* Re: -Tman with -mman
2011-10-03 22:00 ` Ingo Schwarze
@ 2011-10-03 22:18 ` Kristaps Dzonsons
2011-10-05 9:36 ` Kristaps Dzonsons
1 sibling, 0 replies; 6+ messages in thread
From: Kristaps Dzonsons @ 2011-10-03 22:18 UTC (permalink / raw)
To: tech; +Cc: Ingo Schwarze
On 03/10/2011 23:00, Ingo Schwarze wrote:
> Hi Kristaps,
>
> Kristaps Dzonsons wrote on Sun, Oct 02, 2011 at 09:53:33PM +0200:
>> On 02/10/2011 17:40, Joerg Sonnenberger wrote:
>>> On Sun, Oct 02, 2011 at 05:04:18PM +0200, Kristaps Dzonsons wrote:
>
>>>> What sort of behaviour is most reasonable when providing a man
>>>> manual to the new -Tman? I think this should be ironed out before
>>>> the release. The mandoc manual says "-Tman only with mdoc" and, if
>>>> man is provided to -Tman, outputs nothing (like -Tlint).
>>>>
>>>> It makes the most sense to me to output the original file (after
>>>> preprocessing, I guess). All this requires is a duplicate buffer in
>>>> read.c (conditional on -Tman being present to avoid overhead in
>>>> normal usage). I don't like the idea of putting out an error (or
>>>> nothing at all) because it special-cases modes for input.
>
>>> Agreed, just giving back the original input (maybe reformatted) sounds
>>> like the best approach.
>
> I agree that makes sense if the code is parsed (.man -> AST)
> and then freshly formatted (AST -> .man) in the sense of a validating,
> normalizing parser.
>
> Oops, now that i think about it, maybe i went the wrong way with
> mdoc -> man? Maybe i should have implemented that as a two-step
> translation? First mdoc-AST -> man-AST, then man-AST -> man-code?
> Maybe i should start over? Hm...
>
>> Putting some code to the idea...
>
> Not that i'm really opposed to your small patch - but i'm not
> sure this is terribly useful either, so far it's mostly
> a fancy cat(1).
>
> In case we maybe want to implement a real normalizing parser later,
> i'm both OK with erroring out until we get there or putting in
> your fancy cat(1).
>
>> (Sorry for the churn, I ordered the
>> mandoc.h functions. They're the same except for the mandoc_keep()
>> and mandoc_getkeep() additions). This isn't quite commit-ready, but
>> a start. Basically, I ask for a keep buffer if OUTT_MAN is
>> specified. This needed some machinery to expose the parser to the
>> front-end.
>>
>> Thoughts?
>
> Well, as a placeholder for something better later, the general
> idea seems OK.
Ingo,
I think the approach you have is fine---since I don't foresee this mode
as being heavily used, let's keep it as small as possible. I think
adding normalised-output modes to mandoc is just bloat.
Meanwhile, I'll put in the fancy-cat. This fits my expectation of an
output mode always producing output. (It also doesn't cost anything.)
Incidentally, the docs need to be updated to note that the `so'
invocations are resolved prior to output and the roff in mdoc input is
stripped. This may not be expected behaviour, so...
Thanks,
Kristaps
--
To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: -Tman with -mman
2011-10-03 22:00 ` Ingo Schwarze
2011-10-03 22:18 ` Kristaps Dzonsons
@ 2011-10-05 9:36 ` Kristaps Dzonsons
1 sibling, 0 replies; 6+ messages in thread
From: Kristaps Dzonsons @ 2011-10-05 9:36 UTC (permalink / raw)
To: tech
[-- Attachment #1: Type: text/plain, Size: 2364 bytes --]
On 04/10/2011 00:00, Ingo Schwarze wrote:
> Hi Kristaps,
>
> Kristaps Dzonsons wrote on Sun, Oct 02, 2011 at 09:53:33PM +0200:
>> On 02/10/2011 17:40, Joerg Sonnenberger wrote:
>>> On Sun, Oct 02, 2011 at 05:04:18PM +0200, Kristaps Dzonsons wrote:
>
>>>> What sort of behaviour is most reasonable when providing a man
>>>> manual to the new -Tman? I think this should be ironed out before
>>>> the release. The mandoc manual says "-Tman only with mdoc" and, if
>>>> man is provided to -Tman, outputs nothing (like -Tlint).
>>>>
>>>> It makes the most sense to me to output the original file (after
>>>> preprocessing, I guess). All this requires is a duplicate buffer in
>>>> read.c (conditional on -Tman being present to avoid overhead in
>>>> normal usage). I don't like the idea of putting out an error (or
>>>> nothing at all) because it special-cases modes for input.
>
>>> Agreed, just giving back the original input (maybe reformatted) sounds
>>> like the best approach.
