incubator-cassandra-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jonathan Ellis <jbel...@gmail.com>
Subject Re: Column types rather pointless now
Date Mon, 19 Jul 2010 22:26:09 GMT
We are moving towards making greater use of type information, as in #749.
It's also necessary to provide (eventually) a command line client that is
actually useable.

On Jul 19, 2010 5:02 PM, "Folke Behrens" <folke.behrens@gmx.de> wrote:

Hi!

Since Cassandra switched to byte arrays for both keys and column names
marshalling has been completely offloaded to app devs.

Wouldn't it then be better to drop all but BytesType? (LongType could be
replaced by a comparator that accepts integers of arbitrary byte length,
like https://issues.apache.org/jira/browse/CASSANDRA-1282)

My advice:
- drop AsciiType because even Americans use a lot of words with
diacritcs, like fiancée, über, née, jalapeño, etc. It's also much slower
than BytesType.
- replace LongType with CASSANDRA-1282. This lets Cassandra accept other
integer values, i.e., ints from 32-bit PHP.
- drop UTF8Type because it's just BytesType with fancy log messages.
- drop LexicalUUIDType. Ditto.
- deprecate TimeUUIDType. Since users now have to serialize UUIDs
themselves they could just rearrange the bits of the timestamp and use
the fast BytesType, no?
- and finally, merge BytesType and AbstractType. No point in having the
standard comparator span over two classes.

Thoughts?

Regards,
Folke

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message