Feature discussion: Search field and local search engine

Feature discussion: Search field and local search engine

[Sébastien Gendre]

[-- Attachment #1: Type: text/plain, Size: 2536 bytes --]

Hello every one,

* Context

At the beginning of September, I have started a discussion about adding
multiple new features to ox-html exporter. This discussion lead to also
discuss new features about org-mode itself.

To avoid the confusion of having multiples features discussed in the
same thread, following the suggestion of Ihor Radchenko, I will create a
separated thread for each discussed feature.

The original message can be found here:
https://list.orgmode.org/87frqbel30.fsf@localhost/



* Feature description and summary of previous discussion

The goal of this feature is to add, on a website generated with
org-publish, a local search engine.

The idea is to have a simple solution that can be easily enabled with an
org-publish option set to "t".

The search engine, it's search field and how the website is indexed is
gonna be implemented through a pluggable system. Like that, a user can
choose between different solutions. And if the chosen default solution
is no longer maintained, it's more easy to switch to another one.

The search field is gonna be included in a new section, present on each
web page. This section serve to website navigation, can be displayed at
top or side and will include:

- The exported website name and/or logo

- A website navigation menu (discussed in another thread I will create
  later)

- The search field



** Search engines

For now, the first search engine tested is PageFind:
https://pagefind.app/

Their was discussions about the risk of no longer maintained search
engine, that when Ihor Radchenko suggested the pluggable system.



* What's next on this feature ?

First, I opened this thread to discuss about how we want this search
engine feature to be.

In my next message of this thread, I will quotes remarks from Ihor
Radchenko, Max Nikulin and Orm Finnendahl to continue the discussion. I
will also include my replies.

When we have decided how this new feature should work, I will write some
patches to implement them. (I think I already sign the document for the
FSF).

Note that I'm on my last year as a student, so I'm may take some time to
reply to message and also write patches.


* And about the other features ?

How do you want to discuss the other features ?

One by one and only start to discuss the next one when the previous is
implemented ?

Or do you prefer I create new threads for each of themes in the next
days ?



Best regards

-------
Gendre Sébastien

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 849 bytes --]

Re: Feature discussion: Search field and local search engine

[Sébastien Gendre]

[-- Attachment #1: Type: text/plain, Size: 10367 bytes --]


Here is some quote of the previous discussion, with my reply below.



* Remarks #1, by Ihor Radchenko at Sat, 07 Sep 2024 11:53

Ihor Radchenko writes:
> > Sébastien Gendre writes:
> > I can search for another project.
> > 
> > I don't know how many work it would be needed to develop a search motor
> > specifically for Org-mode. But doing the indexing on Org-mode files could
> > let the user control the indexation of each page and section directly
> > from them. With buffer settings and heading properties.
> 
> Let me clarify.
> Unless an indexer is very simple, we do not really want it in Org - it
> will put a lot of extra load on maintainers.
> 
> On the other side, we do not want to tie things to some project that
> may fade out in 10 years into the future.
> 
> The way I see search engine support in Org mode is either:
> 
> 1. Using some really established project that we can expect to last for
>    many years.
> 
> 2. Implementing pluggable search support where users can choose which
>    indexer/searcher to use
> 
> (2) will be the best.
> 
> So, may you please search across available search engines and see what
> they have in common. Then, we can work out some infrastructure that is
> generic enough to plug a custom engine.
> 
> Then, pagefind can be the default (it is MIT license - GPL compatible),
> unless someone proposes a better alternative.

Yes, you are right, a pluggable search support is the best choice. I
include it in the summary of this discussion, on the first message of
this thread.

I will continue my search of other search engine and do a summary of
what they have in common. I will publish it on this thread, attached to
the first message.



