Re: vi: count occurrences of a word

O 04/09/21 ás 14:26, ropers escribiu:
> On 04/09/2021, Parodper <parodper@gmail.com> wrote:
>> To use newlines with sed I use tr and a char I know does not appear
>> on the text, like '|' or '`'. I just tested :!sed s/abc/abc€/g % |
>> tr '€' '\n' | grep -c abc and it worked fine.
>
> That's a neat trick -- IFF you can be *sure* that character won't
> show up in the text. I also feel it's a workaround, and I don't
> really understand *why* substituting/inserting newlines is something
> that seemingly cannot be made to work in OpenBSD sed, or whether the
> information on the sed(1) man page is really consistent with actual
> behaviour.
>
> Observed oddities:
>
> 1. ^V[return] generally yielding CR across the board (on BSD, Linux,
> etc.) even though the Unix-like end-of-line character is LF, not CR,
>
> 2. \n working perfectly fine in some base utilities like tr but not
> in others like sed, and
>
> 3. the sed(1) documentation being seemingly or potentially
> misleading or at odds with observed program behaviour w/r/t
> newlines.
>
> Is anyone able to make sense of this? Does anyone know if there's a
> reason or rationale behind the BSD sed implementation when it comes
> to newlines? Can someone perhaps even say why ^V[return] produces
> CR instead of LF or if there's even a way to enter \n in a way sed
> will accept?
>
> Thank you, Ian
>

I think I found something. From POSIX
(https://pubs.opengroup.org/onlinepubs/9699919799/utilities/sed.html):
> A line can be split by substituting a <newline> into it. The
> application shall escape the <newline> in the replacement by preceding
> it by a <backslash>.

So I wrote
:!sed s/abc/abc\/g % | grep -c abc
and then went back and pressed <ENTER> after that backslash, i.e.
:!sed s/abc/abc\<ENTER>/g % | grep -c abc
And it gave me a correct number of abc's for my test text.