e_spflag spflag)
 		/* stdin? */
 		if (files->fname == NULL) {
 			if (inplace != NULL)
-				errx(1, "-i may not be used with stdin");
+				errx(1, "-I or -i may not be used with stdin");
 			infile = stdin;
 			fname = "stdin";
 			outfile = stdout;
@@ -380,6 +388,10 @@ mf_fgets(SPACE *sp, enum e_spflag spflag)
 			fchown(fileno(outfile), sb.st_uid, sb.st_gid);
 			fchmod(fileno(outfile), sb.st_mode & ALLPERMS);
 			outfname = tmpfname;
+			if (!ispan) {
+				linenum = 0;
+				resetranges();
+			}
 		} else {
 			outfile = stdout;
 			outfname = "stdout";
@@ -448,7 +460,7 @@ lastline(void)
 {
 	int ch;
 
-	if (files->next != NULL)
+	if (files->next != NULL && (inplace == NULL || ispan))
 		return (0);
 	if ((ch = getc(infile)) == EOF)
 		return (1);

usr.bin/sed/process.c

@@ -316,6 +316,19 @@ applies(struct s_command *cp)
 	return (cp->nonsel ? ! r : r);
 }
 
+/*
+ * Reset all inrange markers.
+ */
+void
+resetranges(void)
+{
+	struct s_command *cp;
+
+	for (cp = prog; cp; cp = cp->code == '{' ? cp->u.c : cp->next)
+		if (cp->a2)
+			cp->inrange = 0;
+}
+
 /*
  * substitute --
  *	Do substitutions in the pattern space.  Currently, we build a

usr.bin/sed/sed.1

@@ -46,6 +46,7 @@
 .Op Fl Ealn
 .Op Fl e Ar command
 .Op Fl f Ar command_file
+.Op Fl I Ar extension
 .Op Fl i Ar extension
 .Op Ar
 .Sh DESCRIPTION
@@ -95,7 +96,7 @@ Append the editing commands found in the file
 .Ar command_file
 to the list of commands.
 The editing commands should each be listed on a separate line.
-.It Fl i Ar extension
+.It Fl I Ar extension
 Edit files in-place, saving backups with the specified
 .Ar extension .
 If a zero-length
@@ -105,6 +106,36 @@ It is not recommended to give a zero-length
 .Ar extension
 when in-place editing files, as you risk corruption or partial content
 in situations where disk space is exhausted, etc.
+.Pp
+Note that in-place editing with
+.Fl I
+still takes place in a single continuous line address space covering
+all files, although each file preserves its individuality instead of
+forming one output stream.
+The line counter is never reset between files, address ranges can span
+file boundaries, and the
+.Dq $
+address matches only the last line of the last file.
+(See
+.Sx "Sed Addresses" . )
+That can lead to unexpected results in many cases of in-place editing,
+where using
+.Fl i
+is desired.
+.It Fl i Ar extension
+Edit files in-place similarly to
+.Fl I ,
+but treat each file independently from other files.
+In particular, line numbers in each file start at 1,
+the
+.Dq $
+address matches the last line of the current file,
+and address ranges are limited to the current file.
+(See
+.Sx "Sed Addresses" . )
+The net result is as though each file were edited by a separate
+.Nm
+instance.
 .It Fl l
 Make output line buffered.
 .It Fl n
@@ -140,13 +171,28 @@ Some of the functions use a
 .Em "hold space"
 to save all or part of the pattern space for subsequent retrieval.
 .Sh "Sed Addresses"
-An address is not required, but if specified must be a number (that counts
+An address is not required, but if specified must have one of the
+following formats:
+.Bl -bullet -offset indent
+.It
+a number that counts
 input lines
-cumulatively across input files), a dollar
+cumulatively across input files (or in each file independently
+if a
+.Fl i
+option is in effect);
+.It
+a dollar
 .Pq Dq $
-character that addresses the last line of input, or a context address
-(which consists of a regular expression preceded and followed by a
-delimiter).
+character that addresses the last line of input (or the last line
+of the current file if a
+.Fl i
+option was specified);
+.It
+a context address
+that consists of a regular expression preceded and followed by a
+delimiter.
+.El
 .Pp
 A command line with no addresses selects every pattern space.
 .Pp
@@ -533,7 +579,7 @@ utility is expected to be a superset of the
 specification.
 .Pp
 The
-.Fl E , a
+.Fl E , I , a
 and
 .Fl i
 options are non-standard