>
> I agree that makes sense if the code is parsed (.man -> AST)
> and then freshly formatted (AST -> .man) in the sense of a validating,
> normalizing parser.
>
> Oops, now that i think about it, maybe i went the wrong way with
> mdoc -> man? Maybe i should have implemented that as a two-step
> translation? First mdoc-AST -> man-AST, then man-AST -> man-code?
> Maybe i should start over? Hm...
>
>> Putting some code to the idea...
>
> Not that i'm really opposed to your small patch - but i'm not
> sure this is terribly useful either, so far it's mostly
> a fancy cat(1).
>
> In case we maybe want to implement a real normalizing parser later,
> i'm both OK with erroring out until we get there or putting in
> your fancy cat(1).
>
>> (Sorry for the churn, I ordered the
>> mandoc.h functions. They're the same except for the mandoc_keep()
>> and mandoc_getkeep() additions). This isn't quite commit-ready, but
>> a start. Basically, I ask for a keep buffer if OUTT_MAN is
>> specified. This needed some machinery to expose the parser to the
>> front-end.
>>
>> Thoughts?
>
> Well, as a placeholder for something better later, the general
> idea seems OK.
Hi,
Here's a polished patch for this. It's more or less what you saw
earlier, but has to strip `so' calls from -mman input (else you'd get an
`so' followed by the file contents).
No-ways? Oks?
Kristaps
[-- Attachment #2: patch.txt --]
[-- Type: text/plain, Size: 9957 bytes --]
Index: main.c
===================================================================
RCS file: /usr/vhosts/mdocml.bsd.lv/cvs/mdocml/main.c,v
retrieving revision 1.164
diff -u -r1.164 main.c
--- main.c 17 Sep 2011 15:00:51 -0000 1.164
+++ main.c 5 Oct 2011 09:34:15 -0000
@@ -127,6 +127,12 @@
curp.mp = mparse_alloc(type, curp.wlevel, mmsg, &curp);
+ /*
+ * Conditionally start up the lookaside buffer before parsing.
+ */
+ if (OUTT_MAN == curp.outtype)
+ mparse_keep(curp.mp);
+
argc -= optind;
argv += optind;
@@ -252,6 +258,7 @@
break;
case (OUTT_MAN):
curp->outmdoc = man_mdoc;
+ curp->outman = man_man;
break;
case (OUTT_PDF):
/* FALLTHROUGH */
Index: main.h
===================================================================
RCS file: /usr/vhosts/mdocml.bsd.lv/cvs/mdocml/main.h,v
retrieving revision 1.14
diff -u -r1.14 main.h
--- main.h 18 Sep 2011 14:14:15 -0000 1.14
+++ main.h 5 Oct 2011 09:34:15 -0000
@@ -42,6 +42,7 @@
void tree_man(void *, const struct man *);
void man_mdoc(void *, const struct mdoc *);
+void man_man(void *, const struct man *);
void *locale_alloc(char *);
void *utf8_alloc(char *);
Index: man.c
===================================================================
RCS file: /usr/vhosts/mdocml.bsd.lv/cvs/mdocml/man.c,v
retrieving revision 1.111
diff -u -r1.111 man.c
--- man.c 28 Jul 2011 14:17:11 -0000 1.111
+++ man.c 5 Oct 2011 09:34:15 -0000
@@ -648,3 +648,11 @@
if (m && m->first == n)
m->first = NULL;
}
+
+const struct mparse *
+man_mparse(const struct man *m)
+{
+
+ assert(m && m->parse);
+ return(m->parse);
+}
Index: man.h
===================================================================
RCS file: /usr/vhosts/mdocml.bsd.lv/cvs/mdocml/man.h,v
retrieving revision 1.58
diff -u -r1.58 man.h
--- man.h 23 Mar 2011 12:33:01 -0000 1.58
+++ man.h 5 Oct 2011 09:34:15 -0000
@@ -105,6 +105,7 @@
const struct man_node *man_node(const struct man *);
const struct man_meta *man_meta(const struct man *);
+const struct mparse *man_mparse(const struct man *);
__END_DECLS
Index: mandoc.1
===================================================================
RCS file: /usr/vhosts/mdocml.bsd.lv/cvs/mdocml/mandoc.1,v
retrieving revision 1.93
diff -u -r1.93 mandoc.1
--- mandoc.1 26 Sep 2011 20:47:23 -0000 1.93
+++ mandoc.1 5 Oct 2011 09:34:15 -0000
@@ -176,11 +176,11 @@
See
.Sx Locale Output .
.It Fl T Ns Cm man
-Produce output in
+Produce
.Xr man 7
-format; only useful when applied to
-.Fl m Ns Cm doc
-input.
+format output.
