Conkeror-like functionality for EWW

Conkeror-like functionality for EWW

[Andrey Kotlarski]

here https://github.com/m00natic/eww-lnum

There were voices before for such, so if it seems worthy, I'd gladly add
it to GNU ELPA.  For those who wonder what this conkeror shmazi-pazi is
- sort of combination between Isearch and ace-jump that operates only
over links, buttons and forms.  It makes for quick and precise
alternative to mouse browsing.

Re: Conkeror-like functionality for EWW

[Ted Zlatanov]

On Sun, 11 May 2014 00:40:24 +0300 Andrey Kotlarski <m00naticus@gmail.com> wrote: 

AK> here https://github.com/m00natic/eww-lnum
AK> There were voices before for such, so if it seems worthy, I'd gladly add
AK> it to GNU ELPA.  For those who wonder what this conkeror shmazi-pazi is
AK> - sort of combination between Isearch and ace-jump that operates only
AK> over links, buttons and forms.  It makes for quick and precise
AK> alternative to mouse browsing.

Hi Andrey,

your code is neat and useful. Could it integrate with the existing
`next-error' and `previous-error' Emacs functionality, though? That's
not just for errors, but for visiting points of interest in general.

Many people have keys bound to those functions and they would make sense
in your context.

Ted

Re: Conkeror-like functionality for EWW

[Óscar Fuentes]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Sun, 11 May 2014 00:40:24 +0300 Andrey Kotlarski <m00naticus@gmail.com> wrote: 
>
> AK> here https://github.com/m00natic/eww-lnum
> AK> There were voices before for such, so if it seems worthy, I'd gladly add
> AK> it to GNU ELPA.  For those who wonder what this conkeror shmazi-pazi is
> AK> - sort of combination between Isearch and ace-jump that operates only
> AK> over links, buttons and forms.  It makes for quick and precise
> AK> alternative to mouse browsing.
>
> Hi Andrey,
>
> your code is neat and useful. Could it integrate with the existing
> `next-error' and `previous-error' Emacs functionality, though? That's
> not just for errors, but for visiting points of interest in general.
>
> Many people have keys bound to those functions and they would make sense
> in your context.

I don't think so. next-error/previous-error are for traversing
sequences, while the OP's package is for directly jumping to/acting upon
links.

next-error/previous-error would require to "well order" the set of links
on a web page, on a way that is both correct and intuitive for the user.
That's far from trivial on the general case. OTOH, what the OP's package
does is simpler and more effective at the same time. Traversing a
sequence on links can be ok on Info pages, but it is a pain on web
pages, except the most simple ones.

Re: Conkeror-like functionality for EWW

[Ted Zlatanov]

On Mon, 03 Nov 2014 15:07:04 +0100 Óscar Fuentes <ofv@wanadoo.es> wrote: 

ÓF> I don't think so. next-error/previous-error are for traversing
ÓF> sequences, while the OP's package is for directly jumping to/acting upon
ÓF> links.

I think those are closely related.

ÓF> next-error/previous-error would require to "well order" the set of links
ÓF> on a web page, on a way that is both correct and intuitive for the user.
ÓF> That's far from trivial on the general case.

I don't know why that's so hard, but at least in EWW we have
`shr-next-link' and `shr-previous-link' to do this kind of traversal.

ÓF> Traversing a sequence on links can be ok on Info pages, but it is a
ÓF> pain on web pages, except the most simple ones.

I find it useful with the already-bound TAB in EWW buffers.

Ted

Re: Conkeror-like functionality for EWW

[Andrey Kotlarski]

[  3 ноември 2014, 08:42 -0500, понеделник ] Ted Zlatanov:

> Could it integrate with the existing `next-error' and `previous-error'
> Emacs functionality, though? That's not just for errors, but for
> visiting points of interest in general.

