Login | Register
My pages Projects Community openCollabNet

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Catacomb] [Ann] Catacomb ODBC



Sung Kim wrote:
> This is an interesting announce.
> I have two questions about the work.

> 1. Does ODBC affect performance?

I haven't perceived any noticeable performance loss. Of course, an objective
answer to this requires benchmarking but on the surface I'd say performance
isn't affected.

As Gianugo pointed out, a well-behaved ODBC driver may actually perform
better thanks to client-side caching, statement reuse and other niceties.

> 2. We used some MySQL biased SQL and was that the reason why
> you passed only 80% or litmus tests?

As it figures, our SQL is RDBMS-agnostic. All original MySQL-specific
statements were properly rephrased in terms of standard SQL. This may call
for a little extra configuration on the RDBMS side (like defining a
"SUBSTRING" function for Oracle or a "CONCAT" one for Postgresql) but the
SQL *is* portable across the leading RBDMS's. For the Oracle case -at least-
I've verified statements to be properly optimized in terms of choosing the
right indexes and the like.

The reason some Litmus tests do not pass is attributable to bugs in our
implementation that result in ODBC drivers behaving differently. Thus, for
example, Postgres passes 25 of the 25 "props" test (100% conformance) while
Oracle flunks 2 of them (and MySQL flunks 7). In other cases, we see tests
being flunked "consistently" across all drivers: a clear indication we're
truly missing something on the non-dbms side of things...

We continue to actively debug our code. A good thing about open source
projects is that by releasing this to the community the number of brains and
eyeballs increases substantially so these bugs should prove to be shallow
:-)


Un saludo,

Ricardo