Re: cut -DF

[Pádraig Brady]

On 25/01/2022 22:50, Rob Landley wrote:
> On 1/25/22 3:39 PM, Pádraig Brady wrote:
> > On 25/01/2022 07:55, Assaf Gordon wrote:
> > > Hello,
> > >
> > > Here's an updated patch for "cut -DF".
> > > Since it's a new code path, it opens the possibility of finally
> > > supporting multibyte characters with "cut -c".
> > >
> > >
> > > comments very welcomed,
> > >     - assaf
> > >
> > >     [PATCH 01/18] cut: set-fields: add no-sort options
> > >     [PATCH 02/18] cut: iniitial -D implmentation, currently only with
> > >     [PATCH 03/18] tests: add 'cut -D' tests
> > >     [PATCH 04/18] cut: extract 'cut -D -f' to a separate function
> > >     [PATCH 05/18] cut: implement -D with -b
> > >     [PATCH 06/18] tests: add 'cut -D -b' tests
> > >     [PATCH 07/18] cut: add -O short-option for --output-delimiter
> > >     [PATCH 08/18] cut: implement -F
> > >     [PATCH 09/18] tests: add 'cut -F' tests
> > >     [PATCH 10/18] cut: extract cut-fields into separate functions
> > >     [PATCH 11/18] cut: implement multibyte -c/--characters
> > >     [PATCH 12/18] cut: change -F regex syntax to BRE
> > >     [PATCH 13/18] cut: change -D long-option equivalent
> > >     [PATCH 14/18] doc: mention 'cut -D' in NEWS
> > >     [PATCH 15/18] doc: mention 'cut -F' in NEWS
> > >     [PATCH 16/18] doc: mention 'cut -O' in NEWS
> > >     [PATCH 17/18] doc: mention multibyte 'cut -c' in NEWS
> > >     [PATCH 18/18] doc: expand 'cut' section
> >
> > Looking great!
> > Some initial notes...
> >
> > 0008
> >     I was surprised that -D implies -s
> >     (as there is no way to undo that without another option).
>
> I was ready to make -s be a toggle (so you can switch it back on after -D) but
> nobody ever asked for it. "I asked for the third and fifth words of each line
> but a line with only one word will always pass it through" turns out not to be a
> behavior in high demand.

I agree that the default should originally have been -s.
Though passing through non delimited lines can be very useful.
Consider maintaining the grouping piped from `uniq --group` for example.

> Let me know if you decide to make -s toggle so I can propagate that to the
> existing two implementations for consistency.

Hrm. Short options can be amenable to toggles, but long options not so much.
We'd probably have to have a separate "-S, --pass-undelimited" option
to get back that functionality.

> >     I'm even more surprised that -F implies -s
> >     Fair enough for compat with existing implementations,
> >     but please double check the -F behavior on existing implementations.
>
> Neither toybox nor busybox have -F imply -s. The behavior of -F matches -f, and
> you can use -D with -b or -f:
>
>    $ echo -e 'one two three\nand' | cut -d' ' -f3,2
>    two three
>    and
>    $ echo -e 'one two three\nand' | toybox cut -d' ' -Df3,2
>    three two

Cool, thanks for confirming that.

cheers,
Pádraig