I've been meaning to generalize this beyond eww and emacs-w3m indeed.
Grep, compile, info, which provide sequential access with next/previous
procedures, could have similar "jump immediately to particular
link/something".  The common functionality has to be extracted and made
easy to extend for other modes/packages.  I have it as TODO for long
while.

Re: Conkeror-like functionality for EWW

[Lars Magne Ingebrigtsen]

Ted Zlatanov <tzz@lifelogs.com> writes:

> AK> here https://github.com/m00natic/eww-lnum
> AK> There were voices before for such, so if it seems worthy, I'd gladly add
> AK> it to GNU ELPA.  For those who wonder what this conkeror shmazi-pazi is
> AK> - sort of combination between Isearch and ace-jump that operates only
> AK> over links, buttons and forms.  It makes for quick and precise
> AK> alternative to mouse browsing.
>
> Hi Andrey,
>
> your code is neat and useful. Could it integrate with the existing
> `next-error' and `previous-error' Emacs functionality, though? That's
> not just for errors, but for visiting points of interest in general.

I just had another look at the code.  If I understand it correctly, you
hit `f', numbers are assigned to the links, and then you enter the
number of the link you want to visit?

I'm afraid I don't find this compelling functionality.  Anything that
makes me read numbers from the screen and then type them in doesn't seem
like something a lot of people would want to do when they can just mouse
or tab to the link.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Conkeror-like functionality for EWW

[Óscar Fuentes]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

>> Hi Andrey,
>>
>> your code is neat and useful. Could it integrate with the existing
>> `next-error' and `previous-error' Emacs functionality, though? That's
>> not just for errors, but for visiting points of interest in general.
>
> I just had another look at the code.  If I understand it correctly, you
> hit `f', numbers are assigned to the links, and then you enter the
> number of the link you want to visit?

I'm not the OP, but if his tool behaves like Conkeror/Vimperator, that's
how it works.

> I'm afraid I don't find this compelling functionality.  Anything that
> makes me read numbers from the screen and then type them in doesn't seem
> like something a lot of people would want to do when they can just mouse
> or tab to the link.

Firefox and Chrome have extensions for this kind of functionality
(Conkeror, Vimperator, Vimium...) They are very popular, with hundreds
of thousands of downloads. I'm a Vimperator user because reaching for
the mouse for clicking a link is anti-ergonomic and slow. I'm also a
happy eww user and I do not miss Conkeror-like navigation there, but
that is just because the type of web pages I visit with eww. For the
rest, I'll rather switch to another web browser than stop using
Conkeror-like navigation.

Re: Conkeror-like functionality for EWW

[Gregor Zattler]

Hi Lars, emacs developers,
* Lars Magne Ingebrigtsen <larsi@gnus.org> [03. Nov. 2014]:
> I just had another look at the code.  If I understand it correctly, you
> hit `f', numbers are assigned to the links, and then you enter the
> number of the link you want to visit?
> 
> I'm afraid I don't find this compelling functionality.  Anything that
> makes me read numbers from the screen and then type them in doesn't seem
> like something a lot of people would want to do when they can just mouse
> or tab to the link.

Actually I used such interface provided as an add-on to firefox
for quite a while.  It's great if suffering from mouse arm
[coll.] [repetitive strain injury syndrome].  

Ciao, Gregor
-- 
 -... --- .-. . -.. ..--.. ...-.-

Re: Conkeror-like functionality for EWW

[Ted Zlatanov]

On Mon, 03 Nov 2014 18:20:06 +0100 Lars Magne Ingebrigtsen <larsi@gnus.org> wrote: 

LMI> Ted Zlatanov <tzz@lifelogs.com> writes:
AK> here https://github.com/m00natic/eww-lnum
AK> There were voices before for such, so if it seems worthy, I'd gladly add
AK> it to GNU ELPA.  For those who wonder what this conkeror shmazi-pazi is
AK> - sort of combination between Isearch and ace-jump that operates only
AK> over links, buttons and forms.  It makes for quick and precise
AK> alternative to mouse browsing.
>> 
>> Hi Andrey,
>> 
>> your code is neat and useful. Could it integrate with the existing
>> `next-error' and `previous-error' Emacs functionality, though? That's
>> not just for errors, but for visiting points of interest in general.

