ikiwiki/doc/todo/edittemplate_should_support_uuid__44___date_variables.mdwn

[[!template id=gitbranch branch=anderbubble/edittemplate author="Jonathon Anderson"]]
[[!tag wishlist patch]]

I use a default template for all new pages:

    \[[!meta title="<TMPL_VAR name>"]]
    \[[!meta author=]]
    \[[!meta date="<TMPL_VAR time>"]]
    \[[!meta guid="urn:uuid:<TMPL_VAR uuid>"]]
    \[[!tag ]]

This encourages me to include useful metadata on the page.  In particular, though, I've modified the `edittemplate` plugin to generate a uuid for use in the guid, for use in `inline`.  Importantly, this keeps `inline` from flooding aggregators when I rename these pages.

I've also noticed that IkiWiki seems to use the creation time for the generated page for the page date.  This means that when I do a rebuild, `inline`d pages get shuffled.  The inclusion of a `time` variable in `edittemplate` (and in a `meta` declaration for all such pages) prevents the date from changing unexpectedly.

I've already made these changes in my installation, and have made my patches available in the `edittemplate` branch of git://civilfritz.net/ikiwiki.git.

Changes to the structure of `$pagestate{$registering_page}{edittemplate}{$pagespec}` mean that a `cgi` rebuild is necessary (for reasons I don't entirely understand); but I think that's preferable to creating an entirely separate `$pagestate` namespace for storing parameters.  That said, I'm not really a perl programmer, so corrections are welcome.

> I like this patch. I hate seeing things I've already read get marked as unread in my rss feed. -- [[JoshBBall]]

>> (I don't have commit access so take this with a pinch of salt -
>> I'm just trying to help deal with the code-review backlog.)
>>
>> I mostly like the first and third patches in the branch (adding v4
>> (random) UUIDs, and adding the timestamps). I'd be tempted to rename
>> `time` and `formatted_time` to `iso_time` and `time`, but that's
>> a matter of taste, and perhaps people with commit access have
>> stronger opinions one way or another. I haven't researched
>> whether there's precendent for any particular naming style
>> elsewhere in ikiwiki.
>>
>> The UUID bit would require adding some reference to libuuid-tiny-perl
>> to the Debian packaging - I think a `Recommends` is too strong
>> but a `Suggests` seems OK.
>>
>> I don't like the second patch (adding URL-namespaced UUID support).
>> I'm having a hard time thinking of any situation in which a v4 UUID
>> would be unsuitable, which means it's unnecessary complexity.
>> FYI, the reason that it makes a rebuild is necessary is that
>> you've restructured `$pagestate`, which is carried over from one
>> refresh to the next (that's its purpose), and you haven't
>> built in any migration or backwards-compatibility code that will
>> cope with it being in the old format. My inclination would be to
>> drop that patch. If there's a really good reason to prefer
>> v3/v5 UUIDs, please describe it and I'll try to suggest some
>> better way based on that, maybe global configuration in `$config`.
>> --[[smcv]]

>>> [[!template id=gitbranch branch=smcv/ready/edittemplate
  browse=http://git.pseudorandom.co.uk/smcv/ikiwiki.git/shortlog/refs/heads/ready/edittemplate
  author="Jonathon Anderson, [[smcv]]"]]
>>> Here is a version of that branch that I [[would merge|users/smcv/ready]] if I could.
>>> Changes since Jonathon's version:
>>>
>>> * only generate a UUID if needed
>>> * read `/proc/sys/kernel/random/uuid` instead of using [[!cpan UUID::Tiny]]
>>>   if available, to avoid the dependency on at least Linux
>>> * remove v3/v5 UUID support since I don't really see the point,
>>>   and if included, it would need a migration path for `$pagestate`
>>> * use RFC 3339 time format for `time` to make the timezone unambiguous
>>> * add `html_time` which is the output of `IkiWiki::displaytime`, e.g.
>>>   a `<time>` element on HTML5 wikis
>>>
>>> I'm not entirely sure what the use-case is for `formatted_time`,
>>> so perhaps either `html_time` or `formatted_time` should be
>>> removed; but it's not as if they really cost anything.
>>>
>>> There doesn't seem to be any strong convention for how we name
>>> timestamp variables/objects, so I left the naming as it was.
>>>
>>> --[[smcv]]