import sys
import flickrapi
from operator import itemgetter
import urllib
from xml.dom import minidom
import re
import nltk 
from nltk_contrib.readability import textanalyzer
import string

# This script generates a haiku about a particular location based on the
# titles of nearby flickr photographs.

# Sample comand line use:
# > python geo_haiku.py "721 broadway, new york ny"

# Optionally, pass the number of Haikus you want as a second argument:
# > python geo_haiku.py "721 broadway, new york ny" 3

# Set to true to print extra garbage.
debug = 0

# HELPER FUNCTIONS.

# Phonetic syllable counter via nltk_contrib.
# Dictionary-based approaches don't work relibably with the weird words of Flickr.
t = textanalyzer.textanalyzer('eng')
def count_syllables(word):
	temp_list = list()
	temp_list.append(word)
	count = t.countSyllables(temp_list)
	
	if count == 0:
		count = 1
	
	return count

def strip_punctuation(s):
	return s.translate(string.maketrans("",""), string.punctuation)	

def has_vowel(s):
	return re.search(r'[aeiou]+', s)

def is_picture(s):
	return re.search(r'jpg$|gif$|png$|jpeg$', s)

		
# GEOCODE THE ADDRESS.

# Grab a plain-text address from the command line.
address = sys.argv[1]
google_maps_api_key = 'XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX'
geo_url = 'http://maps.google.com/maps/geo?key=' + google_maps_api_key + '&output=xml&q=' + address
geocode = urllib.urlopen(geo_url)
geocode_xml = minidom.parse(geocode)

lon_lat = geocode_xml.getElementsByTagName('coordinates')[0].childNodes[0].nodeValue.split(',')

geocode.close()

latitude = lon_lat[1]
longitude = lon_lat[0]

if debug: print address + ' is at lat: ' + latitude + ' lon: ' + longitude

# FIND THE TAGS.

# Get the Flickr lib here: http://stuvel.eu/projects/flickrapi
flickr_api_key = 'XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX'
flickr = flickrapi.FlickrAPI(flickr_api_key, format='etree')

tags = dict()
tags_wanted = 150
tags_found = 0
search_radius = 1
current_page = 1

while tags_found < tags_wanted:
	# Search Flickr for images near the address given.
	# Moar info on thie API call: http://www.flickr.com/services/api/flickr.photos.search.html
	geo_photos = flickr.photos_search(page = str(current_page), min_taken_date = '1900-01-01 00:00:00', lat = str(latitude), lon = str(longitude), accuracy = '16', radius = str(search_radius), per_page = '250', extras = 'tags')

	total_pages = int(geo_photos[0].attrib['pages'])
	
	if debug: print 'reading page ' + str(current_page) + ' of ' + str(total_pages)

	for photo in geo_photos.getiterator('photo'):
		# Originally planned on using tags, seems to give more interesting results
		# with photo titles instead.
		photo_tags = photo.attrib['title'].encode('ascii', 'ignore').split(' ')
		#photo_tags = photo.attrib['tags'].encode('ascii', 'ignore').split(' ')
		
		for tag in photo_tags:
			tag = strip_punctuation(tag.lower())
			# Add to the dictionary, otherwise increment the frequency count.
			# Make sure the tag exists, and don't let numbers through (they're boring),
			# and that is has a vowel, and that it doesn't look like a meta-reference to a filename.
			if (len(tag) >= 1) and not re.search(r'\d+', tag) and has_vowel(tag) and not is_picture(tag):		
				if tag not in tags:
					tags[tag] = 1
				else:
					tags[tag] += 1
				
	# To get truly regional, need to subtract wider-area tags from a smaller search radius?
	# Maybe next iteration.
		
	# Update tags found.
	tags_found = len(tags)
	
	if debug: print 'Found ' + str(tags_found) + ' unique tags in a ' + str(search_radius) + ' mile radius'
	
	# If we didn't get everything we wanted, then we'll have to go to the next
	# page of Flickr result, or increase the search radius for the next round.
	# I found that I have to set a hard limit around 20 pages per search radius,
	# otherwise it's too tough to find unique tags.
	if (current_page >= total_pages) or (current_page >= 20):
		# Reset current page.
		current_page = 1
		
		# Bump the search radius.
		if debug: print 'Searching a larger area...'
		search_radius += 1
	else:
		# Search the next page.
		if debug: print 'Going to the next page'
		current_page += 1

# Perhaps a cleaner way to sort dictionaries?
# http://code.activestate.com/recipes/304440/
if debug:
	# Print each word found, and its frequency.
	print '{'
	for tag in sorted(tags.items(), key = itemgetter(1), reverse = True):
		print '"' + tag[0] + '": '+ str(tag[1]) + ', '
	print '}'	
	print '\n'

# GENERATE THE HAIKU.
used_words = list() # Make sure we only use words once.
rows = [5, 7, 5] # Set the syllable scheme and row cout.
haiku = ''

# Allow just one boring word per line, and don't let the line end in a boring word.
boring_words = ['i', 'the', 'of', 'and', 'to', 'or', 'a', 'on', 'at', 'for', 'is', 'in', 'your', 'you', 'it', 'that', 'this']


# Figure out how many haikus to write on the topic.
haikus_wanted = 1
if len(sys.argv) >= 3:
	haikus_wanted = int(sys.argv[2])
	
for i in range(haikus_wanted):

	# Build each row.
	for syllables_wanted in rows:
		syllables_in_row = 0
		used_boring = 0
	
		# Go through each tag, sorted by frequency.		
		for tag in sorted(tags.items(), key = itemgetter(1), reverse = True):
			word = tag[0]

			# Make sure the word won't put us past the syllable limit for the row.
			syllables_in_word = count_syllables(strip_punctuation(word))
			tentative_syllables_in_row = syllables_in_row + syllables_in_word
		
			if (word in boring_words) and used_boring:
				# don't get to use more than one boring word per line.
				if debug: print word + ' is boring, and we already spent our boring allowance, skip it'			
				continue
		
			if word in used_words:
				# keep looking if we already used the word
				continue
			
			if (tentative_syllables_in_row == syllables_wanted) and (word in boring_words):
				# don't end a line with a boring word
				continue

			if tentative_syllables_in_row <= syllables_wanted:

				if word in boring_words:
					if debug: print word + ' is boring...'
					used_boring = 1 # we used our boring card for this line...
				
				haiku += word + ' '
				syllables_in_row = tentative_syllables_in_row
				used_words.append(word)
			
				# Optionally print syllable count for a sanity check.
				if debug: print "\"" + word + "\" has " + str(syllables_in_word)			
	
		haiku = haiku.strip() + '\n'

	# Keep writing haikus!
	haiku += '\n'

print haiku.strip()