[email from 2000-08-17 01:33:37 CDT (Aug Thu) 966494017]

Hi,

For the web, you can write a certain amount of CGI code in /bin/sh (or any
of the other interpreters that can at least handle CGI's somewhat strange
input methods), but the workhorse is typically PERL, and most examples will
be in that language.  C will work just as well as PERL, however, if you
don't know the other yet.

Note that the HTML output you see in a CGI-generated webpage is -not- the
PERL or C or sh code, just the script's HTML output.

If you want to do C-based CGI, then just write a program that coughs up the
correct initial two lines giving the document type, and use puts() or the
like to output HTML.  Play with that for a while (it lets you do a whole
lot, right there), then find out how the GET and POST actions send their
data.  I seem to remember that one uses environment variables mostly, so
getenv() would be a key function, and the other provides and input stream
on stdin of the data from the form, which can be read into a
dynamically-allocated buffer with even just getchar().

After that, parsing the input string is a little weird, since you usually
want to put the results into a dynamically-allocated vector (remember malloc?)
of structs, each with two strings for a name/value pair.

Write a function to scan these vector for a named string, and the rest
is easy again.

So there.  :-)

----------------------------------------------------------------------------
 C. Alexander North-Keys                http://www.talisman.org/~erlkonig/
------------------------------------------ ^*^ -----------------------------

[when questioned about the following....]
| After that, parsing the input string is a little weird, since you usually
| want to put the results (got that part) into a"dynamically-allocated vector
| (remember malloc?)  of structs, each with two strings for a name/value
| pair."
| 
| Write a function to scan these vector for a named string, and the rest is
| easy again.

i.e.

typedef struct _NameValuePair
{
	char *name;
	char *value;
} NameValuePair;

NameValuePair *pairs;

- followed by various functions to fiddle with these, eventually allowing
  you to do stuff like:

char *login = 0, *passwd = 0;

if(   (login = NvpLookup(pairs, "login"))
   && (passwd = NvpLookup(pairs, "password")))
{
   puts("<h1>Login and Password read</h1>");
   printf("<dl>\n<dt>login:<dd>%s<dt>password<dd>%s\n", login, password);
} else {
   puts("<h1>Login and/or Password missing</h1>");
}

- you could also build functions around a linked list:

typedef struct _NameValuePair
{
	struct _NameValuePair *next;  /* yes, the _Thing is finally non-optional */
	char *name;
	char *value;
} NameValuePair;

with functions like

NameValuePair *NvpNew(char *name, char *value);
  /* returns address of new Nvp, with strdup()ed copies of name and value */
NameValuePair *NvpPush(NameValuePair *list, NameValuePair *addme);
  /* returns longer list, new element first */
const char *NvpSearchByName(NameValuePair *list, const char *name);
  /* returns pointer to value, or null */       

sample implementations:

/* free full or partially-complete obj */
NameValuePair *NvpDelete(NameValuePair *doomed)
{
	if(doomed) {
		if(doomed->name)  free(doomed->name);
		if(doomed->value) free(doomed->value);
		free(doomed);
	}
	return 0;
}

NameValuePair *NvpNew(char *name, char *value)
{
	NameValuePair *created = (NameValuePair*)malloc(sizeof(NameValuePair));
	if(created
	   && (created->name = strdup(name))
	   && (created->value = strdup(value)))
	{
		/* yay, it worked */
	} else {
		perror("NvpNew - malloc/strdup");
	    created = NvpDelete(created);
	}
	return created;
}

----------------------------------------------------------------------------
 C. Alexander North-Keys                http://www.talisman.org/~erlkonig/
 Talisman Research / Unix Analyst                   erlkonig@talisman.org
------------------------------------------ ^*^ -----------------------------