Re: WIP: System Versioned Temporal Table

Hi, thanks for working on this. I had planned to work on it and I’m looking forward to this natively in Postgres.

The patch builds with the following warnings:

plancat.c:2368:18: warning: variable 'name' is used uninitialized whenever 'for' loop exits because its condition is false [-Wsometimes-uninitialized]
for (int i = 0; i < natts; i++)
^~~~~~~~~
plancat.c:2379:9: note: uninitialized use occurs here
return name;
^~~~
plancat.c:2368:18: note: remove the condition if it is always true
for (int i = 0; i < natts; i++)
^~~~~~~~~
plancat.c:2363:15: note: initialize the variable 'name' to silence this warning
char *name;
^
= NULL
plancat.c:2396:18: warning: variable 'name' is used uninitialized whenever 'for' loop exits because its condition is false [-Wsometimes-uninitialized]
for (int i = 0; i < natts; i++)
^~~~~~~~~
plancat.c:2407:9: note: uninitialized use occurs here
return name;
^~~~
plancat.c:2396:18: note: remove the condition if it is always true
for (int i = 0; i < natts; i++)
^~~~~~~~~
plancat.c:2391:15: note: initialize the variable 'name' to silence this warning
char *name;
^
= NULL
2 warnings generated.

make check pass without issues, but make check-world fails for postgres_fdw, the diff is attached.

Before going further in the review, I’m a bit surprised by the quantity of code needed here. In https://github.com/xocolatl/periods there is far less code and I would have expected the same here. For example, are the changes to copy necessary or would it be possible to have a first patch the only implement the minimal changes required for this feature?

Thanks a lot!

Rémi