On Wed, Feb 15, 2023 at 12:45:52PM -0600, Justin Pryzby wrote:
> On Mon, Feb 13, 2023 at 07:57:25AM -0500, Andrew Dunstan wrote:
> > On 2023-02-12 Su 15:59, Justin Pryzby wrote:
> > > It seems like if pgindent knows about git, it ought to process only
> > > tracked files. Then, it wouldn't need to manually exclude generated
> > > files, and it wouldn't process vpath builds and who-knows-what else it
> > > finds in CWD.
> >
> > I don't really want restrict this to tracked files because it would mean you
> > can't pgindent files before you `git add` them.
>
> I think you'd allow indenting files which were either tracked *or*
> specified on the command line.
>
> Also, it makes a more sense to "add" the file before indenting it, to
> allow checking the output and remove unrelated changes. So that doesn't
> seem to me like a restriction of any significance.
>
> But I would never want to indent an untracked file unless I specified
> it.
Agreed. I use pgindent three ways:
1. Indent everything that changed between master and the current branch. Most
common, since I develop nontrivial patches on branches.
2. Indent all staged files. For trivial changes.
3. Indent all tracked files. For typedefs.list changes.
That said, pre-2023 pgindent changed untracked files if called without a file
list. I've lived with that and could continue to do so.