* Remarks #2, from Max Nikulin at Sun, 8 Sep 2024 21:46 and 22:55

Max Nikulin writes:

> On 07/09/2024 18:53, Ihor Radchenko wrote:
> > Then, pagefind can be the default (it is MIT license - GPL compatible),
> 
> It might be more tricky:
> 
> <https://yhetil.org/emacs-devel/861q671idt.fsf@gnu.org>
> emacs-devel. Re: [Nicolas Graves] [PATCH v6 01/10] rde: emacs: Start 
> emacs in --daemon mode, with shepherd and pid-file. Sun, 12 May 2024 
> 12:36:46 +0300
> > Nicolas Graves:
> >> The code is given as MIT-0, hence also the two different licenses for
> >> the two functions sd_notify and sd_is_socket. Not an expert on licenses
> >> either, but with a proper flag about what this function's license is, I
> >> guess it should be fine, since other projects also do that.
> 
> Eli Zaretskii:
> > The license is only half of the problem.  Every non-trivial
> > contribution to Emacs must have its copyright assigned to the FSF,
> > because the FSF is in charge of protecting the Emacs sources,
> > something that only the copyright holder can do, at least in some
> > countries.  You will need to assign the copyright as well (a
> > relatively simple procedure of filling a form and emailing it), but if
> > the code is not yours, you cannot assign its copyright.

Max Nikulin writes:

> On 08/09/2024 21:46, Max Nikulin wrote:
> > On 07/09/2024 18:53, Ihor Radchenko wrote:
> >> Then, pagefind can be the default (it is MIT license - GPL compatible),
> > 
> > It might be more tricky:
> 
> Sorry for the noise. Of course, if you are not going to include any 
> pagefind code into Org then it is not an issue.

If we use PageFind, isn't possible to not include it with org-mode but
have an Elisp function that download it ? Like what Elpy do with its Python
dependencies ?



* Remarks #3, by Orm Finnendahl, at Sun, 8 Sep 2024 20:36

Orm Finnendahl writes:

> that makes no sense to me whatsoever: Post processing is already
> possible and built into org-export. pagefind is an external product
> with its own binaries, not written in elisp nor being by any means
> connected to emacs and compiling index files on generated HTML files
> is exactly that: A post process.
> 
> The javascript needed and all processing scripts can easily be
> included in the header, so I don't see any point in this, except
> writing a tutorial, how to integrate pagefind into someone's HTML
> output with the means already available with the existing backend.

It is already possible to add a search field and do PageFind indexation
and JS/CSS installation by set custom HTML in preamble and a custom
post-processing function. That what I have started to do.

But: It require the user to write custom HTML, custom Elisp function,
understand how PageFind work, etc.

The feature I suggest is to let the user having this search engine on its
website by simply set an org-publish option to "t".

A local search engine don't seems to be a niche feature. You have it
with Jupyter Book [1] or Read The Docs documentations [2]. As Org-mode
is a fantastic tool to write online documentation, having local search
engine easy to setup is a good feature. It's also very useful for blogs
to let visitor search information in old posts.

If this feature is simple to use, it let the user concentrate on writing
the content.

And regarding the inclusion, or not, of another software: We can have an
Elisp function that download PageFind ? If not, we can display a message
asking user to install it on its own if Emacs doesn't found PageFind on
$PATH.


Orm Finnendahl writes:

> And that's not even contemplating, why someone would want to throw a
> multipage site search indexer onto single page HTML output which
> doesn't work on static files opened from local disks ;-)

For the single HTML export, indexing is not necessary. But the inclusion
of a search field could be if the page have been generated separately
from the org-publish ?

For the last part about files opened from local disks, what do you mean ?



* Remarks #4, by Ihor Radchenko at 08 Sep 2024 18:42

Ihor Radchenko writes:
> Orm Finnendahl <orm.finnendahl@selma.hfmdk-frankfurt.de> writes:
> 
> > The javascript needed and all processing scripts can easily be
> > included in the header, so I don't see any point in this, except
> > writing a tutorial, how to integrate pagefind into someone's HTML
> > output with the means already available with the existing backend.
> >
> > And that's not even contemplating, why someone would want to throw a
> > multipage site search indexer onto single page HTML output which
> > doesn't work on static files opened from local disks ;-)
> 
> I agree that including indexer is just a question of adding specific
> javascript.
> 
> But I think that it would be useful to provide some default toggle to
> include such a tool without having to know the details of what js to
> include and where. Just as a simple user option.
> 
> It should probably work for single page as well. I do not see why not.

I agree with Ihor, the goal is to have a simple user option to enable
this feature.



