gmane.comp.java.clojure.user

Re: Clojure on Android

blcooley — 2012-05-26T03:47:29


The work by Daniel Solano Gomez at https://github.com/sattvik/neko is the 
state of the art, so to speak. There's a start-up time penalty associated 
with loading the runtime, and you'll need Daniel's fork of clojure if you 
want to do anything dynamic. The readme in the repo is pretty good; you 
should check it out.

There's also a google group (clojure-android) which you may have found 
already, and Alex Yakushev has a blog 
at http://clojure-android.blogspot.com/ to cover his work for Google Summer 
of Code (and maybe beyond?).

You might want to check out Daniel's Clojure REPL in Google Play. You can 
use it to evaluate forms on the device, and it will give you some idea of 
the performance of dynamic compilation.

Best regards,
Brian Cooley

Re: maths, functions and solvers

jlk — 2012-05-26T02:25:22

Hi Zack

I don't really know enough about core.logic to comment on this.  Brent 
mentions symbolic math and it sounds neat but I'm more interested in 
numerical solutions at the moment.

Cheers


On Saturday, May 26, 2012 10:45:09 AM UTC+10, Brent Millare wrote:

Re: maths, functions and solvers

jlk — 2012-05-26T02:19:38

Hi Stephen

Firstly thanks for looking at the code - certainly given me a few tips and 
ideas there.  Also I'll look into monads when I get some spare brain time :)

I didn't realise that the argument list was in a var metadata.  Do you know 
if there is similar metadata for an undefined function (fn [] ...)?

Originally what I was thinking of was a kind of 'partial' function taking 
map arguments and that is where I'm heading.

Playing around this morning I think I've stumbled upon what I was after and 
the code is in the devel branch on github.  It's currently horrible, but 
seems to work.  Basically the maths function is provided as a list of 
symbols, then I do some substitution using clojure.walk and produce a 
function of argument whatever is still undefined.  Then I run this through 
eval.

Speaking of it would be nice if eval could directly take a map of symbols 
as an argument.

I'll tidy all this up when I get a chance.


Cheers

Re: core.logic and aggregates: am i doing it wrong?

Mark — 2012-05-26T01:35:09

Two scotches later, I came up with


and it seems to do the trick.  I'm not sure about leaving r unbound but my 
tests pass.  I'd love some feedback on the approach. 

On Friday, May 25, 2012 4:58:14 PM UTC-7, Mark wrote:

Re: maths, functions and solvers

Brent Millare — 2012-05-26T00:45:09

I originally assumed core.logic would be the key to symbolic math, but now 
I feel that's not really the case. You still need to implement all the 
algebra manipulation algorithms as usual. One can define the numbers in a 
different way as Ambrose did in his talk, but I imagine we would need to 
have much different performance expectations when performing the other 
symbolic operations. That being said, I would like to see more work going 
into that type of implementation.

On Thursday, May 24, 2012 8:37:57 PM UTC-4, Zack Maril wrote:

core.logic and aggregates: am i doing it wrong?

Mark — 2012-05-25T23:58:14

I confess that the last time I cracked open a Prolog book was lo these many 
years ago in college and that I need to re-read The Art but I'm hoping that 
someone can point out how I'm thinking about this problem wrongly:
I'd like to write a goal that succeeds when given a set of required items 
and a set of supplied items, the supplied items is a proper superset of the 
required items.  My intuition is that I want a goal that succeeds and 
'returns' the common subset of the two sets and then another goal that 
tries to unify the subset with the required items:


I'm basically treating the subset operation as a subrouting and using run*  
to aggregate the common elements.  This feels wrong and it feels like I 
should use something less procedural but I can't wrap my brain around how 
else to approach the problem.  Suggestions?  Hints?

Thanks

Re: Lambda: A lniux distro for clojurists

abaitam — 2012-05-25T23:54:16

Great idea. But this be implemented as a pallet or vagrant script instead 
of a ready-made VM? 

On Thursday, May 24, 2012 4:11:21 PM UTC-4, banseljaj wrote:

Re: Do you leave a Swank / nREPL in your production servers?

Sean Corfield — 2012-05-25T22:54:02

I was assuming that network access is pretty locked down on production
servers. For example, the port on which we listen is only open on an
internal IP and therefore only within a small, tightly controlled
subnet when we have VPN active. There's no external IP access open.


I agree that's a nice, secure alternative.

Re: Java Interop Interfaces

Softaddicts — 2012-05-25T22:18:48

Right, I forgot about the $ stuff to refer to internal classes.

Luc P.


--
Softaddicts<lprefontaine< at >softaddicts.ca> sent by ibisMail from my ipad!

Re: Do you leave a Swank / nREPL in your production servers?

Alan Malloy — 2012-05-25T21:56:04

This may be good advice in some cases (eg when all network access to
your server is "trusted"), but on a lot of production servers it
strikes me as very dangerous to apply this suggestion carelessly.
Assuming you have SSH access to the machine, by far the safest way to
do it is to have swank only listen on localhost, and then use SSH port
forwarding to redirect the connection through the localhost interface.
For example:

akm< at >dev-machine $ ssh -NfL 4010:localhost:4005 production-machine

# from emacs on dev-machine
M-x slime-connect RET RET 4010 RET

This way the only people who can connect remotely are those who have
permission to get a local application (usually sshd) to forward the
traffic to the localhost interface on their behalf.

Re: Lambda: A lniux distro for clojurists

Curtis Gagliardi — 2012-05-25T21:52:41

I'll have to check this out when I get home, it sounds like exactly
what I was looking for.  Thanks.

