Predictive Analytics as Micro-services

Running R-as-a-service was something I was obsessed about for a while. I managed to search quite a few websites to come up with the strategy as below:





# GET THE CODE (Please do NOT skip this step and try to install using APTITUDE)

# ----------------------------------------------------------------------------




mkdir rApache

git init

git remote add origin https://github.com/jeffreyhorner/rapache

git pull origin master

sudo apt-get install r-base-dev apache2-mpm-prefork apache2-prefork-dev

# BUILD THE CODE
# ----------------
./configure --with-apxs=/usr/bin/apxs2
sudo make clean
sudo make
sudo make install


# CREATE 2 FILES:
#-----------------------------------------------------
sudo touch /etc/apache2/mods-available/r.conf
sudo touch /etc/apache2/mods-available/r.load


# ENTER THE FOLLOWING CONTENTS INTO THE r.conf FILE
#-----------------------------------------------------
<Location /R>
ROutputErrors
SetHandler r-script
RHandler sys.source
</Location>
<Location /RApacheInfo>
SetHandler r-info
</Location>


# ENTER THE FOLLOWING CONTENTS INTO THE r.load FILE
# -----------------------------------------------------
LoadModule R_module /usr/lib/apache2/modules/mod_R.so


# ENABLE THE RAPACHE MODULE
#----------------------------
sudo a2enmod r


# RESTART WEBSERVER
#----------------------------
#/etc/init.d/apache2 restart
sudo service apache2 restart


# TEST WEBSERVER
#----------------
# ENTER THE FOLLOWING CONTENTS INTO THE /var/www/R/test FILE
# ----------------------------------------------------------

y = rnorm(100)
print(y)


# OPEN BROWSER AND VISIT THE FOLLOWING URL:
# -----------------------------------------
http://localhost/R/test


# THE OUTPUT SHOULD LOOK SOMETHING LIKE THIS:
# ----------------------------------------------------------------

  [1] -1.11999180  0.77388394  0.47741262  0.59742444 -0.90835401 -0.73759768
  [7] -0.98358905  0.17164534 -0.30633769 -0.26356790 -0.72520593  1.09611240
 [13] -0.87002950 -0.68221878 -0.19672396  0.27719816  0.35486869 -2.62321997
 [19]  0.09403267 -0.59546938 -0.71671409 -0.37427584  0.22468514  0.39514394
 [25] -0.02668597 -0.04376564 -0.46238028 -0.26864170  0.55945746  0.18677297
 [31] -0.13377996 -0.14030508  0.02584913  2.25083262  0.10285414 -0.58769284
 [37]  0.88162455 -0.60222629 -0.19106042 -0.33075772  0.88249825  0.20592161
 [43] -0.53860098 -0.29864930 -0.34677542 -1.13457872 -0.52058656 -1.05854668
 [49] -0.87010347  0.47185953 -0.07035073 -0.61571393 -1.53781353 -1.48808031
 [55]  1.70151404  0.55821896 -0.86341586  1.01017915 -1.11602614  1.25460339
 [61] -0.07592987 -0.66240349  0.42753736 -0.87203214  1.13069688  1.10198204
 [67] -0.62701037 -0.91552938  1.34857229  2.00109074 -0.34138944  1.00038689
 [73]  1.94789712 -1.28643020  1.01272894  1.36189124  1.33133999 -0.37285196
 [79] -1.97906684 -1.47064465  1.54959469  1.34553324 -0.59611516  0.47388310
 [85]  2.42228484  0.66866070 -0.40705464 -0.48787802 -0.71283698  1.53617890
 [91]  1.02509900  0.87506909  0.79658073  1.97578280 -0.31492661  0.88690593
 [97]  1.02348963  1.75672929 -1.18559266  0.56992037

# Static Graphics can be created in R and transferred on HTTP response too!
# --------------------------------------------------------------------------

# Change the contents of /var/www/R/test file above and add the following:

setContentType ("image/png")
temp <- tempfile ()
y = rnorm (100)
png (temp, type="cairo")
plot (1:100, y, t='l')
dev.off ()
sendBin (readBin (temp, 'raw', n=file.info(temp)$size))
unlink (temp)


# -----------------------------------------------------------------------


#Of course one can handle other HTTP request methods and pass JSON back and forth
# ------------------------------------------------------------------------------


# Ensure that RJSONIO package is installed in R!
# Change the contents of the /var/www/R/test file above to be:

setContentType("application/json")

library(RJSONIO)
library(utils)
print(GET$data)
#json <- fromJSON(URLdecode(GET$data))
#print (json)
#print(toJSON(rnorm(10)))

#Interactive Graphics?
#-----------------------
# You just need a new apache module that supports ggplot (more on this for later...)



REFERENCES:
https://dzone.com/articles/building-json-webservice-r
https://binfalse.de/2011/05/28/r-for-the-web/