webgen users - Mar 2009 - Markdown support?

I''ve considered switching from Ikiwiki to Webgen, particularly
because the former doesn''t work under Cygwin. There''s a
problem,
however: Webgen doesn''t seem to support proper markdown syntax.
In particular, Maruku seems to require (bullet) lists to be
unintended, which makes for extra conversion work and unreadable
plain text documents. (I have READMEs and such in markdown syntax 
that I convert for the Web. They sould be readable as plain text.)
Changing the processing pipeline to mention markdown doesn''t help.

Is it possible to use a real markdown engine?

-- 
Stop Gnomes and other pests! Purchase Windows today!

I know what you mean.  I tried to switch to Maruku, failed horribly,
wasted a lot of time, and finally stuck with RedCloth.

-- Chad

On Thu, Mar 12, 2009 at 10:42 AM, Tuomo Valkonen <tuomov at iki.fi>
wrote:
>
> I''ve considered switching from Ikiwiki to Webgen, particularly
> because the former doesn''t work under Cygwin. There''s a
problem,
> however: Webgen doesn''t seem to support proper markdown syntax.
> In particular, Maruku seems to require (bullet) lists to be
> unintended, which makes for extra conversion work and unreadable
> plain text documents. (I have READMEs and such in markdown syntax
> that I convert for the Web. They sould be readable as plain text.)
> Changing the processing pipeline to mention markdown doesn''t help.
>
> Is it possible to use a real markdown engine?
>
> --
> Stop Gnomes and other pests! Purchase Windows today!
>
> _______________________________________________
> webgen-users mailing list
> webgen-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/webgen-users
>

On 2009-03-13, Chad Woolley <thewoolleyman at gmail.com>
wrote:
> I know what you mean.  I tried to switch to Maruku, failed horribly,
> wasted a lot of time, and finally stuck with RedCloth.
I actually seem to have found the solution: rdiscount.
It renders my pages mostly allright. Now, all I need
is an ''inline'' extension for collecting FAQ items and
news on a single page... Or maybe there''s another
solution? (Feeds wanted as well:)

-- 
"[Fashion] is usually a form of ugliness so intolerable that we have
 to alter it every six months." -- Oscar Wilde
"The computer industry is the only industry that is more fashion-driven
 than women''s fashion." -- RMS

Am Fri, 13 Mar 2009 01:07:17 +0000 (UTC)
schrieb Tuomo Valkonen <tuomov at iki.fi>:
> On 2009-03-13, Chad Woolley <thewoolleyman at gmail.com> wrote:
> > I know what you mean.  I tried to switch to Maruku, failed horribly,
> > wasted a lot of time, and finally stuck with RedCloth.
> 
> I actually seem to have found the solution: rdiscount.
> It renders my pages mostly allright. Now, all I need
> is an ''inline'' extension for collecting FAQ items and
> news on a single page... Or maybe there''s another
> solution? (Feeds wanted as well:)
RDiscount is a Markdown-only parser but does not have the extensions
provided by Maruku. However, it is definitly the fastest option! And it
complies with the Markdown rules set by the original Markdown
implementation. So I would go for it!

Regarding the FAQ items: if I should help you, I need a little bit more
information on how the FAQ items are structured etc. RSS feed
generation is possible with webgen - see the documentation!


Some thoughts on Markdown/Textile:

I find it rather peculiar that there is only one Textile library for
Ruby but that there are several Markdown libraries (BlueCloth, Maruku,
RDiscount, and maybe some more). I think that this is also because of
some people being not satisfied with the default Markdown syntax - I
certainly really like the basic set of Markdown syntax but it lacks
essential things like assigning header IDs. This is possible with
extensions pioneered by PHP Markdown Extra and implement in Maruku.
And this is the reason for Maruku being the default markup converter in
webgen.

However, since Maruku itself isn''t perfect and there are not definite
rules for the translation of Markdown syntax (and the extra syntax) to
HTML, I have started to implement my own Markdown+Extras
parsing/converting library. It will become the default converter for
webgen once it has all the needed features. The main aspects of
kramdown, the name of this new library, are:

* Based on the basic Markdown syntax but with definite rules stating
  what happens in every possible situation. Therefore the output of
  kramdown may be slightly different to that of rdiscount or any other
  ''conforming'' Markdown converter. The upside is that you know
the
  outcome when you write the markup - you don''t have to test various
  possibilities!

