Re: [lmi] On matching the zero-width vim atom "%V"

[Vadim Zeitlin]

On Mon, 25 Jan 2021 21:55:21 +0000 Greg Chicares <gchicares@sbcglobal.net> 
wrote:

GC> Here's something that I think should work, but doesn't.
GC> 
GC> Goal: in this line:
GC>   InvariantValues().TermSpecAmt[j] = TermSpecAmt;
GC> I want to call dblize() on the RHS, making it this:
GC>   InvariantValues().TermSpecAmt[j] = dblize(TermSpecAmt);

 I know you didn't ask for this, but FWIW here is how I'd do it:

1. Put the cursor anywhere on TermSpecAmt.
2. Do "s/<C-R><C-W>/dblize(&)".

 This might not be the optimal way to do it from the key stroke count
point of view but I find it more convenient than using visual mode (but, in
the interest of full disclosure, I'm cognizant of the fact that I don't use
visual mode enough and should use it more because some things are really
better done with it).

GC> Technique: visually highlight the string to be replaced,
GC> e.g. with this sequence of commands on that line:
GC>   $FTvt;
GC> and then perform the substitution:
GC>   :'<,'>s/\%V.*\%V/dblize(&)/
GC> (The "$FT" part is written to make the command easy to
GC> reproduce--it's not the way I'd normally begin a selection.)

 Another unsolicited advice: I find the fastest way to make such a
selection is by using "viw" or similar.

GC> That fails, yielding:
GC>   InvariantValues().TermSpecAmt[j] = dblize(TermSpecAm)t;
GC> 
GC> It works if I highlight one more character ('f', not 't'):
GC>   $FTvf;
GC>   :'<,'>s/\%V.*\%V/dblize(&)/
GC>   InvariantValues().TermSpecAmt[j] = dblize(TermSpecAmt);
GC> But earlier I excluded that terminal ';' from the selection
GC> because I don't want to operate on it--I don't want
GC>   InvariantValues().TermSpecAmt[j] = dblize(TermSpecAmt;)
GC> which is what I would expect when highlighting includes ';'.
GC> 
GC> Similarly, if the string to be replaced extends to EOL:
GC>   ,NetPmts[Month]
GC> then "$FNvg_" highlights the string I want to replace,
GC> but to use the "s" command above I must instead highlight
GC> it with "$FNv$", which, jarringly, extends the selection
GC> for two (not one) pseudo characters past the end.
GC> 
GC> (I don't think this is caused by any unusual vim setting,
GC> but lmi's 'gwc/.vimrc' gives my full configuration).

 I can confirm that I can reproduce this behaviour and that I see it even
with "vim -u NONE".

GC> What am I misunderstanding?

 As I said, I don't use visual mode often, but it looks like \%V is meant
to work like this, at least the example for it in the help says:

        [...] To make sure the whole pattern is
        inside the Visual area put it at the start and just before the end of
        the pattern, e.g.: >
                /\%Vfoo.*ba\%Vr

so it explicitly says to use "/\%V.*\%V./" in your case. I have no good
explanation for this behaviour, but I think I at least have a slightly less
counterintuitive workaround: you could also use "/\(\%V.\)\+/" which I read
as "match any sequence of characters inside the visual area".

 Regards,
VZ

pgpl820v1ZO_b.pgp
Description: PGP signature