On Mon, Aug 5, 2019 at 10:41 PM Tom Lane <t...@sss.pgh.pa.us> wrote: > There was some discussion of that upthread, and Peter argued that many > patches only need one OID anyway so why try harder. I'm not totally > sure I buy that --- my sense is that even simple patches tend to add > several related functions not just one.
That has been my experience, but it turns out that that was colored by the areas that I work in. I reviewed the history of pg_proc.dat today, and found that adding multiple entries at a time is more common that I thought it was. > But as long as the script > tells you how many OIDs are available, what's the problem? Just run > it again if you want a different suggestion, or make your own choice. Right. Besides, adding something along the lines Michael described necessitates fixing the problems that it creates. We'll run out of blocks of 5 contiguous OIDs (or whatever) far sooner than we'll run out of single OIDs. Now we have to worry about doing a second (actually a third) pass over the OIDs as a fallback when that happens. And so on. -- Peter Geoghegan