* Support for extras like automatic creation of header IDs, fenced code
  blocks, Markdown inside HTML, foot notes etc.

* Fast. In current benchmarks it is around 3 to 4 times faster than
  Maruku on Ruby 1.8. It is still 20x slower than rdiscount but
  provides more features.

* Since it creates an internal representation like Maruku, it is
  possible to create other output formats in the future or manipulate
  the representation before converting to an output format.

Most features are already implemented, I still need to fix some minor
issues with the HTML parsing part. So maybe kramdown will become a
viable alternative to Maruku and/or RedCloth ;-)

-- Thomas

On 2009-03-13, Thomas Leitner <t_leitner at gmx.at>
wrote:
> RDiscount is a Markdown-only parser but does not have the extensions
> provided by Maruku. However, it is definitly the fastest option! And it
> complies with the Markdown rules set by the original Markdown
> implementation. So I would go for it!
Actually, I''m not sure I want the extensions. I haven''t looked
at
them in detail yet, but one of the reasons I like markdown is that
it''s the lightest of the "ASCII markup" language in syntax. I
don''t
like dodging a gazillion cases of special-case syntax everywhere,
as most of these languages have. 
> Regarding the FAQ items: if I should help you, I need a little bit more
> information on how the FAQ items are structured etc. RSS feed
> generation is possible with webgen - see the documentation!
It''s the Ion homepage http://iki.fi/tuomov/ion/faq.html, currently
using Ikiwiki''s ''inline'' plugin to generate both
feeds and pages
containing full file contents and listings.

I have individual FAQ entries in faq/entries/*.mdwn. These
are fully included in faq/some_tag.mdwn based on tags set
in the entry files themselves, as well as listed by heading
in ''faq/entries.mdwn'' that also generates a feed.
(The old version of Ikiwiki I''ve been using doesn''t support
index.html.)

I don''t necessarily need the tagging part. At minimum I
need a way to list all the entries by title; basically
an additional menu. For the news page, it would be very
nice to be able to inline the contents.

Perhaps even tagging could be implemented by support for
multiple menus, and ''in_menu: bugs, features'' -- just one
possibility, probably not the best. 

This is really all part of the planned blog features...


-- 
Stop Gnomes and other pests! Purchase Windows today!

Am Fri, 13 Mar 2009 07:50:19 +0000 (UTC)
schrieb Tuomo Valkonen <tuomov at iki.fi>:
> On 2009-03-13, Thomas Leitner <t_leitner at gmx.at> wrote:
> > RDiscount is a Markdown-only parser but does not have the extensions
> > provided by Maruku. However, it is definitly the fastest option!
> > And it complies with the Markdown rules set by the original Markdown
> > implementation. So I would go for it!
> 
> Actually, I''m not sure I want the extensions. I haven''t
looked at
> them in detail yet, but one of the reasons I like markdown is that
> it''s the lightest of the "ASCII markup" language in
syntax. I don''t
> like dodging a gazillion cases of special-case syntax everywhere,
> as most of these languages have. 
Kramdown will let you choose the used block and span level parsers. For
example, you could select to use foot notes but not use fenced code
blocks.
> > Regarding the FAQ items: if I should help you, I need a little bit
> > more information on how the FAQ items are structured etc. RSS feed
> > generation is possible with webgen - see the documentation!
> 
> It''s the Ion homepage http://iki.fi/tuomov/ion/faq.html, currently
> using Ikiwiki''s ''inline'' plugin to generate both
feeds and pages
> containing full file contents and listings.
> 
> I have individual FAQ entries in faq/entries/*.mdwn. These
> are fully included in faq/some_tag.mdwn based on tags set
> in the entry files themselves, as well as listed by heading
> in ''faq/entries.mdwn'' that also generates a feed.
> (The old version of Ikiwiki I''ve been using doesn''t
support
> index.html.)
This could be done by using a meta information tags in the
faq/entries/*.mdwn files and having a small ERB code fragment in
faq/some_tag.mdwn which selects only those files in faq/entries/*.mdwn
that have a certain tag set. Feed generation should be no problem.
 
