Regeression in RE ("\%>")somewhere after 7.4.560

[Axel Bender]

1) Open the attached file "test.txt".
2) set ts=80
3) /\%>80vGesamtausabe

Contrary to the behavior in 560 (Windows 7, MinGW64), only the term "Gesamtausgabe" in the second line is marked.

[Axel Bender]

Addendum: Typo: Gesamtausabe => Gesamtausgabe

[John Little]

I see this in 7.4.580, linux amd64. In fact, the \%>Nv atom looks quite broken, given the line, and vim -u NONE -N:

abcdefgh

\%>2vc and \%2vd work, but \%>2ve does not. More generally, \%>Nv does not match beyond N + 2, and it used to.

Regards, John Little

[Christian Brabandt]

Hi Axel!

On Sa, 17 Jan 2015, Axel Bender wrote:

> 1) Open the attached file "test.txt".
> 2) set ts=80
> 3) /\%>80vGesamtausabe

I apologize for causing this.

Here is a fix (including a test).

Best,
Christian
--
Man braucht nicht geistreich zu sein, um zu beweisen, daß man begabt
ist. Aber man braucht viel Geist, um zu verbergen, daß man keine
Begabung hat.
-- Marcel Achard

[Bram Moolenaar]

Christian Brabandt wrote:

> Hi Axel!
>
> On Sa, 17 Jan 2015, Axel Bender wrote:
>
> > 1) Open the attached file "test.txt".
> > 2) set ts=80
> > 3) /\%>80vGesamtausabe
>
> I apologize for causing this.
>
> Here is a fix (including a test).

Thanks. It fixes the problem reported, but using vcol_prev won't work
properly. If the engine tries this state at another position, or
another state, it will have the wrong value.

Not sure how to remove vcol_prev without making it slow again.
A rough guess is to multiply the column with 'tabstop'.
Let me do that for now.

--
From "know your smileys":
:-F Bucktoothed vampire with one tooth missing

/// Bram Moolenaar -- Br...@Moolenaar.net -- http://www.Moolenaar.net \\\
/// sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ \\\
\\\ an exciting new programming language -- http://www.Zimbu.org ///
\\\ help me help AIDS victims -- http://ICCF-Holland.org ///

[Bram Moolenaar]

I wrote:

> Christian Brabandt wrote:
>
> > Hi Axel!
> >
> > On Sa, 17 Jan 2015, Axel Bender wrote:
> >
> > > 1) Open the attached file "test.txt".
> > > 2) set ts=80
> > > 3) /\%>80vGesamtausabe
> >
> > I apologize for causing this.
> >
> > Here is a fix (including a test).
>
> Thanks. It fixes the problem reported, but using vcol_prev won't work
> properly. If the engine tries this state at another position, or
> another state, it will have the wrong value.
>
> Not sure how to remove vcol_prev without making it slow again.
> A rough guess is to multiply the column with 'tabstop'.
> Let me do that for now.

Thinking about this again: It should be OK to store the column and the
computed virtual column. These do not depend on the RE state. When
trying several states at the same position the virtual column will be
computed only once. And when the column is higher than the cached
column, we know the virtual column is also higher than the cached
virtual column. That should help avoid useless calls to
win_linetabsize() while still being accurate, instead of the wild guess
using 'tabsize'.

Christian, do you think you could work on this? And write a test.

--
From "know your smileys":

<|-) Chinese
<|-( Chinese and doesn't like these kind of jokes

[Christian Brabandt]

Hi Bram!

On Di, 20 Jan 2015, Bram Moolenaar wrote:

> Thinking about this again: It should be OK to store the column and the
> computed virtual column. These do not depend on the RE state. When
> trying several states at the same position the virtual column will be
> computed only once. And when the column is higher than the cached
> column, we know the virtual column is also higher than the cached
> virtual column. That should help avoid useless calls to
> win_linetabsize() while still being accurate, instead of the wild guess
> using 'tabsize'.

We would also need this for the old engine, don't we?

>
> Christian, do you think you could work on this? And write a test.

I'd like to, but I do not understand the NFA engine good enough to do
this.

Best,
Christian
--
Tausch nicht das Leben gegen den Traum. Besiegen kann man nur die
Wirklichkeit.
-- Lucius Annaeus Seneca (4-65 n.Chr.)