Roman, if you can implement it within a couple of months, it will be great. I also think this feature is big enough to warrant a new branch. This way we can all do periodic reviews and provide comments.
D. On Wed, Mar 23, 2016 at 6:13 PM, Roman Shtykh <rsht...@yahoo.com.invalid> wrote: > Yakov, > Thanks for the pointers, that's exactly what I needed. > RESP (REdis Serialization Protocol) is not that complex, but depending on > my other tasks it may take a month or two. I plan to implement the > integration iteratively starting with the most used generic commands.If you > need it earlier, someone else can pick the ticket, or we can create a > branch to collaborate. > -Roman > > > On Thursday, March 24, 2016 12:44 AM, Yakov Zhdanov < > yzhda...@apache.org> wrote: > > > Roman, > > I like the idea! How complex is the protocol implementation? How long will > it take to implement? > > As far as I remember TcpRestParser switches the correct mode on as soon as > it receives the first message over the connection. You can try adding > Redis mode to it. Is that what you ask for? > > > > > --Yakov > > 2016-03-23 11:53 GMT+03:00 Roman Shtykh <rsht...@yahoo.com.invalid>: > > > Igniters, > > I was thinking on enabling Redis protocol with Ignite [1] so that any > > Redis client can use Ignite transparently. > > Currently I have two ideas. > > 1. Implement it as a core functionality by adding GridRedisProtocol and > > appropriate commands/handlers, similarly to how it is done with REST and > > memcached, and use TCP binary protocol of the current implementation. For > > this I will probably have to do some refactoring in > > ignite/internal/processors/rest by creating > ignite/internal/processors/ext > > to have other external protocols, like > > ignite/internal/processors/ext `-- rest `-- memcached `-- redis > > 2. Implement it as an external module that is started on a local machine > > in a client mode by a user, interpreting Redis commands and sending to > the > > Ignite cluster. > > To be consistent with REST and memcached implementations, I prefer the > 1st > > approach. Any objections or advice? > > Another question -- at first glance memcached protocol is implemented as > a > > part of REST implementation. For instance, GridTcpRestParser is > essentially > > a memcached parser. Should it be refactored? > > [1] https://issues.apache.org/jira/browse/IGNITE-2788 > > > > > >
