Reduce RAM usage in Python script

Question

I've written a quick little program to scrape book data off of a UNESCO website which contains information about book translations. The code is doing what I want it to, but by the time it's processed about 20 countries, it's using ~6GB of RAM. Since there are around 200 I need to process, this isn't going to work for me.

I'm not sure where all the RAM usage is coming from, so I'm not sure how to reduce it. I'm assuming that it's the dictionary that's holding all the book information, but I'm not positive. I'm not sure if I should simply make the program run once for each country, rather than processing the lot of them? Or if there's a better way to do it?

This is the first time I've written anything like this, and I'm a pretty novice, self-taught programmer, so please point out any significant flaws in the code, or improvement tips you have that may not directly relate to the question at hand.

This is my code, thanks in advance for any assistance.

from __future__ import print_function
import urllib2, os
from bs4 import BeautifulSoup, SoupStrainer

''' Set list of countries and their code for niceness in explaining what
is actually going on as the program runs. '''
countries = {"AFG":"Afghanistan","ALA":"Aland Islands","DZA":"Algeria"}

'''List of country codes since dictionaries aren't sorted in any
way, this makes processing easier to deal with if it fails at
some point, mid run.'''
country_code_list = ["AFG","ALA","DZA"]

base_url = "http://www.unesco.org/xtrans/bsresult.aspx?lg=0&c="
destination_directory = "/Users/robbie/Test/"
only_restable = SoupStrainer(class_="restable")

class Book(object):
    def set_author(self,book): 
        '''Parse the webpage to find author names. Finds last name, then
        first name of original author(s) and sets the Book object's 
        Author attribute to the resulting string.'''

        authors = ""
        author_last_names = book.find_all('span',class_="sn_auth_name")
        author_first_names = book.find_all('span', attrs={\
            'class':"sn_auth_first_name"})
        if author_last_names == []: self.Author = [" "]

        for author in author_last_names:
            try: 
                first_name = author_first_names.pop()
                authors = authors + author.getText() + ', ' + \
                    first_name.getText()

            except IndexError:
                authors = authors + (author.getText())
        self.author = authors

    def set_quality(self,book):
        ''' Check to see if book page is using Quality, then set it if 
        so.'''

        quality = book.find_all('span', class_="sn_auth_quality")

        if len(quality) == 0: self.quality = " "

        else: self.quality = quality[0].contents[0]

    def set_target_title(self,book): 
        target_title = book.find_all('span', class_="sn_target_title")
        if len(target_title) == 0: self.target_title = " "
        else: self.target_title = target_title[0].contents[0]

    def set_target_language(self,book): 
        target_language = book.find_all('span', class_="sn_target_lang")
        if len(target_language) == 0: self.target_language = " "
        else: self.target_language = target_language[0].contents[0]

    def set_translator_name(self,book) : 
        translators = ""
        translator_last_names = book.find_all('span', class_="sn_transl_name")
        translator_first_names = book.find_all('span', \
                                               class_="sn_transl_first_name")
        if translator_first_names == [] and translator_last_names == [] :
            self.translators = " "
            return None

        for translator in translator_last_names:
            try: 
                first_name = translator_first_names.pop()
                translators = translators + \
                    (translator.getText() + ',' \
                     + first_name.getText())
            except IndexError:
                translators = translators + \
                    (translator.getText())

        self.translators = translators  

    def set_published_city(self,book) : 
        published_city = book.find_all('span', class_="place")
        if len(published_city) == 0: 
            self.published_city = " "
        else: self.published_city = published_city[0].contents[0]

    def set_publisher(self,book) : 
        publisher = book.find_all('span', class_="place")
        if len(publisher) == 0: 
            self.publisher = " "
        else: self.publisher = publisher[0].contents[0] 

    def set_published_country(self,book) : 
        published_country = book.find_all('span', \
                                        class_="sn_country")
        if len(published_country) == 0: 
            self.published_country = " "
        else: self.published_country = published_country[0].contents[0]

    def set_year(self,book) : 
        year = book.find_all('span', class_="sn_year")
        if len(year) == 0: 
            self.year = " "
        else: self.year = year[0].contents[0]   

    def set_pages(self,book) : 
        pages = book.find_all('span', class_="sn_pagination")
        if len(pages) == 0: 
            self.pages = " "
        else: self.pages = pages[0].contents[0] 

    def set_edition(self, book) :
        edition = book.find_all('span', class_="sn_editionstat")
        if len(edition) == 0: 
            self.edition = " "
        else: self.edition = edition[0].contents[0]

    def set_original_title(self,book) : 
        original_title = book.find_all('span', class_="sn_orig_title")
        if len(original_title) == 0: 
            self.original_title = " "
        else: self.original_title = original_title[0].contents[0]   

    def set_original_language(self,book) :
        languages = ''
        original_languages = book.find_all('span', \
                                         class_="sn_orig_lang")

        for language in original_languages:
            languages = languages + language.getText() + ', '

        self.original_languages = languages

    def export(self, country): 
        ''' Function to allow us to easilly pull the text from the 
        contents of the Book object's attributes and write them to the 
        country in which the book was published's CSV file.'''

