Merge get_posts() and get_pages()

[mikeschinkel]

get_pages() should wrap get_posts() the same way get_page() wraps get_post(). Arguments of different names need to be retained for back compat.

Reasoning for this includes #14823. Querying a nonhierarchical post type should still be allowed with child_of for example, to allow for cross-type relationships.

[nacin]

get_pages() should be hierarchical. get_posts() should be non-hierarchical.

[scribu]

Me thinks get_posts() should handle both. get_pages() should be a wrapper for get_posts().

[scribu]

...similar to how get_categories() is a wrapper for get_terms().

[filosofo]

+1 to scribu's suggestion

[mikeschinkel]

+1

[mikeschinkel]

Replying to scribu:
I'd like to try to create this patch. I think the scope it defined small enough and straightforward enough that I could tackle it. This would give me the first patch beyond trivial for me to contribute.

If you think it's okay for me to take it on please assign to me.

[scribu]

Go for it. You can set it's status to "accepted".

Just remember that it probably won't make it into WP 3.0.

[nacin]

Go for it. In general, just find a ticket and write a patch if you'd like. I never found a need to mess with the owner components for the most part, though I find it helpful as a committer now.

We need to be fully back compat. The args order and sort_order (and orderby and sort_column) must both stay, for example. We'd want to make orderby and order the 'proper' ones since that is our naming convention. The structural differences between the two -- get_pages() relies on its own query, while get_posts() uses WP_Query, for example -- should be interesting to merge.

Just remember that it probably won't make it into WP 3.0.

Yeah, this isn't going to happen for 3.0. Way too much to change and test late in the dev cycle. Consider it blessed for 3.1 as of ryan's comment in #4711.

[mikeschinkel]

@scribu / @nacin

Thanks. Agree with full backward compatibility.

So for clarity we want to start using the parameters that get_posts() uses in addition to the ones get_pages() has always used so that get_pages() can behave like get_posts() but also so it can still behave the same as it used to.

As far as the two queries, do you know if there's any current incompatibilities between the two?

Understood about 3.1 vs. 3.0 and agree. 3.0 is ready to close out, not add new things to it.

Last question: Shouldn't I create a new ticket for the merge and close this one out?

[scribu]

So for clarity we want to start using the parameters that get_posts() uses in addition to the ones get_pages() has always used so that get_pages() can behave like get_posts() but also so it can still behave the same as it used to.

The other way around: add the get_pages() specific parameters ('child_of', 'exclude_tree' etc.) to get_posts(). Then, retrofit get_pages() to use get_posts().

Last question: Shouldn't I create a new ticket for the merge and close this one out?

Sure, just mention this one in the description.

[scribu]

Related: #12873

[mikeschinkel]

Replying to nacin:

Would you like me to attempt to tackle this one? I think it's small enough scope, sufficiently along my expertise and devoid of debatable design decisions it might make a good project for me?

-Mike

[nacin]

Replying to mikeschinkel:

Would you like me to attempt to tackle this one? I think it's small enough scope, sufficiently along my expertise and devoid of debatable design decisions it might make a good project for me?

Go for it.

[jane]

@mikeschinkel: any status on this? Freeze is a few days away, so posting a patch, even if it's not done yet, to get community feedback would be good, so you'd have time to revise based on feedback.

[jane]

Punting due to no patch and entering beta. @mikeschinkel: In general, don't bother asking if you should do a patch, just do the patch. People will put more attention on it if there's a patch to respond to.

[garyc40]

first attempt at merging get_posts() and get_pages()

[markoheijnen]

I looked at the patch and I noticed that the filter for get_pages is removed.
Removing this filter isn't handy since there are plugins that are using that filter.

If it is possible (have no clue) I would recommend to insert it with a deprecated notice.

[garyc40]

restored "get_pages" filter hook, fixed bug introduced in rev1 where arguments are parsed incorrectly in wp_dropdown_pages()

[grandslambert]

I need to know if this issue will fix the following error when trying to use wp_dropdown_pages on a non-hierarchical custom post type:

Notice: Only variable references should be returned by 
reference in /var/www/wpblabber.lan/wp-includes/post.php on line 3313

That seems to be something that needs to be looked at as well. This only shows when debug mode is on, but when using wp_dropdown_pages on non-hierarchical post types, no dropdown appears.

Before anyone complains about this not being related, please see ticket ##14823

[scribu]

In garyc40-12821-rev2.patch​, in get_posts(), is it really necessary to add the default args from get_pages()? Won't those be set in WP_Query anyway?

[scribu]

What I mean is that get_posts() should remain unchanged.

Any get_pages() specific args, like 'include' should be converted to the standard ones ('post__in') inside get_pages(), before calling get_posts() or WP_Query.

[garyc40]

Replying to scribu:

What I mean is that get_posts() should remain unchanged.

Any get_pages() specific args, like 'include' should be converted to the standard ones ('post__in') inside get_pages(), before calling get_posts() or WP_Query.

'include' is not a get_pages() specific argument. The inline documentation of get_posts() specifies that 'include' is an argument, as does ​our codex.

In my patch, get_pages() call get_posts(), so converting args like include inside get_pages() before passing into get_posts() would result into duplicate code.

