On Saturday 02 April 2005 01:17, Johannes Stezenbach wrote: > On Sat, Apr 02, 2005 at 12:28:20AM +0200, Kenneth Aafl?y wrote: > > On Saturday 02 April 2005 00:13, Johannes Stezenbach wrote: > > > Please don't use typedefs unless you are explicitly defining > > > opaque data types. It makes it more difficult to see what's going on. > > > > I'm wondering if you recentment is against the typedef in general or the > > implementation of a clean api. > > I'm neither agains typedefs, where they make sense, nor against clean APIs. :) > > The api will be well documented, and so > > will the new typedefed types? I was merily following the alsa api here, > > so I'm not in any way married to it yet! Any suggestions is very much > > welcome, please prove why your suggestion will produce a cleaner api. > > Typedefs have a purpose: Creating an abstraction, i.e. hiding > information about the type from the user. IMHO it is totally useless to > stick a typedef on every struct or enum definition. All you get is > obfuscation. And the _t suffixes are ugly, too. > Please ask yourself if and why you want to hide from the users > of your API why e.g. vid_dev_hierarchy_t is an enum. I actually agree with you on this, although I feelt like using the interface that alsa used for a while. We have to use typedefs on the abstract types that does not have a real representation in the user application though, as you say. And now that I think back at the getting into the alsa api, I was actually stuck for quite a while when I tried to figure out the details. > > int vid_channel_set_param(vid_channel_param_t, int value); > > int vid_channel_get_param(vid_channel_param_t); > > I would prefer to use a struct, or rather a union of structs for > each transmission system. You'd get nuts when you need a dozen > function calls to tune or update the channel list. Do you have a neat method of checking what struct was used in the above single entry call without to much fuzz? Maybe a enum in concert with a struct? Kenneth
