I've set up my apartment price map to update automatically. What did this entail? (Warning: boring programming stuff.)

There are two scripts involved:

$ crontab -l
...
# fetch the data at 2:02am on the 18th of the month; update the maps
# at 2:02 on the 19th.
02 02 18 * *         python /home/jefftk/query_padmapper.py
02 02 19 * *         python /home/jefftk/draw_heatmaps.sh

The first, query_padmapper.py, pulls apartment data from PadMapper and saves it in a timestamped file like apts-1376823721.txt. Unless Eric changes something, this just does its job. If something goes wrong, cron sends me an email with the error message.

(I really like this method of doing background tasks. Unless they're critical, don't try to recover from errors. Crash, print something informative, and have it show up in my email.)

The second script is a wrapper around draw_heatmap.py. It's safe to run any time because it looks for apartment data dumps that haven't been processed yet, but I intentionally only run it when I know it needs ot do something. It is:

set -e # exit on error

WORKING_DIR="/home/jefftk/jtk/apartment_prices/"

function compute_dates_available() {
  echo '<script type="text/javascript">'
  echo 'var dates_available = ['
  ls $WORKING_DIR/*.boston.bedroom*.png \
     | awk -F. '{print $(NF-1)}' \
     | sort \
     | while read line ; do
    echo '  "'$line'",'
  done
  echo ']'
  echo '</script>'
}

function update_index() {
  INDEX=$WORKING_DIR/index.html
  cp $INDEX $INDEX.pre_$(date +%F)
  cat $INDEX \
    | grep '<!-- begin list of date files -->' -B 10000 \
    > $INDEX.pre
  cat $INDEX \
    | grep '<!-- end list of date files -->' -A 10000 \
    > $INDEX.post
  compute_dates_available > $INDEX.middle

  cat $INDEX.pre $INDEX.middle $INDEX.post > $INDEX
}

for x in $WORKING_DIR/apts-1*.txt ; do
  if [ ! -e $x.started ] ; then
    YYYYMMDD=$(date --date=@$(echo $x | awk -F/ '{print $NF}' | sed \
                              s/apts-// | sed s/.txt//) +%F)
    touch $x.started
    for style in room bedroom ; do
      python /home/jefftk/code/apartment_prices/draw_heatmap.py $x $style
      mv $x.$style.1000.png $WORKING_DIR/apts.boston.$style.$YYYYMMDD.png
    done

    update_index
    touch $x.finished
  fi
done

What's this all doing? The for loop at the bottom considers every apartment data file. It only considers the ones where it hasn't started work, which it tracks by creating a file like apts-1376823721.txt.started. It extracts the timestamp from the filename and converts it to a date like 2013-08-18. Then it runs the real code, draw_heatmap.py, which produces an output file like apts-1376823721.txt.room.1000.png. It renames that to the format the UI expects, then calls update_index.

The update_index function changes a small piece of index.html which has a list of which dates have data available:

...
<!-- begin list of date files -->
<script type="text/javascript">
var dates_available = [
  "2011-06-16",
  "2013-01-29",
  "2013-02-18",
  "2013-03-18",
  "2013-04-18",
  "2013-05-18",
  "2013-06-18",
  "2013-07-18",
  "2013-08-18",
]
</script>
<!-- end list of date files -->
...

It figures out the available dates with compute_dates_available, formats that into a javascript array, then uses some grep to replace the portion of the file between the marker comments with the newly calculated dates.

So now the page stays up to date without me doing anything. Or else I wake up to an error in my email and figure out why.