* Remarks #5, by Orm Finnendahl at 8 Sep 2024 21:25

Orm Finnendahl writes:

> Am Sonntag, den 08. September 2024 um 18:42:51 Uhr (+0000) schrieb
> Ihor Radchenko:
> > I agree that including indexer is just a question of adding specific
> > javascript.
> > 
> > But I think that it would be useful to provide some default toggle to
> > include such a tool without having to know the details of what js to
> > include and where. Just as a simple user option.
> >
> 
>  whatever. The js is already included in the pagefind distribution, so
> it is a simple
> 
> #+HTML_HEAD: <script src="./pagefind/pagefind-ui.js"></script>
> 
> in the org file and the searchbar html in the preamble (or wherever).

This would require the user to write custom HTML code and also write a
custom Elisp function to launch the indexation and installation of
JS/CSS. And the user would also need to understand how PageFind work.

It's a lot of steps and require different skills.

But if we provide a simple option to enable in org-publish, this could
be very simple for the user to have a search engine.

Of course, we are not gonna support every use case of search engine,
only the most simple and useful one. And the pluggable system let to
user the possibility to hack it the way she or he want.


Orm Finnendahl writes:

> > It should probably work for single page as well. I do not see why
> > not.
> 
>  sure it works. I just question the raison d'etre, when single page
> search is already integrated into webbrowsers. But as always there
> will be people arguing that it is necessary to have a search bar with
> pop up results integrated into the page and of course there is nothing
> wrong with that. I use pagefind myself, but the site I'm working on
> (built with the multipage exporter BTW) currently contains more than
> 400 files where the browser search can't help.

A multi-pages search engine is of course useful the most with
multi-pages publication.

Having it for a singe page, I will see it as useful only if you export
one HTML file that you plan to include in an already existing
multi-files website. But it's maybe to specific for us to support it on
a single page export.



* Remarks #6, by Ihor Radchenko at Mon, 09 Sep 2024 16:40

Ihor Radchenko writes:

> Orm Finnendahl <orm.finnendahl@selma.hfmdk-frankfurt.de> writes:
> 
> >> It should probably work for single page as well. I do not see why
> >> not.
> >
> >  sure it works. I just question the raison d'etre, when single page
> > search is already integrated into webbrowsers. But as always there
> > will be people arguing that it is necessary to have a search bar with
> > pop up results integrated into the page and of course there is nothing
> > wrong with that. I use pagefind myself, but the site I'm working on
> > (built with the multipage exporter BTW) currently contains more than
> > 400 files where the browser search can't help.
> 
> Agree. So, having search functionality probably makes more sense within
> ox-publish, where we may also run post-processing to generate search
> terms or whatever extra work is needed to make the indexer work.

I 100% agree with you.


* EOF

That it, I hope to have forgotten no one.


Best regards

-------
Gendre Sébastien




[1] https://jupyterbook.org/

[2] Example here: https://slixmpp.readthedocs.io

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 849 bytes --]

Re: Feature discussion: Search field and local search engine

[Ihor Radchenko]

Sébastien Gendre <seb@k-7.ch> writes:

> If we use PageFind, isn't possible to not include it with org-mode but
> have an Elisp function that download it ? Like what Elpy do with its Python
> dependencies ?

It may be possible - Emacs already downloads tree-sitter grammars, for
example. But can you do it platform-independently, so that things keep
working on Windows, Linux, Guix, DOS (yes, Emacs still supports DOS),
what not?

-- 
Ihor Radchenko // yantar92,
Org mode contributor,
Learn more about Org mode at <https://orgmode.org/>.
Support Org development at <https://liberapay.com/org-mode>,
or support my work at <https://liberapay.com/yantar92>

Re: Feature discussion: Search field and local search engine

[Sébastien Gendre]

[-- Attachment #1: Type: text/plain, Size: 860 bytes --]


For the platform-independence, I think it would depend on the tool used.

PageFind precompiler release, on their Github, is available for x86_64
GNU/Linux, Apple Darwin and Windows and aarch64 GNU/Linux and Apple
Darwin.

But if we use a search indexer written in interpreted language, as long
as the interpreter is available, we can run it.

