Caching large non-unicode dictionary in database?

Question

I have a large dictionary (outputs as string in 366MB, ~383764153 line filetextfile) that I want to store in a database for fast access and to skip the computation time involved in populating the dictionary.

My dictionary consists of a dictionary of dictionaries of filename/contents pairs. Small subset:

{
    'Reuters/19960916': {
        '54826newsML': '<?xml version="1.0"
encoding="iso-8859-1" ?>\r\n<newsitem itemid="54826" id="root"
date="1996-09-16" xml:lang="en">\r\n<title>USA: RESEARCH ALERT -
Crestar Financial cut.</title>\r\n<headline>RESEARCH ALERT - Crestar
Financial cut.</headline>\r\n<text>\n<p>-- Salomon Brothers analyst
Carole Berger said she cut her rating on Crestar Financial Corp to
hold from buy, at the same time lowering her 1997 earnings per share
view to $5.40 from $5.85.</p>\n<p>-- Crestar said it would buy
Citizens Bancorp in a $774 million stock swap.</p>\n<p>-- Crestar
shares were down 2-1/2 at 58-7/8. Citizens Bancorp soared 14-5/8 to
46-7/8.</p>\n</text>\r\n<copyright>(c) Reuters Limited',
        '55964newsML': '<?xml version="1.0" encoding="iso-8859-1"
?>\r\n<newsitem itemid="55964" id="root" date="1996-09-16"
xml:lang="en">\r\n<title>USA: Nebraska cattle sales thin at
$114/dressed-feedlot.</title>\r\n'
    }
}

I thought MongoDB would be a good fit, but it looks like it requires both the key and value need to be Unicode, and since I am grabbing the filenames from namelist() on ZipFile it is not guaranteed to be Unicode.

How would you recommend I serialise this dictionary into a database?

Answer 1

pymongo doesn't require strings to be unicode, it actually sends ascii stings as is and encodes unicodes to UTF8. When retrieving data from pymongo, you always get unicode. @@ http://api.mongodb.org/python/2.0/tutorial.html#a-note-on-unicode-strings

If your input contains "international" byte strings with high-order bytes (like ab\xC3cd) you need to convert these strings to unicode or encode them as UTF-8. Here's a simple recursive converter that handles arbitrary nested dicts:

def unicode_all(s):
    if isinstance(s, dict):
        return dict((unicode(k), unicode_all(v)) for k, v in s.items())
    if isinstance(s, list):
        return [unicode_all(v) for v in s]
    return unicode(s)

Answer 2

If you have the RAM (and you apparently do, because you populated the dictionary to begin with) -- cPickle. Or if you want something requiring less RAM but would be slower -- shelve.