$pagestate{$page}{meta}{authorurl}=$value if safeurl($value);
# fallthrough
}
+ elsif ($key eq 'date') {
+ eval q{use Date::Parse};
+ if (! $@) {
+ my $time = str2time($value);
+ $IkiWiki::pagectime{$page}=$time if defined $time;
+ }
+ }
if (! defined wantarray) {
# avoid collecting duplicate data during scan pass
}
# Metadata collection that happens only during preprocessing pass.
- if ($key eq 'date') {
- eval q{use Date::Parse};
- if (! $@) {
- my $time = str2time($value);
- $IkiWiki::pagectime{$page}=$time if defined $time;
- }
- }
- elsif ($key eq 'permalink') {
+ if ($key eq 'permalink') {
if (safeurl($value)) {
$pagestate{$page}{meta}{permalink}=$value;
push @{$metaheaders{$page}}, scrub('<link rel="bookmark" href="'.encode_entities($value).'" />', $destpage);
* aggregate: If a feed fails to be downloaded, try again immediatly
next time aggregation is run, even if the usual time has not passed.
Closes: #508622 (Michael Gold)
+ * meta: Process meta date during scan pass so that the date will always
+ affect sorting in inlines.
-- Joey Hess <joeyh@debian.org> Mon, 17 Nov 2008 14:02:10 -0500
test case I can run, as instructions like "Create pages that sort different by
mtime and ctime" are not ones that I know how to follow (given that touch sets
*both*). --[[Joey]]
+
+> Sorry. I conflated Unix ctime and ikiwiki's creation time because I
+> didn't think the difference was important to this case. I'm a writer,
+> and I should have known better -- I appologize. Revised steps to
+> reproduce are below; feel free to delete this whole misunderstanding
+> to make the bug report more concise.
+>
+> 1. Create pages in the srcdir that should sort in one order by
+> last-modification time and in a diffent order by original creation
+> time. For example:
+>
+> $ echo -e '\[[!meta date="2007-01-01"]]\nNew.' > test/new.mdwn
+> $ echo -e '\[[!meta date="2006-01-01"]]\nOld.' > test/old.mdwn
+>
+> 2. Create a page that includes the files. For example:
+>
+>
+> $ echo '\[[!inline pages="test/*"]]' > sort-test.mdwn
+>
+> 3. Run ikiwiki. For example `ikiwiki --setup setup_file`
+>
+> 4. Pages are output incorrectly in modification time order. For example,
+> actual partial HTML of the sort-test/index.html from commands used
+> above (whitespace-only lines removed; one whitespace-only line
+> added):
+>
+> <div class="inlinepage">
+> <span class="header">
+> <a href="./../test/old/">old</a>
+> </span>
+> <p>Old.</p>
+> <span class="pagedate">
+> Posted Sun 01 Jan 2006 12:00:00 AM EST
+> </span>
+> </div>
+>
+> <div class="inlinepage">
+> <span class="header">
+> <a href="./../test/new/">new</a>
+> </span>
+> <p>New.</p>
+> <span class="pagedate">
+> Posted Mon 01 Jan 2007 12:00:00 AM EST
+> </span>
+> </div>
+>
+> 5. Run ikiwiki again with the same command line. For example: `ikiwiki --setup setup_file`
+>
+> 6. Pages are output correctly in creation time order. For example,
+> actual partial HTML of the sort-test/index.html from commands used
+> above (whitespace-only lines removed; one whitespace-only line
+> added):
+>
+> <div class="inlinepage">
+> <span class="header">
+> <a href="./../test/new/">new</a>
+> </span>
+> <p>New.</p>
+> <span class="pagedate">
+> Posted Mon 01 Jan 2007 12:00:00 AM EST
+> </span>
+> </div>
+>
+> <div class="inlinepage">
+> <span class="header">
+> <a href="./../test/old/">old</a>
+> </span>
+> <p>Old.</p>
+> <span class="pagedate">
+> Posted Sun 01 Jan 2006 12:00:00 AM EST
+> </span>
+> </div>
+>
+> 7. Create a new page with the current Unix mtime and Unix ctime, but a
+> !meta date before the most recent creation date of another page.
