Is there a Python persitent data store that has the same functionality as dict (or how to coax "Shelve" to get that)?

I am using Python 3. shelve is advertised in the documentation as a persistent version of开发者_如何学Go dict. However, in my usage, it turns out that shelve does not allow tuples as keys, while dict does:

import shelve
def tryShelve():
    db = shelve.open("shelvefile")
    db["word1"] = 1
    db[("word1", "word2")] = 15

tryShelve()

produces this error:

Traceback (most recent call last):
  File "<pyshell#41>", line 1, in <module>
    tryShelve()
  File "<pyshell#40>", line 4, in tryShelve
    db[("word1", "word2")] = 15
  File "C:\Python32\lib\shelve.py", line 125, in __setitem__
    self.dict[key.encode(self.keyencoding)] = f.getvalue()
AttributeError: 'tuple' object has no attribute 'encode'

---

The first paragraph in the shelve module documentation:

A “shelf” is a persistent, dictionary-like object. The difference with “dbm” databases is that the values (not the keys!) in a shelf can be essentially arbitrary Python objects — anything that the pickle module can handle. This includes most class instances, recursive data types, and objects containing lots of shared sub-objects. The keys are ordinary strings.

[emphasis mine]

yaml module allows tuples as keys:

>>> d = {}
>>> d["word1"] = 1
>>> d[("word1", "word2")] = 15
>>> import yaml
>>> yaml.dump(d)
'word1: 1\n? !!python/tuple [word1, word2]\n: 15\n'
>>> yaml.load(_)
{('word1', 'word2'): 15, 'word1': 1}

---

Tuples don't work as keys in the Python 2.7.1 version of shelve either. The 2.7.1 documention suggests looking at the ActiveState Persistent dictionary recipe. Which I tested and it seemed to work using tuples as keys as shown in your example code with the 'csv' format, although not with the 'jason' format (and I didn't try 'pickle'). You'll understand what I mean by 'format' if you look at the recipe.

---

It appears to me that shelve can't serialize tuples for writing to file.
Consider pickle as an alternative.

---

You could use dumps and loads to convert your tuples to strings before using them as keys in a shelve. Alternatively, if your tuple only contains literals, you can use repr(your_tuple) to obtain a string-representation that you can convert back to a tuple using literal_eval.

To make the conversion more convenient, you could subclass Shelf and override __getitem__ and __setitem__ to do the conversions automatically on access.