commit: 68115096138b112147503b11559573256cae2e47
parent 2a3955731ccffee4cb95ce17eeffde861c9a70e0
Author: Haelwenn (lanodan) Monnier <contact@hacktivis.me>
Date: Tue, 3 Sep 2024 16:35:58 +0200
cmd/join: drop compatibility layer for historical options
Using https://codesearch.debian.net/ in regex mode to search
for historical usage:
* `\bjoin (--?[^a] \w+ )?-a\b` => no usage of operand-less -a
* `\bjoin (--?[^j] \w+ )?-j` => <5 actual hits, likely patcheable
* `\bjoin (--?[^o] \w+ )?-o` => <5 hits, easily patcheable
I don't think it's worth keeping non-POSIX option parsing in place.
Historical -j [num] managed by getopt() can stay for now.
Diffstat:
| M | cmd/join.1 | 56 | ++++++++++---------------------------------------------- |
| M | cmd/join.c | 89 | +------------------------------------------------------------------------------ |
2 files changed, 11 insertions(+), 134 deletions(-)
diff --git a/cmd/join.1 b/cmd/join.1
@@ -231,57 +231,21 @@ Also see the note about
.Xr sort 1
above to understand why we need to sort the intermediate result.
.Bd -literal -offset indent
-$ join -t, -1 2 -o 1.2 2.2 nobel_laureates.txt nobel_nationalities.txt | \e
- sort -k2 -t, | join -t, -e "<<NULL>>" -1 2 -o 1.1 2.2 - capitals.txt
+$ join -t, -1 2 -o "1.2 2.2" nobel_laureates.txt nobel_nationalities.txt | \e
+ sort -k2 -t, | join -t, -e "<<NULL>>" -1 2 -o "1.1 2.2" - capitals.txt
Elie Ducommun,<<NULL>>
Jean Henri Dunant,<<NULL>>
.Ed
.Sh COMPATIBILITY
-For compatibility with historic versions of
-.Nm ,
-the following options are available:
-.Bl -tag -width indent
-.It Fl a
-In addition to the default output, produce a line for each unpairable line
-in both
-.Ar file1
-and
-.Ar file2 .
-.It Fl j1 Ar field
-Join on the
-.Ar field Ns 'th
-field of
-.Ar file1 .
-.It Fl j2 Ar field
-Join on the
-.Ar field Ns 'th
-field of
-.Ar file2 .
-.It Fl j Ar field
-Join on the
-.Ar field Ns 'th
-field of both
-.Ar file1
-and
-.Ar file2 .
-.It Fl o Ar list ...
-Historical implementations of
-.Nm
-permitted multiple arguments to the
+Compatibility with
+.Fl a
+and multi-operand
.Fl o
-option.
-These arguments were of the form
-.Ar file_number . Ns Ar field_number
-as described
-for the current
-.Fl o
-option.
-This has obvious difficulties in the presence of files named
-.Pa 1.2 .
-.El
-.Pp
-These options are available only so historic shell scripts do not require
-modification and should not be used.
+from historical versions of
+.Nm
+got dropped, compatibility with
+.Fl j Ar field
+is kept for now.
.Sh SEE ALSO
.Xr awk 1 ,
.Xr comm 1 ,
diff --git a/cmd/join.c b/cmd/join.c
@@ -97,7 +97,6 @@ static void fieldarg(char *);
static void joinlines(INPUT *, INPUT *);
static int mbscoll(const char *, const char *);
static char *mbssep(char **, const wchar_t *);
-static void obsolete(char **);
static void outfield(LINE *, unsigned long, int);
static void outoneline(INPUT *, LINE *);
static void outtwoline(INPUT *, LINE *, INPUT *, LINE *);
@@ -118,15 +117,10 @@ main(int argc, char *argv[])
F2 = &input2;
aflag = vflag = 0;
- obsolete(argv);
- while((ch = getopt(argc, argv, "\01a:e:j:1:2:o:t:v:")) != -1)
+ while((ch = getopt(argc, argv, "a:e:j:1:2:o:t:v:")) != -1)
{
switch(ch)
{
- case '\01': /* See comment in obsolete(). */
- aflag = 1;
- F1->unpair = F2->unpair = 1;
- break;
case '1':
if((F1->joinf = strtol(optarg, &end, 10)) < 1) errx(1, "-1 option field number less than 1");
if(*end) errx(1, "illegal field number -- %s", optarg);
@@ -560,87 +554,6 @@ fieldarg(char *option)
}
static void
-obsolete(char **argv)
-{
- size_t len;
- char **p, *ap, *t;
-
- while((ap = *++argv) != NULL)
- {
- /* Return if "--". */
- if(ap[0] == '-' && ap[1] == '-') return;
- /* skip if not an option */
- if(ap[0] != '-') continue;
- switch(ap[1])
- {
- case 'a':
- /*
- * The original join allowed "-a", which meant the
- * same as -a1 plus -a2. POSIX 1003.2, Draft 11.2
- * only specifies this as "-a 1" and "a -2", so we
- * have to use another option flag, one that is
- * unlikely to ever be used or accidentally entered
- * on the command line. (Well, we could reallocate
- * the argv array, but that hardly seems worthwhile.)
- */
- if(ap[2] == '\0' &&
- (argv[1] == NULL || (strcmp(argv[1], "1") != 0 && strcmp(argv[1], "2") != 0)))
- {
- ap[1] = '\01';
- warnx("-a option used without an argument; "
- "reverting to historical behavior");
- }
- break;
- case 'j':
- /*
- * The original join allowed "-j[12] arg" and "-j arg".
- * Convert the former to "-[12] arg". Don't convert
- * the latter since getopt(3) can handle it.
- */
- switch(ap[2])
- {
- case '1':
- if(ap[3] != '\0') goto jbad;
- ap[1] = '1';
- ap[2] = '\0';
- break;
- case '2':
- if(ap[3] != '\0') goto jbad;
- ap[1] = '2';
- ap[2] = '\0';
- break;
- case '\0':
- break;
- default:
- jbad:
- errx(1, "illegal option -- %s", ap);
- usage();
- }
- break;
- case 'o':
- /*
- * The original join allowed "-o arg arg".
- * Convert to "-o arg -o arg".
- */
- if(ap[2] != '\0') break;
- for(p = argv + 2; *p; ++p)
- {
- if(p[0][0] == '0' || ((p[0][0] != '1' && p[0][0] != '2') || p[0][1] != '.')) break;
- len = strlen(*p);
- if(len - 2 != strspn(*p + 2, "0123456789")) break;
- if((t = malloc(len + 3)) == NULL) err(1, NULL);
- t[0] = '-';
- t[1] = 'o';
- memmove(t + 2, *p, len + 1);
- *p = t;
- }
- argv = p - 1;
- break;
- }
- }
-}
-
-static void
usage(void)
{
(void)fprintf(stderr,