+> For example:
+>
+> $ echo -e '\[[!meta date="2005-01-01"]]\nOlder.' > test/older.mdwn
+>
+> 8. Run ikiwiki again with the same command line. For example: `ikiwiki --setup setup_file`
+>
+> 9. All previously sorted pages output correctly in order of their
+> creation time, but the new page is output incorrectly at the top as
+> if its modification time was its creation time. For example,
+> actual partial HTML of the sort-test/index.html from commands used
+> above (whitespace-only lines removed; two whitespace-only
+> lines added):
+>
+> <div class="inlinepage">
+> <span class="header">
+> <a href="./../test/older/">older</a>
+> </span>
+> <p>Older.</p>
+> <span class="pagedate">
+> Posted Sat 01 Jan 2005 12:00:00 AM EST
+> </span>
+> </div>
+>
+> <div class="inlinepage">
+> <span class="header">
+> <a href="./../test/new/">new</a>
+> </span>
+> <p>New.</p>
+> <span class="pagedate">
+> Posted Mon 01 Jan 2007 12:00:00 AM EST
+> </span>
+> </div>
+>
+> <div class="inlinepage">
+> <span class="header">
+> <a href="./../test/old/">old</a>
+> </span>
+> <p>Old.</p>
+> <span class="pagedate">
+> Posted Sun 01 Jan 2006 12:00:00 AM EST
+> </span>
+> </div>
+>
+> 10. Run ikiwiki again with the same command line. For example: `ikiwiki --setup setup_file`
+>
+> 11. All pages, including new page, are output correctly in creation time
+> order. For example, actual partial HTML of the sort-test/index.html
+> from commands used above (whitespace-only lines removed; two
+> whitespace-only lines added):
+>
+> <div class="inlinepage">
+> <span class="header">
+> <a href="./../test/new/">new</a>
+> </span>
+> <p>New.</p>
+> <span class="pagedate">
+> Posted Mon 01 Jan 2007 12:00:00 AM EST
+> </span>
+> </div>
+>
+> <div class="inlinepage">
+> <span class="header">
+> <a href="./../test/old/">old</a>
+> </span>
+> <p>Old.</p>
+> <span class="pagedate">
+> Posted Sun 01 Jan 2006 12:00:00 AM EST
+> </span>
+> </div>
+>
+> <div class="inlinepage">
+> <span class="header">
+> <a href="./../test/older/">older</a>
+> </span>
+> <p>Older.</p>
+> <span class="pagedate">
+> Posted Sat 01 Jan 2005 12:00:00 AM EST
+> </span>
+> </div>
+>
+> File status after all the above actions:
+>
+> $ stat test/*
+> File: `test/new.mdwn'
+> Size: 33 Blocks: 8 IO Block: 4096 regular file
+> Device: ca20h/51744d Inode: 684160 Links: 1
+> Access: (0644/-rw-r--r--) Uid: ( 1000/ harding) Gid: ( 1000/ harding)
+> Access: 2008-12-20 21:48:32.000000000 -0500
+> Modify: 2008-12-20 21:27:03.000000000 -0500
+> Change: 2008-12-20 21:27:03.000000000 -0500
+> File: `test/older.mdwn'
+> Size: 35 Blocks: 8 IO Block: 4096 regular file
+> Device: ca20h/51744d Inode: 684407 Links: 1
+> Access: (0644/-rw-r--r--) Uid: ( 1000/ harding) Gid: ( 1000/ harding)
+> Access: 2008-12-20 21:48:32.000000000 -0500
+> Modify: 2008-12-20 21:42:10.000000000 -0500
+> Change: 2008-12-20 21:42:10.000000000 -0500
+> File: `test/old.mdwn'
+> Size: 33 Blocks: 8 IO Block: 4096 regular file
+> Device: ca20h/51744d Inode: 684161 Links: 1
+> Access: (0644/-rw-r--r--) Uid: ( 1000/ harding) Gid: ( 1000/ harding)
+> Access: 2008-12-20 21:48:32.000000000 -0500
+> Modify: 2008-12-20 21:27:09.000000000 -0500
+> Change: 2008-12-20 21:27:09.000000000 -0500
+>
+> My ikiwiki configuration file (being used to port a blog from pyblosxom
+> to ikiwiki):
+>
+> harding@mail:~$ sed 's/#.*//; /^[ ]*$/d' .ikiwiki/gnuisance.setup
+> use IkiWiki::Setup::Standard {
+> wikiname => "HardingBlog",
+> adminemail => 'dave@dtrt.org',
+> srcdir => "/srv/backup/git/gnuisance.net",
+> destdir => "/srv/test.dtrt.org",
+> url => "http://srv.dtrt.org",
+> wrappers => [
+> ],
+> atom => 1,
+> syslog => 0,
+> prefix_directives => 1,
+> add_plugins => [qw{goodstuff tag}],
+> disable_plugins => [qw{passwordauth}],
+> tagbase => "tag",
+> }
+>
+> --[David A. Harding](http://dtrt.org/), 2008-12-20
+
+Thank you for a textbook excellent reproduction recipe.