Also, I think get_pages() should become a wrapper of get_posts(), in the sense that get_posts() should process arguments that are associated with hierarchical post types. In other words, whatever arguments that get_pages() can handle, get_posts() itself must also be able to handle.

There are indeed some non-standard arguments in get_pages() such as post_parent, sort_order, number etc. but as you can see in the patch, they all get mapped into standard arguments (parent, order, numberposts etc.) for get_posts().

Other arguments like child_of, depth should be supported by get_posts() in case the developer wants to use it with a hierarchical post type.

[scribu]

Ok, but 'child_of' and 'depth' will actually be handled in WP_Query, no? So it's redundant to set them as defaults in get_posts().

[garyc40]

Replying to scribu:

Ok, but 'child_of' and 'depth' will actually be handled in WP_Query, no? So it's redundant to set them as defaults in get_posts().

Actually depth is an argument for wp_dropdown_pages(), so I'm removing it from get_posts() in an upcoming patch.

But there are other arguments such as offset, meta_key, meta_value etc. which are being set as defaults in get_posts() but are also checked and handled in WP_Query. Putting child_of as a default there does no harm and simply makes it more obvious which arguments and default values a WP developer can use with get_posts(). That being said, I don't really mind if it's considered redundant and gets removed.

I'm more concerned with whether there's any wrong assumption I made when merging the two functions. I tested the patch and it seems like it returns the correct results. But there might be edge cases of get_pages() that I don't know about, let alone performance issues.

[garyc40]

depth is not an argument of get_pages() or get_posts()

[kevinB]

Please remember to include 'suppress_filters' => false in the get_posts() args.

[wlindley]

On line 343 you set the 'order_by' attribute (with underline) but note that the routine actually uses 'orderby' (no underline) (see line 1334).

Also, the 'sort_order' attributes to get_pages() differ from the 'orderby' attributes of get_posts() -- the original get_pages() needs, for example, ('sort_order' => 'post_date') while the equivalent in get_posts() would be ('orderby' => 'date')... but only for certain columns (date, modified, author, name, title; but not menu_order, ID, etc.) The exact logic is in this code, file wp-includes/query.php in get_posts():

         switch ( $orderby ) {
                 case 'menu_order':
                         break;
                 case 'ID':
                         $orderby = "$wpdb->posts.ID";
                         break;
[...other special cases...]
                 default:
                         $orderby = "$wpdb->posts.post_" . $orderby;
         }

That case-specific test for "ID" should probably change, too, surely 'id' should work as well?

[dbernar1]

Hello,

I've removed duplication between WP_Query::get_posts() and get_pages(), while preserving the unique interface get_pages() provides. I've not pulled in any of the unique functionality of get_pages() into WP_Query::get_posts() at this time. Please help me understand which functionality I should pull into WP_Query, and what interface I should provide.

I will then implement that, and perform testing on the code to ensure it works as expected.

I am providing the following patch as a reference for anyone who would like to read the code and provide the feedback I need, do not rely on it to work at this time - it has not been tested.

[scribu]

Eliminating the duplication, but keeping the hierarchical parts inside get_pages() seems like the right way to go.

TODO extract this to something like is_a_hierarchical_post_type( $post_type )

We already have is_post_type_hierarchical()

[dbernar1]

Replying to scribu:

Eliminating the duplication, but keeping the hierarchical parts inside get_pages() seems like the right way to go.

TODO extract this to something like is_a_hierarchical_post_type( $post_type )

We already have is_post_type_hierarchical()

Thank you very much. I could even leave the further evolvement of get_pages() for later, just provide what I have after testing as a first step.

Also thanks for the tip about is_post_type_hierarchical. I saw another place in the code where it wasn't being used, and I might provide another patch for that after.

Great to be collaborating with you, scribu.

[scribu]

Thank you very much. I could even leave the further evolvement of get_pages() for later, just provide what I have after testing as a first step.

You're welcome. You should upload an improved patch as soon as you have it; old versions will still be available, if needed.

[dbernar1]

I made a note while refactoring the code:

"get_pages() had this additional sorting field. Figure out what to do about this one: $allowed_keys = array( 'modified_gmt', 'post_modified_gmt' );"

Basically, WP_Query::get_posts has:

"$allowed_keys = array('name', 'author', 'date', 'title', 'modified', 'menu_order', 'parent', 'ID', 'rand', 'comment_count');" on line 2332

which means it does not support post_modified_gmt, from what I can tell. What shall we do about that?

[scribu]

We should make WP_Query support ordering by 'modified_gmt' (and 'date_gmt', while we're at it).

[atimmer]

We really need unit tests before we change this around. Patch needs refresh + a lot of expanding.

I will be working on this, goal will be 3.8.

[mdawaffe]

WP_Query

[mdawaffe]

Not knowing about this ticket or the previous work, I created a patch that makes get_pages() a wrapper for WP_Query (as get_posts() is already).

Oddities in attachment:12821-wp-query.diff​​:

• Having to use array_intersect_key() on the arguments since wp_dropdown_pages() has a name argument which is not compatible with WP_Query's. There are other ways to resolve this.
• Lack of post_modified sort in WP_Query.
• As written in this patch, the cacheing is handled a bit differently. I think the post terms/meta are cached with this patch, but not in the current version. We could change that with the update_post_meta_cache and update_post_term_cache WP_Query arguments.

[helen]

#32077 was marked as a duplicate.