LMI> I just had another look at the code.  If I understand it correctly, you
LMI> hit `f', numbers are assigned to the links, and then you enter the
LMI> number of the link you want to visit?

LMI> I'm afraid I don't find this compelling functionality.  Anything that
LMI> makes me read numbers from the screen and then type them in doesn't seem
LMI> like something a lot of people would want to do when they can just mouse
LMI> or tab to the link.

I'd like `next-error' and `previous-error' bound, if that's OK with you.
They can take a numeric argument to indicate how far to move in either
direction.  That would be useful either way, right?

Absolute jumps definitely have their uses, so I think Andrey's proposal
is a good addition. But it could just be a call to `next-error-function'
with the RESET parameter (which just means "go to (point-min) before
applying the numeric argument"); it doesn't need special machinery. I
can implement that if you agree; it's not much work.

Ted

Re: Conkeror-like functionality for EWW

[Eric Abrahamsen]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> Ted Zlatanov <tzz@lifelogs.com> writes:
>
>> AK> here https://github.com/m00natic/eww-lnum
>> AK> There were voices before for such, so if it seems worthy, I'd gladly add
>> AK> it to GNU ELPA.  For those who wonder what this conkeror shmazi-pazi is
>> AK> - sort of combination between Isearch and ace-jump that operates only
>> AK> over links, buttons and forms.  It makes for quick and precise
>> AK> alternative to mouse browsing.
>>
>> Hi Andrey,
>>
>> your code is neat and useful. Could it integrate with the existing
>> `next-error' and `previous-error' Emacs functionality, though? That's
>> not just for errors, but for visiting points of interest in general.
>
> I just had another look at the code.  If I understand it correctly, you
> hit `f', numbers are assigned to the links, and then you enter the
> number of the link you want to visit?
>
> I'm afraid I don't find this compelling functionality.  Anything that
> makes me read numbers from the screen and then type them in doesn't seem
> like something a lot of people would want to do when they can just mouse
> or tab to the link.

I'm a conkeror user, and find this system a wonderful way to avoid the
mouse. One key feature is that, in addition to typing the number, you
can type a substring of the anchor text, and conkeror will progressively
narrow down the set of matching links, often until there's only one, and
you just hit RET. For certain types of website, you could wear out your
TAB key before you found the link you wanted.

Even farther afield, but worth considering: this link-selection behavior
works exactly the same across many different functions: copying link
hrefs, downloading linked files, following links, following links in a
new tab, etc etc. It's pretty nice.

Eric

Re: Conkeror-like functionality for EWW

[Lars Magne Ingebrigtsen]

Ted Zlatanov <tzz@lifelogs.com> writes:

> I'd like `next-error' and `previous-error' bound, if that's OK with you.
> They can take a numeric argument to indicate how far to move in either
> direction.  That would be useful either way, right?

Well, `C-u 3 M-x next-error' would just be the same as `3 TAB', right?
So if it's not absolute, it seems even less useful, and we already have
that functionality on `TAB'...

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Conkeror-like functionality for EWW

[Lars Magne Ingebrigtsen]

Eric Abrahamsen <eric@ericabrahamsen.net> writes:

> I'm a conkeror user, and find this system a wonderful way to avoid the
> mouse. One key feature is that, in addition to typing the number, you
> can type a substring of the anchor text, and conkeror will progressively
> narrow down the set of matching links, often until there's only one, and
> you just hit RET. For certain types of website, you could wear out your
> TAB key before you found the link you wanted.

I haven't used Conkeror, but it's basically a browser like Firefox,
right?  So you don't have a cursor that you can move around?

