I would extract compute tasks into separate scope. It is better to keep focus on protocol things and basic language support for now. Once we have basic client API in production-ready state, we could consider adding JavaScript to our core compute feature set and then extend it to the clients (which should be trivial provided that core part is ready). We should be ready to spend considerable efforts to prior R&D because dynamic code execution is not very simple thing, especially in terms of security, native compilation, etc..
On Thu, Mar 1, 2018 at 5:17 PM, Ilya Kasnacheev <[email protected]> wrote: > With regards of thin clients for dynamically typed languages, I think > Ignite needs two following features to shine: > > - Ability to pass JSON to such clients, turn JSON Objects into a > BinaryObjects, which will give ability to index top-level keys in such JSON > with SQL Indexing. Of course this should be integrated with QueryEntities. > - Ability to pass JavaScript snippets to invoke() and affinityCall() > families of calls. On Server node they should be interpreted by Nashorn > (since we've upgraded to Java 8). We should also cache such snippets > pre-interpreted, in this case it can be pretty fast since Nashorn compile > to JVM bytecode. > > WDYT? > > -- > Ilya Kasnacheev > > 2018-02-20 0:23 GMT+03:00 Alexey Kosenchuk <[email protected] > >: > > > Hi All! > > > > Let us join the party, please ;) > > > > As we see, there is Binary Client Protocol to communicate with Ignite > > cluster and a concept of Thin (lightweight) client. > > > > If there are no objections or duplicated plans, we would like to develop > > Thin Client libs for: > > - Node.js > > - Python > > - PHP > > > > Please add us as contributors and provide access to the Ignite Jira > > components. > > > > Usernames in the Apache Jira: > > alexey.kosenchuk > > ekaterina.vergizova > > pavel.petroshenko > > > > Thanks! > > -Alexey > > >
