Re: Making all nbtree entries unique by having heap TIDs participate in comparisons

On Wed, Sep 19, 2018 at 9:56 PM, Andrey Lepikhov
<a(dot)lepikhov(at)postgrespro(dot)ru> wrote:
> Note, that the interface of _bt_moveright() and _bt_binsrch() functions with
> combination of scankey, scantid and nextkey parameters is too semantic
> loaded.
> Everytime of code reading i spend time to remember, what this functions do
> exactly.
> May be it needed to rewrite comments.

I think that it might be a good idea to create an "BTInsertionScankey"
struct, or similar, since keysz, nextkey, the scankey array and now
scantid are all part of that, and are all common to these 4 or so
functions. It could have a flexible array at the end, so that we still
only need a single palloc(). I'll look into that.

> What do you think about submitting the patch to the next CF?

Clearly the project that you're working on is a difficult one. It's
easy for me to understand why you might want to take an iterative
approach, with lots of prototyping. Your patch needs attention to
advance, and IMV the CF is the best way to get that attention. So, I
think that it would be fine to go submit it now.

I must admit that I didn't even notice that your patch lacked a CF
entry. Everyone has a different process, perhaps.

--
Peter Geoghegan