[PEAK] Including PySQLite in PEAK

[Phillip J. Eby]

At 01:06 PM 12/7/04 -0500, Phillip J. Eby wrote:
>At 07:28 PM 12/7/04 +0200, Niki Spahiev wrote:
>>Phillip J. Eby wrote:
>>
>>>>>However, I personally have only extremely limited experience with 
>>>>>SQLite's actual performance and reliability in practice, so I would be 
>>>>>very interested to hear from anyone who has more experience using it.
>>>>
>>>>
>>>>FYI pysqlite is in transition between sqlite2 and sqlite3. There are 
>>>>currently several incompatible versions.
>>>
>>>So, are you saying that it would be a good idea for PEAK to bundle a 
>>>specific version for stability, or that it would be bad because it might 
>>>make it impossible for people to read their existing databases?
>>
>>pysqlite 1.0 uses sqlite2 which is dying.
>>pysqlite 1.1 uses sqlite3 with same python API as 1.0
>>pysqlite 2.0 will have better API using new features of sqlite3, but is 
>>unstable.
>>
>>PEAK can start with 1.1 and upgrade to 2.0 later.

It sounds like the sqlite2 and sqlite3 files are incompatible.  So I guess 
the relevant question here is, does anybody on the list need backward 
compatibility?

My inclination is to use PySQLite 1.1, but put it in peak.util.  That way, 
it won't interfere with a direct install of PySQLite 1.0, if somebody needs 
that.  And, if they need to use it with PEAK it's not that hard to add an 
alternate address scheme.