There is not a way to do this. Should there be? Could there be? The relevant code is here:
https://github.com/racket/web-server/blob/master/web-server-lib/web-server/http/request.rkt#L52 and https://github.com/racket/web-server/blob/master/web-server-lib/web-server/http/request.rkt#L219 It would be challenging to do this. Suppose that a request struct contained the input port (the obvious thing to do.) The problem is that the servlet MUST consume it all before the Web server can look for the next HTTP request (because in 1.1 there's multiple per connection.) So what if the servlet doesn't? Do we detect when it loses the reference and then slurp it up and keep going on the connection? It is currently allowed to store a request in a global data-structure for later, for instance. Let's suppose we ignore that problem and only enable this when #:connection-close? is true and there's only one request per connection... then there's a backwards compatibility problem because a request already contains the raw-bytes. We could delay them, but getting them will require having a copy of the result of reading the input port, which is exactly what you are trying to avoid. I think we're stuck as far as the Web server's defaults go (but maybe you can think of something.) I think the best thing to do is to add the right knob to make-read-request and make it easy to configure a special dispatcher that does what you want and use URI case matching to get into the right code path for you. (I.e. Have your server default to the new behavior and then switch to the old behavior when you are going to do a normal non-streaming read before handing the request to your servlet.) Jay On Fri, Sep 1, 2017 at 10:57 PM, George Neuner <gneun...@comcast.net> wrote: > On Fri, 1 Sep 2017 14:38:21 -0400, Neil Van Dyke > <n...@neilvandyke.org> wrote: > >> ... *except when a GC cycle kicks in*. > > Speaking of web servers and GC ... > > > I have a http web-server application that needs to be able to up/down > load fairly large files. The application is somewhat memory > constrained, and I'd like to handle these large data transfers in > chunks using (relatively) small buffers. > > Outgoing is not [much of] a problem because the client TCP connection > is available in response/output ... but with incoming data, the > web-server framework has slurped in everything before my code even > runs. > > Is there a way in the web-server to stream incoming data? Or maybe to > get at the request before the (multipart) form data is read in? > > George > > -- > You received this message because you are subscribed to the Google Groups > "Racket Users" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to racket-users+unsubscr...@googlegroups.com. > For more options, visit https://groups.google.com/d/optout. -- -=[ Jay McCarthy http://jeapostrophe.github.io ]=- -=[ Associate Professor PLT @ CS @ UMass Lowell ]=- -=[ Moses 1:33: And worlds without number have I created; ]=- -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