> I don''t necessarily need the tagging part. At minimum I
> need a way to list all the entries by title; basically
> an additional menu. For the news page, it would be very
> nice to be able to inline the contents.
As I said above, a small ERB fragment can be used to list the
entries by title, description, ... Inlining content can be done by
using the <webgen:block /> tag with a dynamically generated chain
attribute, eg.

    <webgen:block name="content" chain="<%=
entry.absolute_lcn %>" />

where `entry` refers to the node of the currently processed faq entry.

-- Thomas

On 2009-03-13, Thomas Leitner <t_leitner at gmx.at>
wrote:
> As I said above, a small ERB fragment can be used to list the
> entries by title, description, ... Inlining content can be done by
> using the <webgen:block /> tag with a dynamically generated chain
> attribute, eg.
>
>     <webgen:block name="content" chain="<%=
entry.absolute_lcn %>" />
>
> where `entry` refers to the node of the currently processed faq entry.
Hm... I''d very much prefer an existing extension/tag, or even ready
snippet of code, to starting to learn Ruby/ERB/webgen internals at 
this point in time. (Would be quicker to write a custom shell script
hack than something cleaner.) Already wasted too much time on 
computer/OS switch, and just want to get the site back into a 
condition where it can be updated.

-- 
Tuomo

On 2009-03-13, Tuomo Valkonen <tuomov at iki.fi>
wrote:
>> entries by title, description, ... Inlining content can be done by
>> using the <webgen:block /> tag with a dynamically generated chain
>> attribute, eg.
>>
>>     <webgen:block name="content" chain="<%=
entry.absolute_lcn %>" />
>>
>> where `entry` refers to the node of the currently processed faq entry.
>
> Hm... I''d very much prefer an existing extension/tag, or even
ready
> snippet of code, 
I think that nobody knows how to actually loop over the entries or
whatever is needed to use that. 

-- 
In 1995, Linux was almost a bicycle; an alternative way of live to the
Windows petrol beasts that had to be taken to the dealer for service.
By 2008, Linux has bloated into a gas-guzzler, and local vendors and
artisans have had to yield to "all under one roof" big box
hypermarkets.

Am Thu, 26 Mar 2009 11:33:37 +0000 (UTC)
schrieb Tuomo Valkonen <tuomov at iki.fi>:
> On 2009-03-13, Tuomo Valkonen <tuomov at iki.fi> wrote:
> >> entries by title, description, ... Inlining content can be done by
> >> using the <webgen:block /> tag with a dynamically generated
chain
> >> attribute, eg.
> >>
> >>     <webgen:block name="content" chain="<%=
entry.absolute_lcn
> >> %>" />
> >>
> >> where `entry` refers to the node of the currently processed faq
> >> entry.
> >
> > Hm... I''d very much prefer an existing extension/tag, or even
ready
> > snippet of code, 
> 
> I think that nobody knows how to actually loop over the entries or
> whatever is needed to use that. 
This can be done by using ERB in page or template files (see the
documentation for the ERB content processor). However, you would need
basic knowledge of the Ruby programming language.

-- Thomas

On 2009-03-28, Thomas Leitner <t_leitner at gmx.at>
wrote:
> This can be done by using ERB in page or template files (see the
> documentation for the ERB content processor). However, you would need
> basic knowledge of the Ruby programming language.
Yes, I know that. But more importantly, I would have to learn
the webgen internals, which I''m not really interested in doing.
The documentation isn''t exactly straightforward. And if it''s
not just a simple two-line for-loop or so, it''s really too
much to embed... And I expect it to be a huge blob, because
nobody just gives the answer. 