+See
+.Sx Man Output .
.It Fl T Ns Cm pdf
Produce PDF output.
See
@@ -305,6 +305,26 @@
See
.Sx ASCII Output
for font style specification and available command-line arguments.
+.Ss Man Output
+Translate input format into
+.Xr man 7
+output format.
+This is useful for distributing manual sources to legancy systems
+lacking
+.Xr mdoc 7
+formatters.
+.Pp
+If
+.Xr mdoc 7
+is passed as input, it is translated into
+.Xr man 7 ;
+if the input format is
+.Xr man 7 ,
+it is parsed and re-outputted.
+In either case, the
+.Xr roff 7
+.Sq so
+macros are processed prior to producing output.
.Ss PDF Output
PDF-1.1 output may be generated by
.Fl T Ns Cm pdf .
Index: mandoc.3
===================================================================
RCS file: /usr/vhosts/mdocml.bsd.lv/cvs/mdocml/mandoc.3,v
retrieving revision 1.13
diff -u -r1.13 mandoc.3
--- mandoc.3 18 Aug 2011 08:58:44 -0000 1.13
+++ mandoc.3 5 Oct 2011 09:34:15 -0000
@@ -22,6 +22,7 @@
.Nm mandoc ,
.Nm mandoc_escape ,
.Nm man_meta ,
+.Nm man_mparse ,
.Nm man_node ,
.Nm mchars_alloc ,
.Nm mchars_free ,
@@ -33,6 +34,8 @@
.Nm mdoc_node ,
.Nm mparse_alloc ,
.Nm mparse_free ,
+.Nm mparse_getkeep ,
+.Nm mparse_keep ,
.Nm mparse_readfd ,
.Nm mparse_reset ,
.Nm mparse_result ,
@@ -55,6 +58,10 @@
.Fo man_meta
.Fa "const struct man *man"
.Fc
+.Ft "const struct mparse *"
+.Fo man_mparse
+.Fa "const struct man *man"
+.Fc
.Ft "const struct man_node *"
.Fo man_node
.Fa "const struct man *man"
@@ -100,6 +107,14 @@
.Fo mparse_free
.Fa "struct mparse *parse"
.Fc
+.Ft void
+.Fo mparse_getkeep
+.Fa "const struct mparse *parse"
+.Fc
+.Ft void
+.Fo mparse_keep
+.Fa "struct mparse *parse"
+.Fc
.Ft "enum mandoclevel"
.Fo mparse_readfd
.Fa "struct mparse *parse"
@@ -244,6 +259,8 @@
Obtain the meta-data of a successful parse.
This may only be used on a pointer returned by
.Fn mparse_result .
+.It Fn man_mparse
+Get the parser used for the current output.
.It Fn man_node
Obtain the root node of a successful parse.
This may only be used on a pointer returned by
@@ -294,6 +311,15 @@
.It Fn mparse_free
Free all memory allocated by
.Fn mparse_alloc .
+.It Fn mparse_getkeep
+Acquire the keep buffer.
+Must follow a call of
+.Fn mparse_keep .
+.It Fn mparse_keep
+Instruct the parser to retain a copy of its parsed input.
+This can be acquired with subsequent
+.Fn mparse_getkeep
+calls.
.It Fn mparse_readfd
Parse a file or file descriptor.
If
Index: mandoc.h
===================================================================
RCS file: /usr/vhosts/mdocml.bsd.lv/cvs/mdocml/mandoc.h,v
retrieving revision 1.95
diff -u -r1.95 mandoc.h
--- mandoc.h 23 Jul 2011 18:44:42 -0000 1.95
+++ mandoc.h 5 Oct 2011 09:34:15 -0000
@@ -400,30 +400,31 @@
__BEGIN_DECLS
-void mparse_free(struct mparse *);
-void mparse_reset(struct mparse *);
-struct mparse *mparse_alloc(enum mparset,
- enum mandoclevel, mandocmsg, void *);
-enum mandoclevel mparse_readfd(struct mparse *, int, const char *);
-void mparse_result(struct mparse *, struct mdoc **, struct man **);
-const char *mparse_strerror(enum mandocerr);
-const char *mparse_strlevel(enum mandoclevel);
-
void *mandoc_calloc(size_t, size_t);
+enum mandoc_esc mandoc_escape(const char **, const char **, int *);
void *mandoc_malloc(size_t);
void *mandoc_realloc(void *, size_t);
char *mandoc_strdup(const char *);
char *mandoc_strndup(const char *, size_t);
-
-enum mandoc_esc mandoc_escape(const char **, const char **, int *);
-
struct mchars *mchars_alloc(void);
+void mchars_free(struct mchars *);
char mchars_num2char(const char *, size_t);
int mchars_num2uc(const char *, size_t);
