Tom Lane wrote: > Bruce Momjian <pgman@candle.pha.pa.us> writes: > > OK, reverted, but I saw it using contsel() so I figured we were allowing > > it, but I see contsel() is used by our "box", so ltree was just using > > something that was already there. Let me see if I can break out the new > > selectivity function into /contrib. > > What really needs to happen next is to think about which bits of > selfuncs.c should be exposed --- what's generally useful, and do we > think that it has an API clean/stable enough to expose? The reason I > kept all that stuff static so far was because it got whacked around > every release or two, and I didn't want to be constrained by worries > about breaking outside modules. I'd still prefer to minimize the number > of routines exposed, so some thought is needed.
Well, the ltree routine just needs struct VariableStatData and macro ReleaseVariableStats(), and we need to change one function from static to global. That and some #includes, and it works. Do we want to be more formal about it? -- Bruce Momjian http://candle.pha.pa.us EnterpriseDB http://www.enterprisedb.com + If your life is a hard drive, Christ can be your backup. + ---------------------------(end of broadcast)--------------------------- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