In that case, jumping to numbered links makes more sense, but in Emacs,
you can just hold down the `down' key a bit to move past masses of
links, or whatever.  Moving around in en Emacs buffer is usually not
really that much work.

In Firefox, though, it's pretty awful.  If you don't use the mouse
there, you have to TAB forever.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Conkeror-like functionality for EWW

[Ted Zlatanov]

On Mon, 03 Nov 2014 20:15:29 +0100 Lars Magne Ingebrigtsen <larsi@gnus.org> wrote: 

LMI> Ted Zlatanov <tzz@lifelogs.com> writes:
>> I'd like `next-error' and `previous-error' bound, if that's OK with you.
>> They can take a numeric argument to indicate how far to move in either
>> direction.  That would be useful either way, right?

LMI> Well, `C-u 3 M-x next-error' would just be the same as `3 TAB', right?
LMI> So if it's not absolute, it seems even less useful, and we already have
LMI> that functionality on `TAB'...

For me personally, the keypad / and * keys are bound to next/previous
error. It's easier than S-TAB to move backwards, and IMO a little easier
than TAB to move forwards. Many modes don't even bind TAB because it's
an editing command. So it's a personal preference. I can certainly add
next/previous-error support in a hook if you don't want it in EWW
itself.

I rarely use the numeric arguments, so I have no opinion there...

Ted

Re: Conkeror-like functionality for EWW

[Eric Abrahamsen]

On 11/03/14 20:20 PM, Lars Magne Ingebrigtsen wrote:
> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>> I'm a conkeror user, and find this system a wonderful way to avoid the
>> mouse. One key feature is that, in addition to typing the number, you
>> can type a substring of the anchor text, and conkeror will progressively
>> narrow down the set of matching links, often until there's only one, and
>> you just hit RET. For certain types of website, you could wear out your
>> TAB key before you found the link you wanted.
>
> I haven't used Conkeror, but it's basically a browser like Firefox,
> right?  So you don't have a cursor that you can move around?
>
> In that case, jumping to numbered links makes more sense, but in Emacs,
> you can just hold down the `down' key a bit to move past masses of
> links, or whatever.  Moving around in en Emacs buffer is usually not
> really that much work.
>
> In Firefox, though, it's pretty awful.  If you don't use the mouse
> there, you have to TAB forever.

Yeah, it's actually built on xulrunner, which is firefox's guts. But
it's heavily influenced by emacs (uses a bunch of the same keybindings,
concepts and terminology), so cross-pollination would be amusing. Not
using the mouse at all is a stated goal.

Obviously some people will find link-jumping crucial, others not. But
those who like it *really* like it.

Perhaps a plug-in for ace-jump-mode would be sufficient.

Eric

Re: Conkeror-like functionality for EWW

[Andrey Kotlarski]

[  3 ноември 2014, 12:13 -0800, понеделник ] Eric Abrahamsen:

>> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>>
>>> I'm a conkeror user, and find this system a wonderful way to avoid the
>>> mouse. One key feature is that, in addition to typing the number, you
>>> can type a substring of the anchor text, and conkeror will progressively
>>> narrow down the set of matching links, often until there's only one, and
>>> you just hit RET. For certain types of website, you could wear out your
>>> TAB key before you found the link you wanted.
> ...
> Perhaps a plug-in for ace-jump-mode would be sufficient.

Woops, sorry to interrupt but eww-lnum.el does just that (modulo it's
not ace-jump plugin).  You may even not need hit RET with single match
left.

Re: Conkeror-like functionality for EWW

[Eric Abrahamsen]

Andrey Kotlarski <m00naticus@gmail.com> writes:

> [  3 ноември 2014, 12:13 -0800, понеделник ] Eric Abrahamsen:
>
>>> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>>>
>>>> I'm a conkeror user, and find this system a wonderful way to avoid the
>>>> mouse. One key feature is that, in addition to typing the number, you
>>>> can type a substring of the anchor text, and conkeror will progressively
>>>> narrow down the set of matching links, often until there's only one, and
>>>> you just hit RET. For certain types of website, you could wear out your
>>>> TAB key before you found the link you wanted.
>> ...
>> Perhaps a plug-in for ace-jump-mode would be sufficient.
>
> Woops, sorry to interrupt but eww-lnum.el does just that (modulo it's
> not ace-jump plugin).  You may even not need hit RET with single match
> left.

