Re: Database notes

[Chris Albertson <chris@topdog.pas1.logicon.com>]

Stupendous Man wrote:
> 
>   Thanks for the information.  I'll make the fix to my Postgres server
> this weekend.
> 
> > 3) I ran into what is arguably a bug in Postscript.
> 
>   Um, do you mean "Postgres?"  I wouldn't doubt that you could find
> some bugs in Postscript, too, of course :-)

That's good.  I've crossed up a couple threads.  Yes I've found
problems with Adobe Postscript.  Find Ghostscript to be a little
more robust.  It has to do with the fact that GPL'd software gets
a kind or "peer review".  Anyone who wants to can open it up and
see any mistakes in the code and report/fix them.

Now, the reason I keep posting software before it is done is for
the same reason.  I keep hopping that people will detect the 
bugs now rather then later.

As posted there is zero documentation so here goes a few notes.

1) to compile you need the "Driver_libs.tar" file that was posted
to /pub/incoming some time ago.  You need to edit the Makefile to
point to whereever you unpacked this tar file to.  
/usr/local{lib|include} is a good place.

2) Running "merge" will destroy the content of starlist_dat and
starlist_hdr tables.  Do something if you done want them wiped
clean.

3) "merge -h" or "merge --help" will print a list a accepted
command line switches and then quit.

4) the "-w" switch controls the "slice width".  I process the
database in asending "slices" of RA.  You need to pick a
width such that there are no more then about 30,000 catalog
entries in each slice.  This include new ones to be created
durring the run.  It defaults pretty wide.  Try setting it to 
1.0 (this 1.0 degree of RA per slice)

Reason for all this is a stupid limitation in Postgresql of
only 65,000 SQL statments betwen a BEGIN..END pair.  I'll
change the algorithm to work around this limit later.

5) There should be a script called "import" that runs everything.
It builds the tables, runs the Perl Import then runs the C
merge program.  It builds some indexes along the way to.
Just type "./import foobar/*"  where * is a bunch of files like
B1_0456.432

6) As everyone is running systems with the same CPU I could
post a binary.  I prefer not to as compilation forces testing
of the Makefile on more systems and flushes out bugs there.
-- 

--Chris Albertson

  chris@topdog.logicon.com                Voice:  818-351-0089  X127
  Logicon RDA, Pasadena California          Fax:  818-351-0699