Re: [PATCH] config: rework code to use GKeyFile instead of GHashTable+GValue

[Lionel Landwerlin <lionel g landwerlin linux intel com>]

On Tue, 2011-01-18 at 14:58 +0000, Iago Toral wrote:
> Hi,
> 
>  there is still one minor problem, which was partially present in the 
>  previous implementation:
> 
>  if the client requests the value for a configuration parameter that is 
>  not present in the configuration, it gets a default value and no hint 
>  that the parameter did not exist. This was already the case in the 
>  previous code, however, when in need to know if the parameter actually 
>  existed, the client could use grl_config_get, which would return a 
>  GValue (if the parameter existed) or NULL (otherwise). This was used in 
>  the filesystem plugin to check the maximum depth allowed for search 
>  operations. In this plugin, if this limit is not set through a config 
>  parameter, a sane default is used by the plugin, but with this patch 
>  this situation can no longer be detected, since grl_config_get does not 
>  exist any more.
> 
>  My proposal to fix this situation is to add one more parameter to the 
>  'get' APIs. The new parameter would be a pointer to a location where the 
>  result should be stored, and the function result would be a boolean to 
>  let clients know if the parameter did exist in the configuration.
> 
>  So, for example, we would have this:
>  gboolean grl_config_get_string (GrlConfig *config, const gchar *param, 
>  gchar **value);
> 
>  Instead of this:
>  gchar *grl_config_get_string (GrlConfig *config, const gchar *param);
> 
>  Sounds good?
> 

GKeyFile has g_key_file_has_key() function, which we could wrap in a
grl_config_has_key/value/config function.
We could also use the GError framework as GKeyFile does.

Regards,

--
Lionel Landwerlin