Re: Refactoring of compression options in pg_basebackup

On Wed, Jan 19, 2022 at 08:35:23AM -0500, Robert Haas wrote:
> I think that this will reject something like --compress=nonetheless by
> telling you that 't' is not a valid separator. I think it would be
> better to code this so that we first identify the portion preceding
> the first colon, or the whole string if there is no colon. Then we
> check whether that part is a compression method that we recognize. If
> not, we complain.

Well, if no colon is specified, we still need to check if optarg
is a pure integer if it does not match any of the supported methods,
as --compress=0 should be backward compatible with no compression and
--compress=1~9 should imply gzip, no?

> If so, we then check whatever is after the separator
> for validity - and this might differ by type. For example, we could
> then immediately reject none:4, and if in the future we want to allow
> lz4:fast3, we could.

Okay.

> I think the code that handles the bare integer case should be at the
> top of the function and should return, because that code is short.
> Then the rest of the function doesn't need to be indented as deeply.

Done this way, I hope.
--
Michael