Oh sure, I'd just strayed too far from the original topic. I was
originally posting all this in support of your plugin!

Re: Conkeror-like functionality for EWW

[Yuri Khan]

On Tue, Nov 4, 2014 at 1:20 AM, Lars Magne Ingebrigtsen <larsi@gnus.org> wrote:
>
> In Firefox, though, it's pretty awful.  If you don't use the mouse
> there, you have to TAB forever.

Actually in Firefox you can press F7 and it gives you a cursor that
you can move with cursor navigation keys.

But it’s still pretty awful, because tables.

But Firefox also has incremental search for general text on / [slash]
and for links on ' [straight single quote]; pretty close to Emacs’
isearch.

Re: Conkeror-like functionality for EWW

[Tassilo Horn]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> I just had another look at the code.  If I understand it correctly,
> you hit `f', numbers are assigned to the links, and then you enter the
> number of the link you want to visit?
>
> I'm afraid I don't find this compelling functionality.  Anything that
> makes me read numbers from the screen and then type them in doesn't
> seem like something a lot of people would want to do when they can
> just mouse or tab to the link.

If the displayed page contains dozens of links, hitting f 20 RET can be
much better than TAB twenty times or navigating to the link.  Note that
C-u 20 TAB is usually not workable because it requires you to know that
the link you want to follow is exactly 20 links after point.

There are similar plugins that assign characters to links instead of
numbers, so the first 26 visible links are just one key away.

Or another nice and similar approach was to have a command
`eww-isearch-links' which does what isearch does but considers only link
texts.  Probably, that would also be easier to implement as you don't
need to add dozens of overlays displaying link numbers/chars.

Bye,
Tassilo

Re: Conkeror-like functionality for EWW

[Andrey Kotlarski]

[  4 ноември 2014, 09:54 +0100, вторник ] Tassilo Horn:

> Or another nice and similar approach was to have a command
> `eww-isearch-links' which does what isearch does but considers only link
> texts.  Probably, that would also be easier to implement as you don't
> need to add dozens of overlays displaying link numbers/chars.

Just to mention, eww-lnum does this too, just like Conkeror.  You can
type characters and matching links will be narrowed down (and still
assigned numbers).

Sometimes I go to pages with too many links and just know I'm looking
for "Doc[s]" section or something.  I don't even bother looking at the
page, just press `f`, type "doc" and immediately see what's of interest.
Often in this situation all left is to hit RET or if there are several
matches left and the right one is not selected, just hit a digit.

Re: Conkeror-like functionality for EWW

[Tassilo Horn]

Andrey Kotlarski <m00naticus@gmail.com> writes:

Hi Andrey,

>> Or another nice and similar approach was to have a command
>> `eww-isearch-links' which does what isearch does but considers only
>> link texts.  Probably, that would also be easier to implement as you
>> don't need to add dozens of overlays displaying link numbers/chars.
>
> Just to mention, eww-lnum does this too, just like Conkeror.  You can
> type characters and matching links will be narrowed down (and still
> assigned numbers).

Oh, cool.

> Sometimes I go to pages with too many links and just know I'm looking
> for "Doc[s]" section or something.  I don't even bother looking at the
> page, just press `f`, type "doc" and immediately see what's of
> interest.  Often in this situation all left is to hit RET or if there
> are several matches left and the right one is not selected, just hit a
> digit.