-- 
Tuomo

Am Sat, 28 Mar 2009 06:30:22 +0000 (UTC)
schrieb Tuomo Valkonen <tuomov at iki.fi>:
> On 2009-03-28, Thomas Leitner <t_leitner at gmx.at> wrote:
> > This can be done by using ERB in page or template files (see the
> > documentation for the ERB content processor). However, you would
> > need basic knowledge of the Ruby programming language.
> 
> Yes, I know that. But more importantly, I would have to learn
> the webgen internals, which I''m not really interested in doing.
> The documentation isn''t exactly straightforward. And if
it''s
> not just a simple two-line for-loop or so, it''s really too
> much to embed... And I expect it to be a huge blob, because
> nobody just gives the answer. 
It''s hard to provide a general solution for the problem. I have
included a generic ERB script below that inserts links to all page files
sorted by their `sort_info` meta information in a sub folder called
`faqs`:

<% context.content_node.tree[''/faqs''].children.select {|c|
c[''kind''] == ''page''}.sort.each do |c| %>
* <%= context.dest_node.link_to(c) %>   
<% end %>

-- Thomas

On 2009-03-28, Thomas Leitner <t_leitner at gmx.at>
wrote:
> It''s hard to provide a general solution for the problem. 
That''s fine, just some template to get started with.
(Much easier to start work fro templates than from scratch
with too extensive documentation to read.)
><% context.content_node.tree[''/faqs''].children.select
{|c| c[''kind''] == ''page''}.sort.each do |c|
%>
> * <%= context.dest_node.link_to(c) %>   
><% end %>
Replacing the link_to -line with

<webgen:block name="content" chain="<%= c.absolute_lcn
%>" />

does not seem to be working. I get the path to the html
in chain, but the tag just gets rendered, instead of expanded.
I suspect rdiscount is mangling it. Indeed, I tried just
running the page through both rdiscount and good old markdown,
and the former mangled the starting < into &lt; while the
latter didn''t. Of course, the ERB stuff not being executed
affects the results.

Would it be possible to have this code as a separate unprocessed
block or something? Or how difficult would it be to turn it
into an extension/tag?

One thing that came into my mind: Will this webgen:block stuff
actually rebase the links?

Another thing I''ve noticed that the feed extension seems
to choose the 10 earliest entries!

-- 
In 1995, Linux was almost a bicycle; an alternative way of live to the
Windows petrol beasts that had to be taken to the dealer for service.
By 2008, Linux has bloated into a gas-guzzler, and local vendors and
artisans have had to yield to "all under one roof" big box
hypermarkets.

> Replacing the link_to -line with
> 
> <webgen:block name="content" chain="<%= c.absolute_lcn
%>" />
> 
> does not seem to be working. I get the path to the html
> in chain, but the tag just gets rendered, instead of expanded.
> I suspect rdiscount is mangling it. Indeed, I tried just
> running the page through both rdiscount and good old markdown,
> and the former mangled the starting < into &lt; while the
> latter didn''t. Of course, the ERB stuff not being executed
> affects the results.
> 
> Would it be possible to have this code as a separate unprocessed
> block or something? Or how difficult would it be to turn it
> into an extension/tag?
You can easily create a separate block for this. Just add a new block
in your page file and specify the pipeline that you wish to use. For
example, to allow only erb and webgen blocks processing you would do it
like this:

--- name:entries pipeline:erb,blocks
<% context.content_node.tree[''/faqs''].children.select {|c|
c[''kind''] == ''page''}.sort.each do |c| %>
<webgen:block name="content" chain="<%= c.absolute_lcn
%>" />
<% end %>

Then you can include this block in the content section of your page
file. However, since you will always need to use the block tag in the
main content block, the rdiscount processor will always mangle it. So
this seems to be a problem with the rdiscount processor and not webgen
per se.

Your second idea of using a special tag for this naturally also works,
the code for it would look like this (put it into ext/init.rb):

----------------------------------------------------------------
config = Webgen::WebsiteAccess.website.config
require ''webgen/tag''
class FAQEntriesTag

  include Webgen::Tag::Base

  def call(tag, body, context)
    result = ''''
    context.content_node.tree[''/faqs''].children.select {|c|
c[''kind''] == ''page''}.sort.each do |c|
      context.dest_node.node_info[:used_nodes] << c.absolute_lcn
      result +=
c.node_info[:page].blocks[''content''].render(context.clone(:chain
=> [c])).content + "\n"
    end
    result
  end

end

config[''contentprocessor.tags.map''][''faqentries'']
= ''FAQEntriesTag''
----------------------------------------------------------------

This would create a tag called `faqentries`. Everytime it is used, it
searches for all page files under the `/faqs` directory and inserts the
rendered content of the found nodes into document. The third line of
the method adds the found node to a special array sothat the page gets
automatically re-generated when the node changes. The fourth line
renders the found node.
> One thing that came into my mind: Will this webgen:block stuff
> actually rebase the links?
Yes, it will correctly convert all webgen generated links into relative
links.

