gistfile1.py

#-*- coding: utf-8 -*-
import sys
import pprint

import internetarchive

def _search_collection(collection_name):
    """ Searches the internet archive for the specified collection.
    if no items are found for the collection it returns None otherwise
    the Search object is returned.
    """
    collection = internetarchive.search_items('collection:{}'.format(collection_name))
    if collection.num_found == 0:
        return None
    else:
        return collection

def _get_item_data(item):
    data = internetarchive.get_item(item.get('identifier'))
    total_size = 0
    file_info  = []
    for f in data.files:
        if f['source'] == 'original':
            file_info.append(f)
            total_size += int(f.get('size',0))
    return {'identifier' : data.identifier,
            'files'      : file_info,
            'size'       : total_size}

if __name__ == '__main__':
    if len(sys.argv) < 2:
        print('Missing parameter: collection name')
        sys.exit(-1)

    collection = sys.argv[1]
    print('Getting data for the collection {}'.format(collection))

    collection_data = _search_collection(collection)
    if not collection:
        print('No collection {} found'.format(collection))
        sys.exit(-1)
    else:
        results = []
        # Note the internetarchive library does a http request for each item so this
        # could take some time.
        for item in collection_data:
            print('Proccessing item {0}\n'.format(item['identifier']))
            results.append(_get_item_data(item))
        collection_original_size = 0
        for item in results:
            collection_original_size += int(item['size'])
            print('Item {}:'.format(item['identifier']))
            print('\tTotal original files: {}'.format(len(item['files'])))
            print('\tTotal size: {}'.format(item.get('size')))
        print('Total collection original file size: {}'.format(collection_original_size))