Well, one think that I don't like is that `f <something>' only considers
the links that are currently visible.  For example, when visiting the
Emacs homepage and point being on top, I expect that `f imag' will show
me the "ImageMagick" link in the Releases section.  But since that's not
in the visible buffer portion, I can only type `f i' and the link "JOIN
THE FSF" is highlighted as first match.  I can't even input the next
char of "imagemagick", i.e., m, anymore, because there's no such match
in the visible part of the page.

So basically I like the feature a lot but I'd prefer if it operated on
the whole page, not only the visible buffer portion.  That is, I think
it would be ok if the numbers started with 1 at the beginnig of the
visible buffer portion but extended to the end of the page so that `f
<something>' has a search-forward behavior without bound.

Bye,
Tassilo

Re: Conkeror-like functionality for EWW

[Andrey Kotlarski]

The following message is a courtesy copy of an article
that has been posted to gmane.emacs.devel as well.

[  5 ноември 2014, 11:18 +0100, сряда ] Tassilo Horn:

> Well, one think that I don't like is that `f <something>' only considers
> the links that are currently visible.  For example, when visiting the
> Emacs homepage and point being on top, I expect that `f imag' will show
> me the "ImageMagick" link in the Releases section.  But since that's not
> in the visible buffer portion, I can only type `f i' and the link "JOIN
> THE FSF" is highlighted as first match.  I can't even input the next
> char of "imagemagick", i.e., m, anymore, because there's no such match
> in the visible part of the page.

Yes, this is sore.

> So basically I like the feature a lot but I'd prefer if it operated on
> the whole page, not only the visible buffer portion.  That is, I think
> it would be ok if the numbers started with 1 at the beginnig of the
> visible buffer portion but extended to the end of the page so that `f
> <something>' has a search-forward behavior without bound.

Problem is user can't refer to something invisible (by number or further
narrow) unless magical powers.  However, in case narrowing filters out
all visible links, automatic scrolling could happen until/if there are
matches further down.  I'll give it a go these days.

Re: Conkeror-like functionality for EWW

[Tassilo Horn]

Andrey Kotlarski <m00naticus@gmail.com> writes:

Hi Andrey,

>> So basically I like the feature a lot but I'd prefer if it operated
>> on the whole page, not only the visible buffer portion.  That is, I
>> think it would be ok if the numbers started with 1 at the beginnig of
>> the visible buffer portion but extended to the end of the page so
>> that `f <something>' has a search-forward behavior without bound.
>
> Problem is user can't refer to something invisible (by number or
> further narrow) unless magical powers.

Well, a user might know what he's looking for.  Or do you mean that
statement in a more technical sense?

In the latter case, `helm-occur' does something similar namely
dynamically narrowing down a buffer to lines that match a given regex as
you type it.  It does that in a separate buffer and prefixes the the
matching lines with some metadata (file, line), but I think that should
be doable also on the current buffer.

> However, in case narrowing filters out all visible links, automatic
> scrolling could happen until/if there are matches further down.  I'll
> give it a go these days.

Yeah, that would be an alternative approach.

Bye,
Tassilo

Re: Conkeror-like functionality for EWW

[Andrey Kotlarski]

The following message is a courtesy copy of an article
that has been posted to gmane.emacs.devel as well.

[  5 ноември 2014, 12:17 +0100, сряда ] Tassilo Horn:

> Well, a user might know what he's looking for.  Or do you mean that
> statement in a more technical sense?

If there are matches further down, they have to be shown somehow so user
doesn't blindly RET on the first without knowing what it is.  This
interface makes sense only on visible part of page.

> In the latter case, `helm-occur' does something similar namely
> dynamically narrowing down a buffer to lines that match a given regex as
> you type it.  It does that in a separate buffer and prefixes the the
> matching lines with some metadata (file, line), but I think that should
> be doable also on the current buffer.

helm-occur is very nice (probably even separate help-eww-links source)
but it shows just one line context and somewhat restricted way to jump
directly to particular match if many.  It is good in many cases but
sufficiently different than the conkeror-style interface to somehow
combine them.

