un-top-posting your reply... Le mercredi 08 octobre 2014 à 12:45 -0700, Jim Nelson a écrit : > On Wed, Oct 8, 2014 at 12:14 PM, Jérémy Lal <[email protected]> wrote: > > Le mercredi 08 octobre 2014 à 16:04 -0300, Gustavo Noronha Silva a > > écrit : > >> On Dom, 2014-08-31 at 23:31 +0200, Jérémy Lal wrote: > >> > webkit_web_resource_get_data is used to get a response data, but > >> in case > >> > of big files, video for example, it can easily eats a lot of > >> memory. > >> > The "received-data" signal would be a very nice place to receive > >> the > >> > data chunks as they are downloaded. Is this possible or does it > >> goes > >> > against webkit architecture ? > >> > >> Hmm. I think it could be. Perhaps what we need is a way of getting > >> an > >> InputStream out of WebResource, so you could use the regular read > >> operations to fetch the contents, how about that? > > > > Looks good to me... i'll happily test it to stream the data to my > > node-webkitgtk module :) > > Shotwell ran into this problem with Soup and multipart uploads some > time back. Dan Winship pointed out that memory-mapped I/O would > solve that problem, and I suspect it would work here too.
Yes, that answers part of the problem, (if it works as expected), but being able to access the response stream would enable a lot of other applications... Processing data as soon as it is received (not having to wait for the whole file to be done loading) is a huge advantage. Also i fear bindings to webkit in other languages might have to make a copy at some point of the data, and it might as well do that on smaller chunks than on one big file. Jérémy. _______________________________________________ webkit-gtk mailing list [email protected] https://lists.webkit.org/mailman/listinfo/webkit-gtk