+
+What appears to be going on here is that meta directives are not processed
+until the leaf pages are rendered, and thus the ctime setting is not
+available at the time that they are inlined, and the newer unix ctime is
+used. On the second build, the meta data has already been recorded.
+
+This can probably be avoided by processing meta date at scan time.
+
+Verified, fix works. [[done]]
+--[[Joey]]
-Known issues with the [[plugins/comments]] plugin:
+# Known issues with the [[plugins/comments]] plugin
-* There is some common code cargo-culted from other plugins (notably inline and editpage) which
- should probably be shared
-
- > Actually, there's less of this now than there used to be - a lot of simple
- > things that were shared have become unshareable as they became more
- > complex. --[[smcv]]
+## Unimplemented
* Previews always say "unknown IP address"
-* Add `COMMENTOPENID`: the authenticated/verified user name, if and only if it was an OpenID
+* Instead of just a link to add a comment, it could have a form to enter
+ the title, similar to the form for adding a new blog post.
- > Done in my comments git branch --[[smcv]]
+ > I'm not sure this is so useful? On Livejournal titles are allowed on
+ > comments, but very rarely used (and indeed usually not very useful);
+ > it's hard enough to get some people to title their blog posts :-)
+ > --[[smcv]]
- > Not seeing it there, which branch? --[[Joey]]
+* If a spammer posts a comment, it is either impossible or hard to clean
+ up via the web. Would be nice to have some kind of link on the comment
+ that allows trusted users to remove it (using the remove plugin of
+ course).
- >> Bah, git push --all is not the default... 'comments' branch now (I've also rebased it).
- >> Sorry, I'm on mobile Internet at the moment... --[[smcv]]
+ > Won't the remove plugin refuse to remove internal pages? This would be
+ > a good feature to have, though. --[[smcv]]
+
+* Now that inline has some comments-specific functionality anyway, it would
+ be good to output `<link rel="comments">` in Atom and the equivalent in RSS.
+
+## Patches pending merge
* The default template should have a (?) icon next to unauthenticated users (with the IP address
as title) and an OpenID icon next to OpenIDs
>> directory (untested!) --[[smcv]]
>>> The new code produces links like /wikiisons/openid.png, which
- >>> fail if ikiwiki is not at the root of the web server. --[[Joey]]
+ >>> fail if ikiwiki is not at the root of the web server. --[[Joey]]
+
+ >>>> Sorry, I should have spotted that (the assumption failed on my demo
+ >>>> site, but the push to that site was when I was on the way out, so I
+ >>>> didn't have time to investigate). As a note for other ikiwiki hackers,
+ >>>> I should have used
+ >>>> `<img src="<TMPL_VAR NAME=BASEURL>wikiicons/openid.png" />`. --[[smcv]]
>>> I got to wondering if the icons are needed. On my comments branch
>>> (not master), I've dropped the icons and info can be seen by hovering
>>> over the author's name. Idea being that you probably don't care how
>>> they authenticated unless something is weird, and in that case you
>>> can hover to check. Does that make sense, should I merge it?
- >>> --[[Joey]]
-
-* Should the comments be visually set off more from the page above?
- Rather than just a horizontal rule, I'm thinking put the comments
- in a box like is used for inlined pages.
-
- > I did put them in a box in the CSS... I agree the default template
- > could do with visual improvement though. --[[smcv]]
+ >>> --[[Joey]]
-* Instead of just a link to add a comment, it could have a form to enter
- the title, similar to the form for adding a new blog post.
