** This code is Copyright (c) 2002, by the authors of nmh. See the
** COPYRIGHT file in the root directory of the nmh distribution for
** complete copyright information.
+**
+** This code compiles the format strings (documented in mh-format(5)) into
+** an internal form to be later processed by fmt_scan.c.
+**
+** What happens here is that the format strings are parsed and an array
+** of struct format structures are returned. Each format structure is
+** a single operation interpreted by the the routines in fmt_scan.c.
+**
+** There is a NOT a one-to-one correspondence between format strings and
+** format instructions; some functions have side effects that can result
+** in multiple instructions being generated. The exact list of instructions
+** generated by a format string can be seem with the nmh fmtdump utility.
+**
+** A list of format instructions can be found in fmt_compile.h.
+**
+** If you wish to add a new function, you will need to do the following
+** things:
+**
+** - Add a new instruction to the list of instructions in fmt_compile.h.
+** Note that test instructions (starting with FT_IF_S_NULL) have special
+** handling, so if you are NOT writing a test function then you need
+** to insert it into the list before that _and_ bump all of the
+** following instruction numbers.
+**
+** - Add the function name to the functable[] array below, and write any
+** special code that your function may require in terms of parsing
+** (it very well may not need anything).
+**
+** - Add the code in fmt_scan.c to handle your new function.
+**
+** - Document the new function in the mh-format(5) man page.
+**
*/
#include <h/mh.h>
+#include <h/utils.h>
#include <h/addrsbr.h>
#include <h/tws.h>
#include <h/fmt_scan.h>
#include <h/fmt_compile.h>
+#include <ctype.h>
+#include <sysexits.h>
-#ifdef TIME_WITH_SYS_TIME
+#ifdef HAVE_SYS_TIME_H
# include <sys/time.h>
-# include <time.h>
-#else
-# ifdef TM_IN_SYS_TIME
-# include <sys/time.h>
-# else
-# include <time.h>
-# endif
#endif
+#include <time.h>
/*
** hash table for deciding if a component is "interesting"
#define TFL_PUTS 1 /* implicit putstr if top level */
#define TFL_PUTN 2 /* implicit putnum if top level */
+/*
+** The functable array maps between the text names of format functions and
+** the format instructions interpreted by the engine in fmt_scan.c.
+**
+** The elements of this structure are as follows:
+**
+** name: The name of the function as seen in the format string. This is
+** what maps a particular function name into a format instruction.
+** type: The type of argument this function expects. Those types are
+** listed above (with the TF_ prefix). This affects what gets
+** placed in the format instruction (the f_un union).
+** f_type: The instruction corresponding to this function (from the list
+** in fmt_compile.h).
+** extra: Used by some functions to provide extra data to the compiler.
+** Uses include:
+** - Providing an alternate instruction to combine a load
+** and test operation (see do_if()).
+** - Passed in f_value in the format instruction to provide
+** extra information for the engine (see FT_LV_DAT handling
+** in fmt_scan.c).
+** - Provide a hint as to preprocessing that is required for
+** this instruction (see do_name()).
+** flags: See the definitions for TFL_PUTS & TFL_PUTN above.
+*/
struct ftable {
char *name; /* function name */
char type; /* argument type */
/* Add new component to the hash table */
#define NEWCOMP(cm,name) do { \
- cm = ((struct comp *) calloc(1, sizeof (struct comp)));\
+ cm = ((struct comp *) mh_xcalloc(1, sizeof (struct comp)));\
cm->c_name = name;\
ncomp++;\
i = CHASH(name);\
static char *do_if(char *);
+/*
+** Lookup a function name in the functable
+*/
static struct ftable *
lookup(char *name)
{
- register struct ftable *t = functable;
- register char *nm;
- register char c = *name;
+ struct ftable *t = functable;
+ char *nm;
+ char c = *name;
while ((nm = t->name)) {
if (*nm == c && strcmp(nm, name) == 0)
}
advise(NULL, "\"%s\": format compile error - %s",
- &usr_fstring[errpos-errctx], str);
- adios(NULL, "%*s", errctx+1, "^");
+ &usr_fstring[errpos-errctx], str);
+ adios(EX_SOFTWARE, NULL, "%*s", errctx+1, "^");
