On 13/05/2015 14:19, Rémy Maucherat wrote: > 2015-05-13 0:29 GMT+02:00 <ma...@apache.org>: > >> Author: markt >> Date: Tue May 12 22:29:45 2015 >> New Revision: 1679116 >> >> URL: http://svn.apache.org/r1679116 >> Log: >> Start to process the initial settings frame sent by the client. >> + private boolean processFrame() throws IOException { >> + // TODO: Consider refactoring and making this a field to reduce >> GC. >> + byte[] frameHeader = new byte[9]; >> + if (!getFrameHeader(frameHeader)) { >> + return false; >> + } >> + >> + int frameType = getFrameType(frameHeader); >> + int streamId = getStreamIdentifier(frameHeader); >> + int payloadSize = getPayloadSize(streamId, frameHeader); >> + >> + switch (frameType) { >> + case FRAME_SETTINGS: >> + processFrameSettings(streamId, payloadSize); >> + break; >> + default: >> + // Unknown frame type. >> + processFrameUnknown(streamId, frameType, payloadSize); >> + } >> return false; >> > > That's really a great example on where you can use the new IO calls I > proposed: they can replace with a single async call multiple reads (incl > blocking ones) that are needed to process the frame. > > Comments ?
I'm not sure reading into a 9 byte array is such a good example. I suspect there will be better ones in the header and data frames. I find it easier to see how gathering writes would be useful. Mark --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