By the way helm integration is something I'd like to do for the
eww-lnum-universal command where link/form/button is first selected and
then number of actions offered.

Re: Conkeror-like functionality for EWW

[Tassilo Horn]

Andrey Kotlarski <m00naticus@gmail.com> writes:

>> Well, a user might know what he's looking for.  Or do you mean that
>> statement in a more technical sense?
>
> If there are matches further down, they have to be shown somehow so
> user doesn't blindly RET on the first without knowing what it is.
> This interface makes sense only on visible part of page.

Yes, sure.  So matches further down have to be made visible, e.g., by
excluding lines/links that don't match the current input string.  That
is, `f i' might not show the ImageMagick link already but further typing
`f imag' probably does.

>> In the latter case, `helm-occur' does something similar namely
>> dynamically narrowing down a buffer to lines that match a given regex
>> as you type it.  It does that in a separate buffer and prefixes the
>> the matching lines with some metadata (file, line), but I think that
>> should be doable also on the current buffer.
>
> helm-occur is very nice (probably even separate help-eww-links source)
> but it shows just one line context and somewhat restricted way to jump
> directly to particular match if many.  It is good in many cases but
> sufficiently different than the conkeror-style interface to somehow
> combine them.

Agreed.  I just wanted to point out that narrowing down to show only
lines with links that match a given input in a dynamic as-you-type
manner looks feasible.  Of course, it would still be nice to have
conkeror-style numbers and the first matching link selected.

> By the way helm integration is something I'd like to do for the
> eww-lnum-universal command where link/form/button is first selected
> and then number of actions offered.

Yeah, cool.

Bye,
Tassilo

Re: Conkeror-like functionality for EWW

[Andrey Kotlarski]

[  5 ноември 2014, 12:48 +0200, сряда ] Andrey Kotlarski:

> [  5 ноември 2014, 11:18 +0100, сряда ] Tassilo Horn:
>
>> Well, one think that I don't like is that `f <something>' only considers
>> the links that are currently visible.  For example, when visiting the
>> Emacs homepage and point being on top, I expect that `f imag' will show
>> me the "ImageMagick" link in the Releases section.  But since that's not
>> in the visible buffer portion, I can only type `f i' and the link "JOIN
>> THE FSF" is highlighted as first match.  I can't even input the next
>> char of "imagemagick", i.e., m, anymore, because there's no such match
>> in the visible part of the page.
>
> Yes, this is sore.
>
>> So basically I like the feature a lot but I'd prefer if it operated on
>> the whole page, not only the visible buffer portion.  That is, I think
>> it would be ok if the numbers started with 1 at the beginnig of the
>> visible buffer portion but extended to the end of the page so that `f
>> <something>' has a search-forward behavior without bound.
>
> Problem is user can't refer to something invisible (by number or further
> narrow) unless magical powers.  However, in case narrowing filters out
> all visible links, automatic scrolling could happen until/if there are
> matches further down.  I'll give it a go these days.

For the record, this has been implemented and also searches backward in
case there are no forward matches.

Re: Conkeror-like functionality for EWW

[Tassilo Horn]

Andrey Kotlarski <m00naticus@gmail.com> writes:

Hi Andrey,

>>> So basically I like the feature a lot but I'd prefer if it operated
>>> on the whole page, not only the visible buffer portion.  That is, I
>>> think it would be ok if the numbers started with 1 at the beginnig
>>> of the visible buffer portion but extended to the end of the page so
>>> that `f <something>' has a search-forward behavior without bound.
>>
>> Problem is user can't refer to something invisible (by number or
>> further narrow) unless magical powers.  However, in case narrowing
>> filters out all visible links, automatic scrolling could happen
>> until/if there are matches further down.  I'll give it a go these
>> days.
>
> For the record, this has been implemented and also searches backward
> in case there are no forward matches.

Just tried it out, and it works like a charm.

Thanks,
Tassilo