-- Thomas

On 2009-03-29, Thomas Leitner <t_leitner at gmx.at>
wrote:
> Your second idea of using a special tag for this naturally also works,
> the code for it would look like this (put it into ext/init.rb):
>
This seems to be inserting the entries, but they''re incorrectly
sorted, just like the feed. I have the dates set in created.metainfo
this way:

--- name:paths
Released_ion-3rc-20071220.page:   
  created_at: 2007-12-20T18:57:36Z 
  modified_at: 2007-12-20T18:57:36Z
Released_ion-3rc-20080103.page:   
  created_at: 2008-01-03T18:27:27Z 
  modified_at: 2008-01-03T18:27:27Z

etc. Should this work?

Rebasing also doesn''t seem to be happening. Both
[download]({relocatable: /download.html}) and {link: ../download.html} 
end up pointing outside the target tree, i.e., are not altered.
Also the title should be created (but maybe I''ll figure it out).

-- 
Tuomo

Am Sun, 29 Mar 2009 16:44:20 +0000 (UTC)
schrieb Tuomo Valkonen <tuomov at iki.fi>:
> On 2009-03-29, Thomas Leitner <t_leitner at gmx.at> wrote:
> > Your second idea of using a special tag for this naturally also
> > works, the code for it would look like this (put it into
> > ext/init.rb):
> >
> 
> This seems to be inserting the entries, but they''re incorrectly
> sorted, just like the feed. I have the dates set in created.metainfo
> this way:
> 
> --- name:paths
> Released_ion-3rc-20071220.page:   
>   created_at: 2007-12-20T18:57:36Z 
>   modified_at: 2007-12-20T18:57:36Z
> Released_ion-3rc-20080103.page:   
>   created_at: 2008-01-03T18:27:27Z 
>   modified_at: 2008-01-03T18:27:27Z
> 
> etc. Should this work?
Yes, this should work. I tried it on a sample site and the dates are
set correctly.

However, webgen sorts nodes using the meta information sort_info and
uses the title of a node as fallback. So, since the entries don''t
appear to have the sort_info meta information set, they are sorted by
their titles. If you want to sort by modification time, you need to
replace the `.sort` call in the code from last mail with

   .sort {|a,b| a[''modified_at''] <=>
b[''modified_at'']}
> Rebasing also doesn''t seem to be happening. Both
> [download]({relocatable: /download.html}) and
> {link: ../download.html} end up pointing outside the target tree,
> i.e., are not altered. Also the title should be created (but maybe
> I''ll figure it out).
The fragments are correct. For setting the title attribute of a link
generated with the link tag, have a look at its documentation at
http://webgen.rubyforge.org/documentation/tag/link.html . Are any
warnings or errors displayed when you run webgen?

Could you send me the sources for your website? This would greatly help
figuring out why things are not working the way they should.

-- Thomas

On 2009-03-29, Thomas Leitner <t_leitner at gmx.at>
wrote:
>    .sort {|a,b| a[''modified_at''] <=>
b[''modified_at'']}
I had to change the order of and b in the test to make this work.
The same change in feed.rb fixed feeds.

The result is as follows (sorry, no proper attachment; reading
through gmane with slrn, which has no support):

---
config = Webgen::WebsiteAccess.website.config

require ''webgen/tag''

class InlineTag

  include Webgen::Tag::Base                                                     

  def call(tag, body, context)
    pages = param(''tag.inline.pages'')
    count = param(''tag.inline.count'')
    result = ''''
    nodes = context.content_node.tree[pages].children.\
        select {|c| c[''kind''] == ''page''}.\
        sort {|a,b| b[''created_at''] <=>
a[''created_at'']}
    (count == 0 ? nodes : nodes[0,count]).each do |c|
      context.dest_node.node_info[:used_nodes] << c.absolute_lcn
      result += "<div
class=\"inlinepage\">\n<h2>" \
          + context.dest_node.link_to(c) + "</h2>\n" \
          +
c.node_info[:page].blocks[''content''].render(context.clone(:chain
=> [c])).content \
          + "\n<span class=\"pageinfo\">Posted " \
          + c[''created_at''].strftime("%Y-%m-%d
%H:%M") \
          + "</span>\n</div>\n"
    end
    result
  end
  
end

config[''contentprocessor.tags.map''][''inline'']
= ''InlineTag''

config.tag.inline.pages("/news", :doc => ''page
directory'', :mandatory => ''default'')
config.tag.inline.count(0, :doc => ''page count'')
---