-
- > I'm not sure this is so useful? On Livejournal titles are allowed on
- > comments, but very rarely used (and indeed usually not very useful);
- > it's hard enough to get some people to title their blog posts :-)
- > --[[smcv]]
-
-* If a spammer posts a comment, it is either impossible or hard to clean
- up via the web. Would be nice to have some kind of link on the comment
- that allows trusted users to remove it (using the remove plugin of
- course).
+ >>>> Yeah, go ahead. I preferred my layout with the author before the
+ >>>> comment - perhaps that's Livejournal's influence :-) - but I can always
+ >>>> edit the templates for my own site. As long as the default is something
+ >>>> reasonable and both layouts are possible, I don't really mind.
+ >>>> Minimizing the number of "resource" files in the basewiki also seems
+ >>>> a good goal. --[[smcv]]
- > Won't the remove plugin refuse to remove internal pages? This would be
- > a good feature to have, though. --[[smcv]]
+## Won't fix
-* One can use inline to set up a feed of all comments posted to any page.
- Using template=comment they are displayed right. Only problem
- is there is no indication in that template of what page each comment in the
- feed is a comment on. So, if a comment is inlined into a different page,
- I think it should show a link back to the page commented on.
- (BTW, the rss feed in this situation seems ok; there the link element
- points back to the parent page.
+* There is some common code cargo-culted from other plugins (notably inline and editpage) which
+ should probably be shared
- > done --[[Joey]]
+ > Actually, there's less of this now than there used to be - a lot of simple
+ > things that were shared have become unshareable as they became more
+ > complex. --[[smcv]]
-* It would be useful to have a pagespec that always matches all comments on
+* It would be useful to have a pagespec that always matches all comments on
pages matching a glob. Something like `comment(blog/*)`.
Perhaps postcomment could also be folded into this? Then the pagespec
would match both existing comments, as well as new comments that are
> would also let X edit/delete comments on blog pages (including those
> written by others) in arbitrary ways, which doesn't seem good. --[[smcv]]
- > I had a look at implementing comment() and fell afoul of
+ > I had a look at implementing comment() and fell afoul of
> some optimisations that assume only internal() will be used to match
- > internal pages. So probably this isn't worth doing. --[[Joey]]
+ > internal pages. So probably this isn't worth doing. --[[Joey]]
+
+## Done
+
+* Add `COMMENTOPENID`: the authenticated/verified user name, if and only if it was an OpenID
+
+ > Done in my comments git branch --[[smcv]]
+
+ > Not seeing it there, which branch? --[[Joey]]
+
+ >> Bah, git push --all is not the default... 'comments' branch now (I've also rebased it).
+ >> Sorry, I'm on mobile Internet at the moment... --[[smcv]]
+
+ >>> merged by [[Joey]] in commit 0f03af38 --[[smcv]]
+
+* Should the comments be visually set off more from the page above?
+ Rather than just a horizontal rule, I'm thinking put the comments
+ in a box like is used for inlined pages.
+
+ > I did put them in a box in the CSS... I agree the default template
+ > could do with visual improvement though. --[[smcv]]
+
+ >> I'll consider this solved by [[Joey]]'s changes. --[[smcv]]
+
+* One can use inline to set up a feed of all comments posted to any page.
+ Using template=comment they are displayed right. Only problem
+ is there is no indication in that template of what page each comment in the
+ feed is a comment on. So, if a comment is inlined into a different page,
+ I think it should show a link back to the page commented on.
+ (BTW, the rss feed in this situation seems ok; there the link element
+ points back to the parent page.
+
+ > done --[[Joey]]
* One of Joey's commit messages says "Not ideal, it would be nicer to jump to
the actual comment posted, but no anchor is available". In fact there is
- an anchor - the `\[[_comment]]` preprocessing wraps the comment in a <div>
+ an anchor - the `\[[_comment]]` preprocessing wraps the comment in a `<div>`
with id="comment_123" or something. I'll fix this, unless Joey gets there
first. --[[smcv]]
- > done --[[Joey]]
-
-* Now that inline has some comments-specific functionality anyway, it would
- be good to output '<link rel="comments">' in Atom and the equivalent in RSS.
+ > done --[[Joey]]
sipb-www / ikiwiki.git / commitdiff