Ihor Radchenko <yantar92@posteo.net> writes:

> Sébastien Gendre <seb@k-7.ch> writes:
>
>> If we use PageFind, isn't possible to not include it with org-mode but
>> have an Elisp function that download it ? Like what Elpy do with its Python
>> dependencies ?
>
> It may be possible - Emacs already downloads tree-sitter grammars, for
> example. But can you do it platform-independently, so that things keep
> working on Windows, Linux, Guix, DOS (yes, Emacs still supports DOS),
> what not?

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 849 bytes --]

Re: Feature discussion: Search field and local search engine

[Ihor Radchenko]

Sébastien Gendre <seb@k-7.ch> writes:

> For the platform-independence, I think it would depend on the tool used.
>
> PageFind precompiler release, on their Github, is available for x86_64
> GNU/Linux, Apple Darwin and Windows and aarch64 GNU/Linux and Apple
> Darwin.
> ...
> But if we use a search indexer written in interpreted language, as long
> as the interpreter is available, we can run it.

Not trivial :)
We may be able to do it, and I do not see major obstacles as long as we
make this optional (some users frown upon binaries downloaded
automatically, while other users love such automation)

However, I'd first start from implementing the feature under assumption
that the indexer/search engine is already installed. We can add
auto-downloading after we have the rest.

-- 
Ihor Radchenko // yantar92,
Org mode contributor,
Learn more about Org mode at <https://orgmode.org/>.
Support Org development at <https://liberapay.com/org-mode>,
or support my work at <https://liberapay.com/yantar92>

Re: Feature discussion: Search field and local search engine

[Sébastien Gendre]

[-- Attachment #1: Type: text/plain, Size: 1370 bytes --]

Hello,

I'm back, for a few days. I've started to list candidates for a search
engine, compare their characteristics, etc.

But I need to define the criteria of selection.
Here is my suggestion:

- Libre Software

- Exist for more than a few years

- Still maintained

- Can index content from command line

- Can search with only frontend components (no backend needed)


What do you think ? Did I miss something ?



Best regards

-------
Gendre Sébastien


Ihor Radchenko <yantar92@posteo.net> writes:

> Sébastien Gendre <seb@k-7.ch> writes:
>
>> For the platform-independence, I think it would depend on the tool used.
>>
>> PageFind precompiler release, on their Github, is available for x86_64
>> GNU/Linux, Apple Darwin and Windows and aarch64 GNU/Linux and Apple
>> Darwin.
>> ...
>> But if we use a search indexer written in interpreted language, as long
>> as the interpreter is available, we can run it.
>
> Not trivial :)
> We may be able to do it, and I do not see major obstacles as long as we
> make this optional (some users frown upon binaries downloaded
> automatically, while other users love such automation)
>
> However, I'd first start from implementing the feature under assumption
> that the indexer/search engine is already installed. We can add
> auto-downloading after we have the rest.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 849 bytes --]

Re: Feature discussion: Search field and local search engine

[Ihor Radchenko]

Sébastien Gendre <seb@k-7.ch> writes:

> But I need to define the criteria of selection.
> Here is my suggestion:
>
> - Libre Software
>
> - Exist for more than a few years
>
> - Still maintained
>

+1

> - Can index content from command line
>
> - Can search with only frontend components (no backend needed)

These are probably not necessary, unless it makes your life easier for
the purposes of implementation.
 

-- 
Ihor Radchenko // yantar92,
Org mode contributor,
Learn more about Org mode at <https://orgmode.org/>.
Support Org development at <https://liberapay.com/org-mode>,
or support my work at <https://liberapay.com/yantar92>

Re: Feature discussion: Search field and local search engine

[Ihor Radchenko]

Sébastien Gendre <seb@k-7.ch> writes:

> * And about the other features ?
>
> How do you want to discuss the other features ?
>
> One by one and only start to discuss the next one when the previous is
> implemented ?
>
> Or do you prefer I create new threads for each of themes in the next
> days ?

If you can make them independent, especially with sparse free time, I
suggest focusing on one thing at a time. Or you risk spreading yourself
thin across many simultaneous discussions.

-- 
Ihor Radchenko // yantar92,
Org mode contributor,
Learn more about Org mode at <https://orgmode.org/>.
Support Org development at <https://liberapay.com/org-mode>,
or support my work at <https://liberapay.com/yantar92>