MESSAGE
| DATE | 2004-09-10 |
| FROM | Billy
|
| SUBJECT | Re: [hangout] Re: two dimensional arrays passed to functions
|
On Fri, Sep 10, 2004 at 12:50:05PM -0400, Ruben Safir Secretary NYLXS wrote:
>
> > Functions should be kept clean enough that it isn't a problem. So
> > abusing goto is very bad, but so is abusing any other feature. I
> > know that when I use it, it's as a sort of last resort when the
> > alternatives would be worse (read: more bug-prone).
> >
>
> Like longjump and short jump for errors?
You're thinking of setjmp/longjmp ? Those are hard to deal with, but
they can be used to implement a primitive exception throwing facility.
> >> In addition, there is always signals, which is how I originally
> >> handled errors. The thing is here, the 'error' is not necessarily
> >> an error. It could be a normal response from the database, like an
> >> attempt to insert already entered data on a unique key. In which
> >> case, I want to switch over and do an UPDATE instead.
> >
> > Sound thinking. You're talking about gtk signals or Unix signals?
>
> I meant unix signals, singnals.h I think in the C library. Also,
> consider GTK, in their obsession to force an Object-like notation onto
> everything, one of the benifits is that instead of capturing the whole
> of a structured program in if statements, you should be able to emit a
> GTk signal to the top level window, or carry the error handling up the
> tree, which already exists, without the need for break break break.
Right, but you have to be careful about carelessly exiting a function in
which you've allocated resources.
f() {
char *p = malloc(1);
if( risky_function() == SCREWUP ){
/* congratulations!
* You've got a memory leak! */
return;
}
free(p);
return;
}
or:
f() {
int fd = open("/tmp/foo", O_RDONLY);
if( risky_function() == SCREWUP ){
/* congratulations!
* You've got a stale file descriptor! */
return;
}
close(fd);
return;
}
> >> > You were doing a return if anything went wrong, and you never
> >> > cleaned up the MYSQL session, so I suppose you didn't see a need
> >> > for the goto before.
> >>
> >> Yeah, it is not apparent yet from the code, but the MYSQL struct
> >> instance, there is a flag called reconnect which I want to check
> >> for. I believe if it is 1, then you need to reconnect. I actually
> >> want to make that data structure global and to cache the database
> >> connection, eventually closing it in the destructor.
> >
> > Make a big struct (of which the MYSQL connection is one member) and
> > pass it around everywhere. Much cleaner than using globals...
>
> That's a thought but the last time I did that I had to malloc the
> structure and I really don't know where to free it since it is being
> used in all the lower Widgets.
You have to decide early on. EVERYTHING has a lifecycle.
destruct at the end of main() if you have to, or in the handler
for Gtk's application_close event or whatever they call it.
> Otherwise it disapears from the tact.
> That's why, just for exepedency, I created those dreadful global
> structures to store data. You can't just create a structure and then
> passing to the callback function at g_pointer a because it disapears
> from the darn stack. It took me two days to figure that out this
> time, and a week last time (although RMS figured it out in about 15
> seconds in the debugger.)
I think it's a pretty common beginner mistake.
> The wild factor here is that gtk_main() is actually interceeding in
> everything here. Roughly speaking, when you click on a button, your
> talking to gtk_main() which 'senses' the clicked event and then looks
> for the proper callback and then runs it. In this case at least, it
> is a darn mine field.
Why is it a mine field?
Loosely coupled objects can make for difficult debugging.
It's one of the headaches of abstraction.
> >> I don't know what mysql_init(&MYSQL) does, but it does not allocate
> >> any memory.
> >
> >One thing it does is allocates space for a MYSQL if you pass it a
> >NULL.
>
> Say again? Where did you read that?
http://dev.mysql.com/doc/mysql/en/mysql_init.html
> Then why did I have to make sure the memory was allowed for it?
There are 2 ways to call it.
{
MYSQL db_conn_storage;
MYSQL *db_conn = mysql_init(&db_conn_storage);
...
mysql_close(db_conn);
}
Or if you don't need to manage your own memory:
{
MYSQL *db_conn = mysql_init(0);
...
mysql_close(db_conn);
}
In the second form, the memory pointed to by db_conn
will be free()'ed by mysql_close.
> What the heck does mysql_init() do then. It completely didn't work
> otherwise. I'm totally confused now.
I don't know.. Presumably it sets all of the fields of the MYSQL struct
to consistent values.
> >> That is the reason for the assignment of the pointer. I hope I'm
> >> missing something of the MYSQL API, because otherwise I might end
> >> up creating a global flagging system in order to cache the database
> >> connection.
> >
> >Go for it. But not global define a struct. You'll be happy you did.
> I need some rest. I feel like I'm back in College again. I've
> crammed for finals relearning 2 dimensional arrays, array function
> parameters, the string library, the mysql C API, the GTK C API, patch,
> diff, vi advanced features gdb, and C macros all in the last 48 hours,
> in addition to trying to remain with a sane program design and boning
> up on integration for Biopharmacuetics all in the last 48 hours.
> SOMEBODY better be giving me a masters degree when this is done.
Sounds like you're learning a LOT. It's good for you.
____________________________
NYLXS: New Yorker Free Software Users Scene
Fair Use -
because it's either fair use or useless....
NYLXS is a trademark of NYLXS, Inc
|
|
