Rom

[added] support for Python 3.3 thanks to the six library and porting effort
from Justin Mayfield https://github.com/mayfield, which also includes some
additional tests.
[remove] String column from Python 3.3, use the Text column instead.
[changed] Both String and Text columns can use unique columns.
[fixed] rom's behavior with long ints in some situations.
[fixed] rom's behavior with unicode suffixes in Python 2.x.
[added] tests for unicode text indexes, unique indexes, prefix indexes, and
suffix indexes.
[updated] relevant docs regarding the Text column and unique indexes.
[fixed] improperly labeled 0.25.1 release in this changelog.

[fixed] an issue where Lua writing wouldn't properly update unique indexes on
deletion was fixed. Thank you to https://github.com/thedrow for the
report.
[changed] added a link to the full Python docs to the Readme.

[changed] version numbers to account for bugfixes vs. feature updates.
[added] columns can now be defined to allow for prefix and/or suffix queries.
Enabling prefix queries also enables arbitrary pattern matching over your
data.
[fixed] in some cases, rom would allow the definition of multiple primary
keys, of which only one would ever be used (inconsistently). This will now
result in an error.
[changed] defaulted to assume Lua is available on Redis, which has been
released for over 15 months at this point. You can disable support via
a call to rom._disable_lua_writes().
[added] the ability to cache and get the key that holds the result of a query,
which can be used for pagination, etc. See: Query.cached_result()
[warning] using rom versions of 0.23 with 0.25.0 when prefix and suffix
indexes are enabled can result in improper results from prefix, suffix,
and/or pattern queries, and can result in orphan data living in prefix or
suffix indexes. Upgrade all of your clients!
[changed] temporary keys for queries are now prefixed with the name of the
model over which queries are being executed on. This should effect
basically zero people, but can allow for query cleanup in the off chance
of a failure during execution.

[changed] reduced number of round trips for single-filter queries by 1, thanks
to https://github.com/MickeyKim for the report.

[fixed] size estimation for intersection ordering when filtering has now been
fixed, thank you to https://github.com/MickeyKim for the report and the
change (should improve performance).
[fixed] an issue with some types when trying to update attributes has now been
fixed, thank you to https://github.com/denisvolokh for the report.
[changed] improved performance for simple numeric range queries of the form
Model.get_by(attr=value) or Model.get_by(attr=(min, max)) by roughly a
factor of 60x or better in some cases. Thank you to
https://github.com/MickeyKim for the report on poor performance.

[fixed] upload for rom 0.20 was missing new columns.py, now fixed