New param for search in WP Query

[sebastian.pisula]

s => search post with phrase
without_s => without phrase

For example:

1. I want find movies without phrase "blood" in title or content. (without_s => blood)
2. I wand find movies with phrase "tiger" and without "die" in title or content. (without_s => die, s => tiger)

I think that good idea is, if param "s" will be have support for array. For example:

I can find post with words s = array('tiger', 'lion', 'dog')

[boonebgorges]

I think that good idea is, if param "s" will be have support for array.

Use search_terms. get_posts( array( 'search_terms' => array( 'tiger', 'lion', 'dog' ) ) )

I like the idea of NOT LIKE searches, but I wonder if we can come up with a more elegant syntax than 'without_s'. For example, we have 'search_terms'; maybe we could add something like 'search_terms__not_in'. Or maybe we could add boolean operators to the search, so that 'search_terms' => array( '-tiger', 'lion' ) will match posts *with* 'lion' but *without* 'tiger'. See ​https://support.google.com/websearch/answer/2466433?hl=en for a list of what Google supports.

[DrewAPicture]

Replying to boonebgorges:

I think that good idea is, if param "s" will be have support for array.

Use search_terms. get_posts( array( 'search_terms' => array( 'tiger', 'lion', 'dog' ) ) )

I like the idea of NOT LIKE searches, but I wonder if we can come up with a more elegant syntax than 'without_s'. For example, we have 'search_terms'; maybe we could add something like 'search_terms__not_in'. Or maybe we could add boolean operators to the search, so that 'search_terms' => array( '-tiger', 'lion' ) will match posts *with* 'lion' but *without* 'tiger'. See ​https://support.google.com/websearch/answer/2466433?hl=en for a list of what Google supports.

What about search_terms_exclude as an argument name? I also like the boolean operator approach, especially since it's already in use other places in core. The latter seems like a more-elegant approach.

[akibjorklund]

Terms that begin with a dash made most sense to me too.

I've attached a patch that implements this functionality as a first draft of what it might look like.

[boonebgorges]

So, it turns out that search_terms is listed in the docs, but doesn't actually work. WP_Query does pass around an internal tool called search_terms, but it can't be configured by parameter.

[boonebgorges]

Thanks, akibjorklund! Looks like it ended up being easier than I thought it'd be. 33988.diff​ cleans up the first part of your patch, and adds a unit test.

I think that your changes in parse_search_order() aren't quite right. parse_search_order() converts s to a couple of ORDER BY clauses: one for an exact match in the title, a few for combinations of single words in the title, and one for an exact match in the content. We don't want any exact matches, though. If you search foo -bar baz, you don't want to match a post that contains the sentence foo -bar baz. And you also don't want to match a post that has foo baz, which is what your patch will do. I think the right answer in these cases is to skip the sentence matches, but to keep the single word matches (negative terms already having been filtered out by this point). Does this seem right to others?

This is a neat little improvement. Let's try to do it for 4.4.

[akibjorklund]

Thanks for the feedback @boonebgorges! Your version looks good to me.

Only one small thought: If we skip the sentence match in ordering, then search results with negative terms will have different sorting. Consider first querying for foo bar and then altering that to filter out baz: foo bar -baz – now we present the results quite possibly in different order and also not in the best possible order of relevance – intuitively at the very least. But most likely the issue I described is so minor that it is not worth looking into.

[boonebgorges]

Only one small thought: If we skip the sentence match in ordering, then search results with negative terms will have different sorting. Consider first querying for foo bar and then altering that to filter out baz: foo bar -baz – now we present the results quite possibly in different order and also not in the best possible order of relevance – intuitively at the very least. But most likely the issue I described is so minor that it is not worth looking into.

Hrm. Yeah, I agree it's slightly strange. But at the same time, I think it makes sense that the NOT operator would sorta override the rest of the query. It doesn't seem obvious to me what the expected behavior in this case *ought* to be. So I think that what's in the patch is probably as good as anything.

[boonebgorges]

In 34934:

Allow excluded keywords when searching posts.

Pass a keyword with a leading hyphen to exclude posts containing that keyword.
For example, 'taco -onions' will return posts that contain the word 'taco' but
do not contain the word 'onions'.

Props akibjorklund.
Fixes #33988.

[GaryJ]

Fantastic improvement folks.

Is a unit test needed to clarify behaviour when the search field only contains negated terms? i.e.

• -foo
• -foo -bar

[boonebgorges]

Replying to GaryJ:

Fantastic improvement folks.

Is a unit test needed to clarify behaviour when the search field only contains negated terms? i.e.

• -foo
• -foo -bar

I don't know that it's needed, but if you wrote these tests, I would add them :)