workflow

[d^2]

all that talk about perl at the the PHPUG social last night reminded
me of this graphic:
http://revolution-computing.typepad.com/.a/6a010534b1db25970b0120a5968d6a970b-popup

that's actually exactly what has happened to me. i used to use perl/
bash to clean data, SAS or MATLAB to monkey around with it, and then
my final analysis would be done in MATLAB.

SAS and MATLAB are super expensive, though, and i can't guarantee i'll
have them at any employer (or myself if i ever start a business). R
fixes this problem very handily by being opensource, free to install,
actually superior to both offerings in terms of functionality, and
available on all OS platforms.

R can run as an application server (or simply in command line batch
mode), too, so there is plenty of opportunity to build an app in PHP
that uses R as the statistical backend. if anyone has to do any
mathematical/statistical modeling in their PHP work, i'd highly
recommend looking into this! i would expect* that R can do any of the
math in a more simple and intuitive way than PHP libraries (and i'd
also expect that R has more math libraries available than PHP)

* i haven't done any math modeling in PHP, so all i can do is
conjecture here

[Bradley Holt]

Dan,

Interesting stuff. Rasmus Lerdorf (the original creator of PHP) once
said something like, "use PHP as little as possible" (I can't find the
exact quote). What I think he was talking about was something very
similar to what you're talking about here with R: use PHP for the web
layer but integrate with another tool if it's better at solving the
specific problem you're working with.

Thanks,
Bradley

--
http://bradley-holt.blogspot.com/

[Matthew Weier O'Phinney]

On Fri, Sep 25, 2009 at 9:19 AM, Bradley Holt <bradle...@gmail.com> wrote:
> Interesting stuff. Rasmus Lerdorf (the original creator of PHP) once
> said something like, "use PHP as little as possible" (I can't find the
> exact quote). What I think he was talking about was something very
> similar to what you're talking about here with R: use PHP for the web
> layer but integrate with another tool if it's better at solving the
> specific problem you're working with.

Interestingly, today Rasmus posted an entry about setting up and using
Gearman with PHP:

http://toys.lerdorf.com/archives/51-Playing-with-Gearman.html

Gearman is a really interesting technology -- basically, it's a message
server, and you send messages to it from clients, and workers will then
process those messages. You can run it on the same machine, or have a
farm of Gearman servers that your frontend talks to (Digg, for instance,
does this). It currently has wrappers for most dynamic languages,
including a PHP extension in PECL.

The fun part about this is that you can implement the Gearman workers
in any language you want -- PHP, perl, C, python, Java, whatever. This
gives you tremendous flexibility when designing your application
architecture, and also makes it possible to make your frontend really,
really responsive by having it do basically next to nothing other than
communicating with Gearman and/or a caching layer (such as Memcached).

> On Fri, Sep 25, 2009 at 9:12 AM, d^2 <dan...@gmail.com> wrote:
>>
>> all that talk about perl at the the PHPUG social last night reminded
>> me of this graphic:
>> http://revolution-computing.typepad.com/.a/6a010534b1db25970b0120a5968d6a970b-popup
>>
>> that's actually exactly what has happened to me. i used to use perl/
>> bash to clean data, SAS or MATLAB to monkey around with it, and then
>> my final analysis would be done in MATLAB.
>>
>> SAS and MATLAB are super expensive, though, and i can't guarantee i'll
>> have them at any employer (or myself if i ever start a business). R
>> fixes this problem very handily by being opensource, free to install,
>> actually superior to both offerings in terms of functionality, and
>> available on all OS platforms.
>>
>> R can run as an application server (or simply in command line batch
>> mode), too, so there is plenty of opportunity to build an app in PHP
>> that uses R as the statistical backend. if anyone has to do any
>> mathematical/statistical modeling in their PHP work, i'd highly
>> recommend looking into this! i would expect* that R can do any of the
>> math in a more simple and intuitive way than PHP libraries (and i'd
>> also expect that R has more math libraries available than PHP)
>>
>> * i haven't done any math modeling in PHP, so all i can do is
>> conjecture here

--
Matthew Weier O'Phinney
mweiero...@gmail.com
http://weierophinney.net/matthew/