I can embed an nrepl-server in a driver program and submit that jar and
then connect the nrepl-server to my emacs cider and work with it. I have
just never done it. May be it is time to understand how instead of just
doing
M-x cider-jack-in
Thanks Asim for pointing in the right direction.
It was too simple to setup .. I just had not spent time .. and also with
respect to running the spark-app in yarn-client mode and connecting to it
was a good suggestion. Thank you! Sunil.
On Tue, Nov 3, 2015 at 7:46 PM, Asim Jalis wrote:
> Hi Sunil,
>
> Yes. This should
Hi Sunil,
Yes. This should work. I have used something much simpler in the past such
as server-socket. This starts listening on a specific port and then you can
telnet to that port from your client and start telnet with rlwrap to get
history and editing.
Asim
On Tue, Nov 3, 2015 at 12:00 AM,
Thanks Asim for this suggestion. I wanted to do it but did not have the
bandwidth to explore how to do it.. I am new to spark.. I usually do
repl-driven dev for local apps... Have you done repl-driven using spark and
clojure? do you have pointers as to how to do it.. I can explore if you
tell me
Hi Sunil,
Here is something to think about. Can you use the REPL for your workflow
instead?
If the changes are all in the Spark driver code you can run the driver
inside the REPL and then test out the changes in the REPL. Once you have
something that works you can save that and deploy an
Hi Sunil,
I am wondering why you want to speed up lein uberjar.
How frequently do you need to run this?
What is your use case and workflow around this?
Thanks.
Asim
On Mon, Oct 26, 2015 at 2:04 AM, Sunil S Nandihalli <
sunil.nandiha...@gmail.com> wrote:
> Hi Everybody,
> My lein uberjar
Hi Asim,
I needed to run lein uberjar to submit the jar to run a distributed job on
a apache-spark cluster. and my uberjar is currently taking anywhere from 2
to 4 minutes to finish. I have have had to run uberjar fairly frequently
during my iterations ( 4 to 5 times in an hour) . I am new to
I think Matching Socks has definitely got the right answer -- anything that
prevents your code from being AOT-compiled is going to give you a huge
speedup.
On Sun, Nov 1, 2015 at 10:11 PM Sunil S Nandihalli <
sunil.nandiha...@gmail.com> wrote:
> Hi Asim,
> I needed to run lein uberjar to submit
Hi,
I think the main reason of it is due to the fact that aot compilation is
transitive and compiles all dependent namespaces.
You will have to wait for this issue
http://dev.clojure.org/jira/browse/CLJ-322 to get some closure :)
On Monday,
October 26, 2015 at 2:35:19 PM UTC+5:30, Sunil
If you do not want much to be compiled ahead-of-time and represented in the
jar as class files, you may use the :impl-ns option on :gen-class.
e.g.,
1) project.clj says
:aot [bla.bla-aot]
2) bla.bla_aot.clj says
(ns bla.bla-aot (:gen-class :impl-ns bla.bla))
Ahead-of-time compilation does
Recommend to use libdir plugin: https://github.com/djpowell/lein-libdir
Run lein libdir to copy all the dependencies to lib directory, and just
run lein jar to package the project.
Then you can rsync or copy the project jar with all dependent jars instead
of compiling all dependent namespaces
Hi Everybody,
My lein uberjar takes about 2 minutes to run. It also recompiles all the
clj files everytime. Is there a way to speedup lein-uberjar .. Or may be
make it incremental?
Thanks,
Sunil.
--
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To
12 matches
Mail list logo