}
/*
** Return the number of header components found in the format
** string.
*/
-
int
fmt_compile(char *fstring, struct format **fmt)
{
- register char *cp;
- int i;
+ char *cp;
+ size_t i;
if (format_string)
free(format_string);
i = strlen(fstring)/2 + 1;
if (i == 1)
i++;
- next_fp = formatvec = (struct format *)calloc((size_t) i,
+ next_fp = formatvec = (struct format *)mh_xcalloc((size_t) i,
sizeof(struct format));
if (next_fp == NULL)
- adios(NULL, "unable to allocate format storage");
+ adios(EX_OSERR, NULL, "unable to allocate format storage");
ncomp = 0;
infunction = 0;
static char *
compile(char *sp)
{
- register char *cp = sp;
- register int c;
+ char *cp = sp;
+ int c;
for (;;) {
sp = cp;
}
+/*
+** Process functions & components (handle field width here as well
+*/
static char *
do_spec(char *sp)
{
- register char *cp = sp;
- register int c;
-#ifndef lint
- register int ljust = 0;
-#endif /* not lint */
- register int wid = 0;
- register char fill = ' ';
+ char *cp = sp;
+ int c;
+ int ljust = 0;
+ int wid = 0;
+ char fill = ' ';
c = *cp++;
if (c == '-') {
return (cp);
}
+
+/*
+** Process a component name. Normally this involves generating an FT_COMP
+** instruction for the specified component. If preprocess is set, then we
+** do some extra processing.
+*/
static char *
do_name(char *sp, int preprocess)
{
- register char *cp = sp;
- register int c;
- register int i;
+ char *cp = sp;
+ int c;
+ int i;
static int primed = 0;
while (isalnum(c = *cp++) || c == '-' || c == '_')
CERROR("component used as both date and address");
}
cm->c_tws = (struct tws *)
- calloc((size_t) 1, sizeof(*cm->c_tws));
+ mh_xcalloc((size_t) 1, sizeof(*cm->c_tws));
fp->f_type = preprocess;
PUTCOMP(sp);
cm->c_type |= CT_DATE;
return (cp);
}
+
+/*
+** Generate one or more instructions corresponding to the named function.
+** The different type of function arguments are handled here.
+*/
static char *
do_func(char *sp)
{
- register char *cp = sp;
- register int c;
- register struct ftable *t;
- register int n;
+ char *cp = sp;
+ int c;
+ struct ftable *t;
+ int n;
int mflag; /* minus sign in NUM */
infunction++;
return (cp);
}
+
+/*
+** Handle an expression as an argument. Basically we call one of do_name(),
+** do_func(), or do_if()
+*/
static char *
do_expr(char *sp, int preprocess)
{
- register char *cp = sp;
- register int c;
+ char *cp = sp;
+ int c;
if ((c = *cp++) == '{') {
cp = do_name(cp, preprocess);
return (cp);
}
+
+/*
+** I am guessing this was for some kind of loop statement, which would have
+** looked like %[ .... %]. It looks like the way this would have worked
+** is that the format engine would have seen that FT_DONE had a 1 in the
+** f_un.f_un_value and then decided whether or not to continue the loop.
+** There is no support for this in the format engine, so right now if
+** you try using it you will reach the FT_DONE and simply stop. I'm leaving
+** this here in case someone wants to continue the work.
+*/
static char *
do_loop(char *sp)
{
- register char *cp = sp;
+ char *cp = sp;
struct format *floop;
floop = next_fp;
return cp;
}
+
+/*
+** Handle an if-elsif-endif statement. Note here that the branching
+** is handled by the f_skip member of the struct format (which is really
+** just f_width overloaded). This number controls how far to move forward
+** (or back) in the format instruction array.
+*/
static char *
do_if(char *sp)
{
- register char *cp = sp;
- register struct format *fexpr, *fif = (struct format *)NULL;
- register int c = '<';
+ char *cp = sp;
+ struct format *fexpr, *fif = (struct format *)NULL;
+ int c = '<';
for (;;) {
if (c == '<') { /* doing an IF */