[pmwiki-users] pagelist template brainstorming

[Hladůvka Jiří]

The more complex pagelists are a kind of reports.
In such we do:
- filter records (pages) ... it is said in (:pagelist group=... name= ...:)

- ordering records (pages) ... it is said in (:pagelist ...order=...:) 
   I think this should be in the template otherwise there can happen an 
inconsistence between
   ordering and grouping the records

- grouping records (pages) ... in the template  e.g. (:if equal 
{<$Group} {=$Group}:)

- display and format desired content ... in the template 
   e.g. ||{=$FullName}  ||{=$LastModifiedBy} ||{=$LastModified} ||

The grouping should alow more levels and each group should have enabled
some (:first <level>:) and the (:last <level>:)

I can imagin when authors start to create shopping carts (and have 
proper Math tools )
then they are happy to create pagelists grouped by delivery_date and 
delivery_town


 >So, again my questions are:  Would markups such as these make it
 >easier or more difficult for people creating pagelist templates? 
For people skilled in data processing there is no problem, IMO. The 
others can try and discover the advantage.

 >And is the (:first:)/(:each:)/(:last:) syntax appropriate, or
 >would a different syntax be more desirable?
When pagetextvariables are now to our disposition we can filter and sort 
on them.
Then the template should enable such a structure:
#template
(:first:)  ... start of the pagelist
(:first {=$:1stgroupingkey}:)  ... start of the highest level group
(:first {=$:1stgroupingkey}{=$:2ndgroupingkey}:)  ... ... start of the 
next level group
...
(:each :)  ...  record level (single page)
...
(:last {=$:2ndgroupingkey}{=$:2ndgroupingkey}:) 
(:last {=$:1stgroupingkey}:) 
(:last:)
#templateend

While in  the case of (:first ...:) and (:each :)  the current page 
attributes and textvars are accesible
in the case of (:last...:) the key values should be accessible in some 
temporary variable e.g.$_old2ndgroupingkey

I have no idea about the impact on the performance - number of 
iterations in PHP
but from the author point of view it is very logical and transparent.


