are Perl regexps well-known enough for command-line use?

are Perl regexps well-known enough for command-line use?

[Eric Wong]

I've been thinking a bit about UI/UX for local command-line
tooling, and one thing I've been pondering is exposing Perl5
regexps as a mechanism for filtering
mailboxes/newsgroups/URLs/pathnames, etc...

It ties the UI to Perl5, though PCRE can get pretty close...

NNTP has it's own "wildmat" format in RFC 3977 which we support
(though maybe not 100% completely :x).  It's pretty close to
Bourne shell globbing, I think.

Perl's built-in and stdlib glob operators only work on FS paths,
and (as usual) I'm trying to avoid dependencies or Inline::C for
ease-of-installation and disk/bandwidth considerations.

Re: are Perl regexps well-known enough for command-line use?

[Konstantin Ryabitsev]

On Mon, Dec 14, 2020 at 08:39:38PM +0000, Eric Wong wrote:
> I've been thinking a bit about UI/UX for local command-line
> tooling, and one thing I've been pondering is exposing Perl5
> regexps as a mechanism for filtering
> mailboxes/newsgroups/URLs/pathnames, etc...

I think it's best to stick to the same regexp flavour as used by git for 
things like "git log -S", since people most likely to use this 
functionality are the same people likely to use git.

I believe it's POSIX, not PCRE -- but I don't know how reasonable it is 
to support POSIX regexps in Perl.

-K

Re: are Perl regexps well-known enough for command-line use?

[Eric Wong]

Konstantin Ryabitsev <konstantin@linuxfoundation.org> wrote:
> On Mon, Dec 14, 2020 at 08:39:38PM +0000, Eric Wong wrote:
> > I've been thinking a bit about UI/UX for local command-line
> > tooling, and one thing I've been pondering is exposing Perl5
> > regexps as a mechanism for filtering
> > mailboxes/newsgroups/URLs/pathnames, etc...
> 
> I think it's best to stick to the same regexp flavour as used by git for 
> things like "git log -S", since people most likely to use this 
> functionality are the same people likely to use git.

That's my first choice, too

> I believe it's POSIX, not PCRE -- but I don't know how reasonable it is 
> to support POSIX regexps in Perl.

There's an re engine for POSIX regexps on CPAN, but it's not in
any distros.  A strong dependency on Inline::C would make it
harder to install.

All the stuff that's in Xapian can be filtered using the normal
search terms and wildcard support in native Xapian; which can
includes all of what's in the proposed ls-* commands.

I think that leaves pathnames and URLs (IMAP/NNTP)
for --exclude=PATH/FOLDERS and --include...
globs with '?' and '*' support is sufficient, there, I hope.
No need to introduce users to IMAP vs NNTP wildcard differences.