On May 25, 11:01 am, Tom Maynard <tom.w...< at >gmail.com> wrote:

Clojure on Android

HelmutKian — 2012-05-25T21:45:03

I've recently been on board a project that will targeting Android, and I've 
been given pretty much free reign chose my tools.  I'm an experienced 
Common Lisper and a fairly new Clojure user, but I'd like to know about 
the feasibility of writing real-world Android applications in Clojure.

Re: Do you leave a Swank / nREPL in your production servers?

Karl Krukow — 2012-05-25T21:34:57

Thanks Sean for the detailed information. This will be really useful to me.

/Karl
On 25/05/2012, at 23.30, Sean Corfield wrote:

Re: Do you leave a Swank / nREPL in your production servers?

Sean Corfield — 2012-05-25T21:30:35

Specifically:

Add a dependency on [swank-clojure "1.4.2"]

Then in your code:

(swank.swank/start-server :host "0.0.0.0" :port 4567) ;; use whatever
port you want, default is 4005

You can programmatically stop the server with:

(swank.swank/stop-server)

The :host specifies the IP (or hostname) to listen on so if you want
external access you'll need to listen on an IP that is externally
accessible. If you listen on "0.0.0.0" then it'll listen on any IP
address so you can REPL in from outside as well as directly on server
itself (on 127.0.0.1). Our servers have multiple IP addresses and this
caught me out at first when working with our data center folks and
setting up VPN/DMZ access to the host/port.

Another thing that was a bit of a problem for us was that swank.swank
does a bunch of stuff at initialization that got in our way (I don't
remember details) so we actually resolve the symbols at runtime so
swank.swank is only loaded when we actually start the swank server:

(defn start-swank
  "If a swank por

Re: Do you leave a Swank / nREPL in your production servers?

Karl Krukow — 2012-05-25T20:49:06

Sorry, and thanks for the info.

/Karl
On 25/05/2012, at 22.46, Phil Hagelberg wrote:

Re: Do you leave a Swank / nREPL in your production servers?

Phil Hagelberg — 2012-05-25T20:46:28

Just add swank as a non-dev dependency. The swank-clojure readme
actually covers how to do embedding.

-Phil

Re: Do you leave a Swank / nREPL in your production servers?

Karl Krukow — 2012-05-25T20:41:47

Sorry to ask what may be an obvious question to you. But what is a good way of embedding a swank server inside a production app? I'm kind of an emacs beginning and in development I just do lein swank and slime-connect from Emacs- that works fine.

If I'm embedding swank clojure server in non-development, the code would need to start up swank - how would I do that?

Thanks, 




On 24/05/2012, at 19.19, Sean Corfield wrote:

Re: Do you leave a Swank / nREPL in your production servers?

Sean Allen — 2012-05-25T20:34:30

I've never done this w/ clojure but in my last job when we were using
Gemstone Smalltalk, I made live code changes on a fairly regular basis.

Get the quick fix out. Do the complete release cycle later. It was only a
difference of 25 minutes, but often worth it.

Not for the faint of heart. I felt a little better doing it in Gemstone
because it was transaction based and I had to commit all my code changes.
A wee bit of a net that was.

-Sean-


On Thursday, May 24, 2012, blais wrote:

Re: Lambda: A lniux distro for clojurists

Matti Oinas — 2012-05-25T12:35:28

It could be possible to do all this by using package manager's like
apt or pacman. Create a package that depends on the software that is
allready available from the debian repository and provides all the
missing software. I think pacman in Arch linux could handle the same
by using package groups.

In distributions like Debian and Arch you can install only what you
want and configure it by yourself or use some predefined package group
and configuration. Installing desktop environment like gnome, kde and
xfce is a good example of this. In Arch for example I could install
only the parts of XFCE that I need and configure it afterwards like I
first did. The next time I installed XFCE in arch I just installed
xfce4 and xfce4-goodies groups and everything was ready.

It would probably be much less work to maintain and keep up to date
this kind of package than to maintain the whole distribution.

On Fri, May 25, 2012 at 2:18 PM, Roberto Mannai <robermann< at >gmail.com> wrote:

Re: Lambda: A lniux distro for clojurists

Jake Johnson — 2012-05-25T10:28:25

I like this idea a lot. My suggestions would be to base it on something
like Lubuntu. I have Ubuntu installed as dual-boot, but I tend to do most
of my work in Win7, so having a performance-oriented, small linux distro
installed inside VirtualBox helps if I don't want to reboot. I imagine
there will be others who would like to do the same. My other suggestion is
to base it on a distro that is extremely easy to install (like Ubuntu). I
just think it is always a good idea when creating something like this to
cater to the inexperienced guys, so that you attract as many newcomers as
possible. The experienced guys can set everything up themselves.

I would certainly use it for Overtone dev. Win7 isn't getting along well,
and I don't have the time right not to set up a virtual linux box to get it
running. Downloading and installing a small, simple distro would take care
of a lot of it for me.

On Thu, May 24, 2012 at 4:11 PM, banseljaj <ali.sajid.imami< at >gmail.com>wrote:

Clojure Pretty Printing with Blogger

Jake Johnson — 2012-05-25T10:07:24

Hi, guys. Has anyone had any luck with using
google-code-prettify<http://code.google.com/p/google-code-prettify/>and
http://google-code-prettify.googlecode.com/svn/trunk/src/lang-clj.jstogether
on blogger? I followed
http://www.tkglaser.net/2012/03/syntax-highlighting-in-blogger.html<http://www.tkglaser.net/2012/03/syntax-highlighting-in-blogger.html%20>,
but I can't get Clojure to display correctly, at least yet.