However, the parameters aren''t being recognised. If I remove the
"/news" default, I get

    undefined method `children'' for nil:NilClass

and with it, I still get

    WARN -- Invalid parameter ''count'' for tag
''InlineTag'' in </news.en.html>
    WARN -- Invalid parameter ''pages'' for tag
''InlineTag'' in </news.en.html>

> The fragments are correct. For setting the title attribute of a link
> generated with the link tag, have a look at its documentation at
> http://webgen.rubyforge.org/documentation/tag/link.html . Are any
> warnings or errors displayed when you run webgen?
>
> Could you send me the sources for your website? This would greatly help
> figuring out why things are not working the way they should.
There isn''t much to it. Basically:

src/news.page:

{inline: {pages: /news, count: 3}}

src/news/Released_xxxx.page:

([download]({relocatable: /download.html}))

or

{link: ../download.html}

(plus title in metadata section, and created/modified date in
src/news/created.metainfo, extracted from version control system).

Neither of these link types get rebased for inclusion in the 
news.html one directory up.

-- 
Stop Gnomes and other pests! Purchase Windows today!

> However, the parameters aren''t being recognised. If I remove the
> "/news" default, I get
> 
>     undefined method `children'' for nil:NilClass
> 
> and with it, I still get
> 
>     WARN -- Invalid parameter ''count'' for tag
''InlineTag'' in
> </news.en.html> WARN -- Invalid parameter ''pages''
for tag ''InlineTag''
> in </news.en.html>
The resolution of the configuration parameters is done via the class
name of a tag but it seems that I have forgotten to mention this in
the API documentation - will fix it. So basically, if you have a class
called InlineTag, the default mechanism searches for options under
inlinetag, ie. you have to set the options this way:

config.inlinetag.pages("/news", :doc => ''page
directory'', :mandatory => ''default'')
config.inlinetag.count(0, :doc => ''page count'')

Then the configuration options are found. Note that you also have to
change

    param(''tag.inline.pages'')
    param(''tag.inline.count'')

to

    param(''inlinetag.pages'')
    param(''inlinetag.count'')

Or you can name your tag class Tag::Inline - then you don''t have to
change the rest as stated above.
> There isn''t much to it. Basically:
> 
> src/news.page:
> 
> {inline: {pages: /news, count: 3}}
> 
> src/news/Released_xxxx.page:
> 
> ([download]({relocatable: /download.html}))
> 
> or
> 
> {link: ../download.html}
> 
> (plus title in metadata section, and created/modified date in
> src/news/created.metainfo, extracted from version control system).
> 
> Neither of these link types get rebased for inclusion in the 
> news.html one directory up.
That''s my fault, did not code correctly. Replace the corresponding line
with the following

          +
c.node_info[:page].blocks[''content''].render(context.clone(:chain
=> [c], :dest_node => context.dest_node || context.content_node)).content
\

This sets the extra information :dest_node on the clone context. This
information is needed for the rebasing when the rendered node and the
output node are not the same.

I have tested the above changes and they work with 0.5.7.

-- Thomas

On 2009-03-30, Thomas Leitner <t_leitner at gmx.at>
wrote:
> I have tested the above changes and they work with 0.5.7.
I will try these later. One thing that popped to my mind, though:
would it be possible to extend the feed plugin to define a block
of the feed items (in the style of my init.rb) for inclusion in
other pages? That would seem cleaner than defining both inline
and feed separately... although of some inlines I don''t want a
feed of.

It''s something to consider anyway, if the blog features are to be
extended. I should in any case get my site to a working condition
if the mentioned fixes work.

