Re: What's new in 8.3

Jim Nasby wrote:
> This is the list I'm working off of for the "What's New in 8.3"
> lightning talk at OSCon. I'm going to kind-of lump all the performance
> stuff together (and most of that is going to get a very, very quick
> overview), and blitz through the improved statistics stuff in like 10
> seconds... everything else I'm going to essentially go through in the
> order shown here, and whatever I can fit in in the time remaining is
> what makes it in. So, two questions...
>
> 1) Am I missing anything? (I'm not going to mention anything that hasn't
> been committed unless it's *really* big; namely HOT and tsearch)
> 2) Any *serious* issues with the ordering?
>
> Keep in mind that this is 100% targeted at *end users*, so these need to
> be ranked on how much they'll affect them. There's tons of patches that
> haven't even made this list not because they're not important, but
> because they just won't have that big of a direct impact on end users.
> Performance stuff sort of falls into this category... people care a lot
> about how performance has improved, but far less about how it was
> actually done (plus 5 minutes is nowhere near enough to do any justice
> to most of the performance stuff).
>
> * indicates feature isn't committed yet
>
> Integrated tsearch2*

this will be simply called text search or so once it got integrated ?

> XML
> UUID / ENUM

for UUID we also have contrib/uuid-ossp

> Sort order / NULLs first/last in indexes
> UPDATE/DELETE WHERE CURRENT OF cursor_name

aka updateable cursors (also in plpgsql)

> Arrays of compound types
> Functional costing
> Index advisor -- Does this actually do anything useful yet, or is it
> just backend hooks?

http://pgfoundry.org/projects/pgadviser/

> pg_standby
> improved regex support (regexp_matches, regexp_split_to_array,
> regexp_split_to_table)
> CREATE TABLE LIKE INCLUDING INDEXES
> GSSAPI
> Operator families
> Make CLUSTER MVCC safe

full MSVC++ support (big one for some developers)
improved support for replication solutions (ie the session controlable
trigger firing stuff)
plan invalidation
index support for IS NULL
scrollable cursors in pl/pgsql
hashing support for NUMERIC
RESET SESSION/TEMP/PLANS

>
> Performance
> HOT*
> Autovac workers
> plan cache
> SyncScan
> distributed checkpoint
> cmin/cmax combination
> varvarlena

faster multibyte text matching
multiple temporary tablespaces

>
> Improved stats info
> tuple stats in pg_stat_database
> bgwriter
> lock wait logging
> pg_stat_all_tables.n_live_tuples, n_dead_tuples
> log_temp_files

contrib/pageinspect (probably out of scope for endusers)


Stefan