On Wed, 30-Dec-98, jagged@shaitan.demiurge.org
asked about how to delete Large Objects.
> Now, if you do a
> DELETE FROM foo WHERE descr = 'obj1';
> you'll wind up deleting the tuple in foo, but if you do a
> SELECT lo_export(166219, '/tmp/frotz.txt');
> you'll find that the large object is still in the database, which makes
> sense, since you're only deleting a pointer to the lo. I apologise for the
> pedantry -- I just wanted to make sure that I was on solid ground before I
> asked what will very probably be a silly question: How do you delete a
> large object from the database? I didn't see a built in function, and
there
> wasn't anything in section 15 of the programmer's manual.
I too ran across this problem. I was surprised, given that the design of the
interface was to look like UNIX file operations that there wasn't some kind
of an "unlink" command. I went to the source code and found that indeed
there is exactly that command. I know that Bruce Mojian spent some time
last year making sure that the large objects worked, so I expected that with
release 6.4 the documentation would be updated as well. It doesn't appear to
have been, however, so I'll make the specific request that the documentation
for Large Objects be "reviewed". It looked to me, from the source code, that
everything was in place - its just that it isn't fully documented.
The output from "ctags -x" on .../src/interfaces/libpq/fe-lobj.c shows the
following functions are available:
lo_close 93 fe-lobj.c lo_close(
lo_creat 267 fe-lobj.c lo_creat(
lo_export 448 fe-lobj.c lo_export(
lo_import 383 fe-lobj.c lo_import(
lo_initialize 509 fe-lobj.c lo_initialize(
lo_lseek 218 fe-lobj.c lo_lseek(
lo_open 46 fe-lobj.c lo_open(
lo_read 132 fe-lobj.c lo_read(
lo_tell 305 fe-lobj.c lo_tell(
lo_unlink 343 fe-lobj.c lo_unlink(
lo_write 172 fe-lobj.c lo_write(
In many cases one can guess what the return values will be, however, in a
few cases, the meaning of error-returns is not clear.
I am willing to help work on upgrading some documentation here, but I am
not terribly knowlegeable about the ins and outs of these routines. If
someone who is will work with me (say, I write and the other person reviews
and corrects?) we may be able to update some of this...
Mark Dalphin mdalphin@amgen.com
PS Sorry about any "funny" formating here; this is a brand new mailer to
me and I am not at all comfortable with it!