[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Avoiding svn_iter_apr_hash/array() [was: svn commit: r935579]

From: Hyrum K. Wright <hyrum_wright_at_mail.utexas.edu>
Date: Tue, 20 Apr 2010 10:46:30 -0500

On Tue, Apr 20, 2010 at 10:34 AM, Julian Foad <julian.foad_at_wandisco.com>wrote:

> Daniel Shahaf wrote:
> > Julian Foad wrote on Tue, 20 Apr 2010 at 11:49 +0100:
> > > Hmm... I wonder if we could do something with a macro that takes the
> > > whole block of code to be executed as an argument, like:
> >
> > A code block inside a macro call??
>
> Well, I tried it and it works and I kind of like it... but really I
> tried it for fun and I'm not sure if we should be seriously considering
> anything like this.
>
> > Perhaps we write a macro in the
> > style of SVN_DBG --- one that is "pseudo-variadic"...
> >
> > For example (not tested):
> >
> > #define SVN_ITER_ARRAY2(element_type, element_name, array, iterpool,
> pool, func_name, (arg1, arg2)) \
> > SVN_ITER_ARRAY(element_type, element_name, array, iterpool, pool,
> func_name(arg1, arg2))
>
> I think you mean something like (with a backslash after every line):
>
> #define SVN_ITER_ARRAY2(element_type, element_name, array,
> iterpool, pool,
> func_name, arg_list)
> SVN_ITER_ARRAY(element_type, element_name, array,
> iterpool, pool,
> {
> svn_error_t *__err = func_name arg_list;
> if (__err->apr_err == SVN_ERR_ITER_BREAK)
> {
> svn_error_clear(err);
> break;
> }
> SVN_ERR(__err);
> }
>
> And I think the advantages of your SVN_ITER_ARRAY2 are:
>
> * not taking a code block as an argument, thus avoiding the surprise
> (and possible confusion of syntax-highlighting editors etc.);
>
> and its disadvantages are:
>
> * the body of the loop has to be expressed as a single function call,
> which adds a considerable overhead for simple operations.
>
> - Julian
>
>
> > where SVN_ITER_ARRAY() is the macro you defined:
> >
> > > #define SVN_ITER_ARRAY(element_type, element_name, array, \
> > > iterpool_name, /* as subpool of */ pool, \
> > > code_block) \
> > > { apr_pool_t *iterpool_name = ... \
> > > for (...) \
> > > { \
> > > element_type element_name = ...; \
> > > apr_pool_clear(iterpool_name); \
> > > code_block \
> > > } \
> > > apr_pool_destroy(iterpool_name); \
> > > }
> > >
> >
> > :-)
> >
> > Daniel
> >
> > > Usage:
> > > SVN_ITER_ARRAY(svn_node_t *, node, nodes_array,
> > > iterpool, existing_pool,
> > > {
> > > if (node->kind == svn_node_file)
> > > SVN_ERR(bar(node, blah, iterpool));
> > > else
> > > break; /* breaks the iteration */
> > > } ) /* macro ends here */
> > >
> > > Hmm...
> > >
> > > - Julian
>

While an interesting exercise in extending the C language, I've got serious
reservations about introducing these types of macros. They don't improve
code clarity, and make it more difficult for a novice to the code to
understand what is going on. Please think long and hard before adding these
obfuscating "features" to our code base.

-Hyrum
Received on 2010-04-20 17:47:04 CEST

This is an archived mail posted to the Subversion Dev mailing list.