-const char *mchars_spec2str(struct mchars *, const char *, size_t, size_t *);
-int mchars_spec2cp(struct mchars *, const char *, size_t);
-void mchars_free(struct mchars *);
-
+int mchars_spec2cp(struct mchars *,
+ const char *, size_t);
+const char *mchars_spec2str(struct mchars *,
+ const char *, size_t, size_t *);
+struct mparse *mparse_alloc(enum mparset,
+ enum mandoclevel, mandocmsg, void *);
+void mparse_free(struct mparse *);
+void mparse_keep(struct mparse *);
+enum mandoclevel mparse_readfd(struct mparse *, int, const char *);
+void mparse_reset(struct mparse *);
+void mparse_result(struct mparse *,
+ struct mdoc **, struct man **);
+const char *mparse_getkeep(const struct mparse *);
+const char *mparse_strerror(enum mandocerr);
+const char *mparse_strlevel(enum mandoclevel);
__END_DECLS
Index: mdoc_man.c
===================================================================
RCS file: /usr/vhosts/mdocml.bsd.lv/cvs/mdocml/mdoc_man.c,v
retrieving revision 1.3
diff -u -r1.3 mdoc_man.c
--- mdoc_man.c 30 Sep 2011 00:13:28 -0000 1.3
+++ mdoc_man.c 5 Oct 2011 09:34:15 -0000
@@ -18,6 +18,7 @@
#include <string.h>
#include "mandoc.h"
+#include "man.h"
#include "mdoc.h"
#include "main.h"
@@ -216,6 +217,13 @@
break;
}
}
+}
+
+void
+man_man(void *arg, const struct man *man)
+{
+
+ fputs(mparse_getkeep(man_mparse(man)), stdout);
}
void
Index: read.c
===================================================================
RCS file: /usr/vhosts/mdocml.bsd.lv/cvs/mdocml/read.c,v
retrieving revision 1.23
diff -u -r1.23 read.c
--- read.c 23 Jul 2011 18:41:18 -0000 1.23
+++ read.c 5 Oct 2011 09:34:16 -0000
@@ -63,6 +63,7 @@
mandocmsg mmsg; /* warning/error message handler */
void *arg; /* argument to mmsg */
const char *file;
+ struct buf *secondary;
};
static void resize_buf(struct buf *, size_t);
@@ -411,6 +412,27 @@
of = 0;
+ /*
+ * Maintain a lookaside buffer of all parsed lines. We
+ * only do this if mparse_keep() has been invoked (the
+ * buffer may be accessed with mparse_getkeep()).
+ */
+
+ if (curp->secondary) {
+ curp->secondary->buf =
+ mandoc_realloc
+ (curp->secondary->buf,
+ curp->secondary->sz + pos + 2);
+ memcpy(curp->secondary->buf +
+ curp->secondary->sz,
+ ln.buf, pos);
+ curp->secondary->sz += pos;
+ curp->secondary->buf
+ [curp->secondary->sz] = '\n';
+ curp->secondary->sz++;
+ curp->secondary->buf
+ [curp->secondary->sz] = '\0';
+ }
rerun:
rr = roff_parseln
(curp->roff, curp->line,
@@ -437,6 +459,12 @@
assert(MANDOCLEVEL_FATAL <= curp->file_status);
break;
case (ROFF_SO):
+ /*
+ * We remove `so' clauses from our lookaside
+ * buffer because we're going to descend into
+ * the file recursively.
+ */
+ curp->secondary->sz -= pos + 1;
mparse_readfd_r(curp, -1, ln.buf + of, 1);
if (MANDOCLEVEL_FATAL <= curp->file_status)
break;
@@ -704,6 +732,8 @@
mdoc_reset(curp->mdoc);
if (curp->man)
man_reset(curp->man);
+ if (curp->secondary)
+ curp->secondary->sz = 0;
curp->file_status = MANDOCLEVEL_OK;
curp->mdoc = NULL;
@@ -720,7 +750,10 @@
man_free(curp->pman);
if (curp->roff)
roff_free(curp->roff);
+ if (curp->secondary)
+ free(curp->secondary->buf);
+ free(curp->secondary);
free(curp);
}
@@ -779,4 +812,20 @@
mparse_strlevel(enum mandoclevel lvl)
{
return(mandoclevels[lvl]);
+}
+
+void
+mparse_keep(struct mparse *p)
+{
+
+ assert(NULL == p->secondary);
+ p->secondary = mandoc_calloc(1, sizeof(struct buf));
+}
+
+const char *
+mparse_getkeep(const struct mparse *p)
+{
+
+ assert(p->secondary);
+ return(p->secondary->sz ? p->secondary->buf : NULL);
}
^ permalink raw reply [flat|nested] 6+ messages in thread