        file_name = os.path.join(destination_directory + country + ".csv")

        with open(file_name, "a") as by_country_csv:        
            print(self.author.encode('UTF-8') + " & " + \
                  self.quality.encode('UTF-8') + " & " + \
                  self.target_title.encode('UTF-8') + " & " + \
                  self.target_language.encode('UTF-8') + " & " + \
                  self.translators.encode('UTF-8') + " & " + \
                  self.published_city.encode('UTF-8') + " & " + \
                  self.publisher.encode('UTF-8') + " & " + \

                  self.published_country.encode('UTF-8') + " & " + \
                  self.year.encode('UTF-8') + " & " + \
                  self.pages.encode('UTF-8') + " & " + \
                  self.edition.encode('UTF-8') + " & " + \
                  self.original_title.encode('UTF-8') + " & " + \
                  self.original_languages.encode('UTF-8'), file=by_country_csv)

        by_country_csv.close()

    def __init__(self, book, country):
        ''' Initialize the Book object by feeding it the HTML for its 
        row'''
        self.set_author(book)
        self.set_quality(book)
        self.set_target_title(book)
        self.set_target_language(book)

        self.set_translator_name(book)
        self.set_published_city(book)
        self.set_publisher(book)
        self.set_published_country(book)

        self.set_year(book)
        self.set_pages(book)
        self.set_edition(book)
        self.set_original_title(book)

        self.set_original_language(book)


def get_all_pages(country,base_url):
    ''' Create a list of URLs to be crawled by adding the ISO_3166-1_alpha-3
    country code to the URL and then iterating through the results every 10
    pages. Returns a string.'''

    base_page = urllib2.urlopen(base_url+country)
    page = BeautifulSoup(base_page, parse_only=only_restable)

    result_number = page.find_all('td',class_="res1",limit=1)
    if not result_number:
        return 0

    str_result_number = str(result_number[0].getText())
    results_total = int(str_result_number.split('/')[1])

    page.decompose()

    return results_total


def build_list(country_code_list, countries):
    '''  Build the list of all the books, and return a list of Book objects
    in case you want to do something with them in something else, ever.'''
    for country in country_code_list:

        print("Processing %s now..." % countries[country])
        results_total = get_all_pages(country, base_url)

        for url in range(results_total):
            if url % 10 == 0 :
                all_books = []  
                target_page = urllib2.urlopen(base_url + country \
                                             +"&fr="+str(url))
                page = BeautifulSoup(target_page, parse_only=only_restable)
                books = page.find_all('td',class_="res2")
                for book in books:
                    all_books.append(Book (book,country))
                page.decompose()

                for title in all_books:
                    title.export(country)    
    return

if __name__ == "__main__":
    build_list(country_code_list,countries)
    print("Completed.")

Answer 1

I guess I'll just list off some of the problems or possible improvements in no particular order:

1. Follow PEP 8 .

   Right now, you've got lots of variables and functions named using camel-case like setAuthor . That's not the conventional style for Python; Python would typically named that set_author (and published_country rather than PublishedCountry , etc.). You can even change the names of some of the things you're calling: for one, BeautifulSoup supports findAll for compatibility, but find_all is recommended.

   Besides naming, PEP 8 also specifies a few other things; for example, you'd want to rewrite this:

    if len(resultNumber) == 0 : return 0 

   as this:

    if len(result_number) == 0: return 0 

   or even taking into account the fact that empty lists are falsy:

    if not result_number: return 0 

2. Pass a SoupStrainer to BeautifulSoup .

   The information you're looking for is probably in only part of the document; you don't need to parse the whole thing into a tree. Pass a SoupStrainer as the parse_only argument to BeautifulSoup . This should reduce memory usage by discarding unnecessary parts early.

3. decompose the soup when you're done with it.

   Python primarily uses reference counting, so removing all circular references (as decompose does) should let its primary mechanism for garbage collection, reference counting, free up a lot of memory. Python also has a semi-traditional garbage collector to deal with circular references, but reference counting is much faster.

4. Don't make Book.__init__ write things to disk.

   In most cases, I wouldn't expect just creating an instance of a class to write something to disk. Remove the call to export ; let the user call export if they want it to be put on the disk.

5. Stop holding on to so much data in memory.

   You're accumulating all this data into a dictionary just to export it afterwards. The obvious thing to do to reduce memory is to dump it to disk as soon as possible. Your comment indicates that you're putting it in a dictionary to be flexible; but that doesn't mean you have to collect it all in a list: use a generator, yielding items as you scrape them. Then the user can iterate over it just like a list:

    for book in scrape_books(): book.export() 

   …but with the advantage that at most one book will be kept in memory at a time.

6. Use the functions in os.path rather than munging paths yourself.

   Your code right now is rather fragile when it comes to path names. If I accidentally removed the trailing slash from destinationDirectory , something unintended happens. Using os.path.join prevents that from happening and deals with cross-platform differences:

    >>> os.path.join("/Users/robbie/Test/", "USA") '/Users/robbie/Test/USA' >>> os.path.join("/Users/robbie/Test", "USA") # still works! '/Users/robbie/Test/USA' >>> # or say we were on Windows: >>> os.path.join(r"C:\\Documents and Settings\\robbie\\Test", "USA") 'C:\\\\Documents and Settings\\\\robbie\\\\Test\\\\USA' 

7. Abbreviate attrs={"class":...} to class_=... .

   BeautifulSoup 4.1.2 introduces searching with class_ , which removes the need for the verbose attrs={"class":...} .

I imagine there are even more things you can change, but that's quite a few to start with.

Answer 2

What do you want the booklist for, in the end? You should export each book at the end of the "for url in range" block (inside it), and do without the allbooks dict. If you really need a list, define exactly what infos you will need, not keeping full Book objects.