remember 'known unknowns'

[Martin Pool]

wayned/mpe say

>I think that there needs to be some way to output the ignored files
>summarized on a per-matching rule basis so that people can see if an
>extra rule has started to show up. Perhaps the default output of the
>"bzr ignored" command should be changed to output such a list, and a new
>--verbose option added to make it output each file (like it does now).
>
>Another improvement would be to add some more smarts to the add command
>that would let it differentiate between files that were already-known
>to be ignored, and files that were newly ignored, since these new files
>are the only ones that have a big chance of being erroneously ignored.
>Here's one way to accomplish this:
>
> - Leave "bzr add" chatty about ignored files, but only if the file is
> not in a list of known-to-be-ignored files (perhaps maintained in a
> versioned property). If the command mentions a file that is not in
> this list, it suggests running the command "bzr ignored --new" to
> approve these new finds. For files in the known-to-be-ignored list,
> they are summarized in a single-line count.
>
> - Add the "--new" option to "bzr ignored" that makes it only mention
> files that are not in the known-to-be-ignored list (summarized by
> rule without -v), and to update the list with these new names. If a
> file should not have been ignored, the user can always explicitly add
> the file, which would also remove the name from the ignored list.
>
>I think that this would both cleanup the output of "bzr add" and make
>the interaction of new files with the ignore rules safer.

That's an interesting idea, although it basically comes to down to
ignored files not actually being ignored. ie. In order to detect "new"
ignored files you need to keep a list of all the existing ignored
files .. which means you're not actually ignoring them.

So basically we could have "unversioned" files, that are known but not
managed by bzr, as opposed to ignored. Then when new files appear in
the tree bzr would know exactly which ones are new and could present
them to the user. But I'm not sure, interesting to think about though.

[Martin Pool (mbp)]

I'm not really convinced this would be useful but it's an interesting idea.

[Robert Collins (lifeless)]

It would fall out quite naturally from caching the result of 'ignored' calculations