Regards,
Jiri


Patrick R. Michaud napsal(a):
> Today I had an idea about improving pagelist templates, and
> would like some comments and feedback.  As many of you know,
> pagelist templates can be a bit slow, and I think I have
> an idea that could speed them up as well as make them easier
> to create.  (If you don't do much with pagelist templates,
> feel free to skip this message.  :-)
>
> For purposes of this message, consider the following template
> to create a table of page names, authors, and modification dates,
> with header rows at the top and colspan rows to mark the beginning 
> of each group:
>
>     [[#template]]
>     (:if equal {<$Group}:)
>     ||width=100%
>     ||!Page         ||!Author            ||!Modified        ||
>     (:if equal {<$Group} {=$Group}:)
>     ||{=$Group}     ||||||
>     (:if:)
>     ||{=$FullName}  ||{=$LastModifiedBy} ||{=$LastModified} ||
>     (:if equal {>$Group}:)
>     {$$PageCount} pages.
>     [[#templateend]]
>
> Of course, the (:if:) lines are the canonical (and quite common)
> constructs that people use to identify sections for
>
>     the first page in the pagelist   (:if equal {<$Group} :)
>     the first page of a wikigroup    (:if equal {<$Group} {=$Group}:)
>     each individual page             (:if:)
>     the last page of the pagelist    (:if equal {>$Group} :)
>
> It works fine, but isn't all that desirable from a performance
> perspective, because all of the sections are included and processed
> for every page in the list.  In other words, if a pagelist has 
> 1000 pages over 10 wikigroups, then the above template produces 
> 9000 lines of markup to be processed, even though most of the lines
> will eventually be discarded by the (:if:) statements.
>
> Thus I'm wondering if we can improve things for both performance 
> and readability by introducing special-purpose (:first:), (:each:),
> and (:last:) directives inside of pagelist templates.
>
> The (:first:) directive would identify markup prior to the first 
> page in the pagelist, (:each:) would be markup used for each page 
> in the list, and (:last:) would contain markup used after the 
> last page in the list.  Thus with these markers, our above 
> template would read:
>
>     [[#template]]
>     (:first:)
>     ||width=100%
>     ||!Page         ||!Author            ||!Modified        ||
>     (:each:)
>     (:if equal {<$Group} {=$Group}:)
>     ||{=$Group}     ||||||
>     (:if:)
>     ||{=$FullName}  ||{=$LastModifiedBy} ||{=$LastModified} ||
>     (:last:)
>     {$$PageCount} pages.
>     [[#templateend]]
>
> Now the pagelist template formatter can be much smarter about
> producing the markup.  Unlike the previous version where 1000
> pages in the list produced 9000 lines of markup, this version 
> would end up with only 2 + 1000*4 + 1 == 4003 lines of markup
> to process -- a significant decrease.
>
> Also, since it's very common for templates to produce "control breaks"
> whenever a value changes (e.g., "the first page in a wikigroup"),
> we could improve things even further by allowing (:each:) to take
> a parameter that says "output this section whenever the specified
> parameter changes value".  Thus:
>
>     [[#template]]
>     (:first:)
>     ||width=100%
>     ||!Page         ||!Author            ||!Modified        ||
>     (:each {=$Group}:)
>     ||{=$Group}     ||||||
>     (:each:)
>     ||{=$FullName}  ||{=$LastModifiedBy} ||{=$LastModified} ||
>     (:last:)
>     {$$PageCount} pages.
>     [[#templateend]]
>
> Here, (:each {=$Group}:) says to output the markup that follows
> for every page where {=$Group} results in a value different from
> the previous iteration.  With this formulation our 1000 page, 10 group 
> list now produces 2 + 10*1 + 1000*1 + 1 == 1013 lines of markup, far 
> less than 9000 lines of the original template.
>
> I'm a bit concerned that some authors may try to put full fledged
> conditionals into (:each:), as opposed to simply values to be
> "watched".  So, perhaps (:each:) should take a conditional expression
> instead of a value to be watched...
>
>     (:each ! equal {=$Group} {<$Group} :)
>
> But somehow I think it reads better if (:each:) simply watches
> a value for control breaks, and that we tell authors to always
> use (:if:) when a condtional needs to be evaluated.
>
> Does this approach and markup make sense to others?
>
> Note that all existing pagelist templates will continue to work
> exactly as before.  Any template that doesn't contain (:first:), 
> (:each:), or (:last:) directives will still act as though the 
> entire template is included for each page in the list.  And,
> pagelist templates that need to do more complex sorts of
> operations can still use (:if:) in the templates as before.
>
> While we're on the topic, another desirable feature we can introduce
> with this approach would be the ability to set default pagelist
> options from within a pagelist template itself.  For example,
> to display a list of page titles in alphabetical order, we
> currently have to remember to specify both "fmt=#title" and 
> "order=title" in the pagelist directive.  It would be nice if 
> "fmt=#title" automatically implied "order=title" as well.  
> So, we could add an (:options:) directive to the template, as in
>
>     [[#title]]
>     (:options order=title:)
>     * [[{=$FullName}|+]]
>     [[#titleend]]
>
> Then an author specifying "fmt=#title" in a pagelist command
> would automatically get the pages ordered by title unless 
> overridden by an explicit "order=" parameter.
>
> So, again my questions are:  Would markups such as these make it
> easier or more difficult for people creating pagelist templates?  
> And is the (:first:)/(:each:)/(:last:) syntax appropriate, or 
> would a different syntax be more desirable?
>
> Thanks in advance for any comments and suggestions,
>
> Pm
>
> _______________________________________________
> pmwiki-users mailing list
> pmwiki-users at pmichaud.com
> http://www.pmichaud.com/mailman/listinfo/pmwiki-users
>
> __________ Informacia od NOD32 2067 (20070217) __________
>
> Tato sprava bola preverena antivirusovym systemom NOD32.
> http://www.eset.sk
>
>
>
>