Hi Marcel
> Hi Mohamed,
>
> > > > Add new API to support http download.
> > > > ---
> > > > gweb/gweb.h | 12 ++++++++++++
> > > > 1 files changed, 12 insertions(+), 0 deletions(-)
> > > >
> > > > diff --git a/gweb/gweb.h b/gweb/gweb.h
> > > > index 1ab2a9f..61e488c 100644
> > > > --- a/gweb/gweb.h
> > > > +++ b/gweb/gweb.h
> > > > @@ -38,10 +38,19 @@ typedef enum {
> > > > G_WEB_METHOD_GET,
> > > > } GWebMethod;
> > > >
> > > > +typedef enum {
> > > > + G_WEB_DATA_HEADER,
> > > > + G_WEB_DATA_BODY,
> > > > +} GWebDataType;
> > > > +
> > > > typedef void (*GWebResultFunc)(uint16_t status, gpointer user_data);
> > > >
> > > > typedef void (*GWebDebugFunc)(const char *str, gpointer user_data);
> > > >
> > > > +typedef gboolean (*GWebReceivedFunc)(const guint8 *str, int len,
> > > > + GWebDataType data_type,
> > > > + gpointer user_data);
> > > > +
> > >
> > > so I am thinking this becomes a little bit too complicated and we are
> > > overloading this. Here is my idea on how to make this simpler.
> > >
> > > struct _GWebResult;
> > >
> > > typedef struct _GWebResult GWebResult;
> > >
> > >
> > > typedef gboolean (*GWebResultFunc) (guint16 status, GWebResult *result,
> > > gpointer user_data);
> > >
> > > guint g_web_request(GWeb, GWebMethod method, const char *url,
> > > GWebResultFunc, gpointer user_data);
> > >
> > > So essentially we are creating a new GWebResult object that represents
> > > our transaction/request.
> > >
> > > Then we can have these functions to be used inside the result callback.
> > >
> > > gboolean g_web_result_get_header(GWebResult *result,
> > > const char *header, const char **value);
> > >
> > > gboolean g_web_result_get_fragment(GWebResult *result,
> > > const guint8 **fragment, gsize *length);
> > >
> > > The get_header function will be always valid since we keep all headers
> > > in the GWebResult object for the time of the transaction. And the
> > > get_fragment returns the current buffer value.
> > >
> > > And of course the result function has a return value. If returning FALSE
> > > then we just stop the transaction right away. So after we received the
> > > headers we can call this once with empty fragment buffer and when we
> > > have all details we just stop it. If we return TRUE then the transaction
> > > continues until we have received the full body.
> > >
> > > Also the status will indicate if we are done or not. A status of 100
> > > indicates that there is more to come. And every other status is either
> > > 200 for success or an error code. Of course the return value only
> > > matters if the status is 100 since otherwise the transaction is over
> > > anyway.
> >
> > So now we have one notification function which will be used by GWeb to
> > notify the user. So I guess GWeb will call GWebResultFunc evey time we
> > receive a payload right?
> > This new API will be more flixible to user since they can get the
> > payload on chunk or wait and get the whole payload at once but it will
> > be more work in GWeb since we have to buffer until user ask for data.
>
> no buffering inside the GWebResult. If you don't get the fragment buffer
> in that callback, then that buffer is gone.
>
> If you don't want the website, then don't bother requesting it in the
> first place ;)
>
> > One more question in gboolean g_web_result_get_fragment(GWebResult
> > *result, const guint8 **fragment, gsize *length);
> > I assume the user will responsible to release the memory here right?
>
> It is const. So the buffer is owned by GWebResult. The user can not free
> it. That is what const means. You get a pointer and can use that
> pointer, but it is not yours.
>
> So basically GWebResult needs to have an internal buffer where it stores
> data in from the HTTP transaction. And when requested it just returns a
> pointer to that data.
>
> > I will start on adding this change and please let me know if any of my
> > assumption are wrong so I dont have to redo.
>
> Essentially you have all of this already. Just the header handling is a
> bit different. So you have parse all headers and store them. Using a
> GHastTable internally inside GWebResult looks like a good idea. And once
> you have all headers, then you call into the callback for the first
> time.
Thanks for the response and clearing that out. I will email back with
modified patch.
Mohamed
_______________________________________________
connman mailing list
[email protected]
http://lists.connman.net/listinfo/connman