-- 
"[Fashion] is usually a form of ugliness so intolerable that we have
 to alter it every six months." -- Oscar Wilde
"The computer industry is the only industry that is more fashion-driven
 than women''s fashion." -- RMS

Am Mon, 30 Mar 2009 09:52:57 +0000 (UTC)
schrieb Tuomo Valkonen <tuomov at iki.fi>:
> On 2009-03-30, Thomas Leitner <t_leitner at gmx.at> wrote:
> > I have tested the above changes and they work with 0.5.7.
> 
> I will try these later. One thing that popped to my mind, though:
> would it be possible to extend the feed plugin to define a block
> of the feed items (in the style of my init.rb) for inclusion in
> other pages? That would seem cleaner than defining both inline
> and feed separately... although of some inlines I don''t want a
> feed of.
> 
> It''s something to consider anyway, if the blog features are to be
> extended. I should in any case get my site to a working condition
> if the mentioned fixes work.
I''m thinking about providing an easy method for selecting an arbitrary
set of nodes for further processing. It will allow selecting nodes
by different criterias, sorting them and do everything else that is
needed (hopefully). Then this method will be used for all such
things like selecting nodes for feeds, for menus, ...

-- Thomas

On 2009-03-30, Tuomo Valkonen <tuomov at iki.fi>
wrote:
> On 2009-03-30, Thomas Leitner <t_leitner at gmx.at> wrote:
>> I have tested the above changes and they work with 0.5.7.
>
> I will try these later. 
Ok, with these fixes things seem to work.

There''s one minor bug/annoying feature that I noted: if I add
a virtual (index) page, pages referring to it (through from the 
breadcrumb trail) won''t be updated. (I''m not going to start
moving pages around, and since ikiwiki had index for /foo/ as
/foo.html, I''m adding a virtual /foo/index.html referring to
/foo.html. This makes it easier with the selection of pages
for feeds/inlines too.)

-- 
"[Fashion] is usually a form of ugliness so intolerable that we have
 to alter it every six months." -- Oscar Wilde
"The computer industry is the only industry that is more fashion-driven
 than women''s fashion." -- RMS

> There''s one minor bug/annoying feature that I noted: if I add
> a virtual (index) page, pages referring to it (through from the 
> breadcrumb trail) won''t be updated. (I''m not going to
start
> moving pages around, and since ikiwiki had index for /foo/ as
> /foo.html, I''m adding a virtual /foo/index.html referring to
> /foo.html. This makes it easier with the selection of pages
> for feeds/inlines too.)
Okay, I have tried to reproduce this with the following setup:

    /index.page
    /foo.page
    /foo
    /virtual

Contents of index.page:

    A link to {link: foo/index.html}

Contents of virtual:

    foo/index.html:
      title: My Kung Foo

Then the whole site is rendered. When changing meta information
associated with the virtual file (e.g. the title), the link in
`index.page` is not updated. This is certainly not the intended
behaviour - is this what you mean?

I have fixed the above behaviour and unconvered some other lurking bugs
on the way - the fixes are available with the next release!

-- Thomas

On 2009-03-31, Thomas Leitner <t_leitner at gmx.at>
wrote:
> Contents of virtual:
>
>     foo/index.html:
>       title: My Kung Foo
>
> Then the whole site is rendered. When changing meta information
> associated with the virtual file (e.g. the title), the link in
> `index.page` is not updated. This is certainly not the intended
> behaviour - is this what you mean?
I have

faq/index.virtual:

\--- !omap
- index.html:
    url: /faq.html
    routed_title: FAQ

When I created this and other similar files, already rendered pages
pointing to a non-existent index within the directory (faq/), were
not automatically updated to reflect this "redirection" of
faq/(index.html) to faq.html. It was simplest to delete webgen.cache
to have them updated.

-- 
Tuomo

> I have
> 
> faq/index.virtual:
> 
> \--- !omap
> - index.html:
>     url: /faq.html
>     routed_title: FAQ
> 
> When I created this and other similar files, already rendered pages
> pointing to a non-existent index within the directory (faq/), were
> not automatically updated to reflect this "redirection" of
> faq/(index.html) to faq.html. It was simplest to delete webgen.cache
> to have them updated.
Thanks for the test case! I have added this to my TODO list to
investigate it. This behaviour will probably affect not only directory
index files but links/references to all nodes, but these cases appear
only very seldom.

-- Thomas