TIP: Using "AND" selectors across multiple fields (named selectors to the rescue)

[LMD]

I don't know if this is the right place, it's not really a tutorial, just a tip based on notes I wrote myself in a recent project to get it straight in my own head. I thought it might be useful for others in a similar situation.

Scenario:

Create a search function that will search for keywords "foo" and "bar" in multiple fields, but the keywords do not have to be adjacent, in order, or even all in the same field. For eample, the selector must match if "foo" is in "field_a" and "bar" is in "field_b" -- so long as both keywords are present somewhere, the page match is valid.

It is possible to just split the terms and do multiple queries on each field separately and then combine the results into a single PageArray for pagination (I believe there is a module that helps with this). However, I wanted to see if it was possible to do a basic version with a single query.

Not The Solution:

The following selector does not work when keywords appear separately in different fields (operator '~=' - contains all the words):

$selector = "title|field_a|field_b~=foo bar";

What the selector is saying:

FIND BOTH "foo" AND "bar" IN title
OR
FIND BOTH "foo" AND "bar" IN field_a
OR
FIND BOTH "foo" AND "bar" IN field_b

In this case, both "foo" and "bar" have to be in the same field (but not adjacent or in order) to match.

The Actual Solution

What we need to use is "named selectors" to let us match each individual keyword separately while still using one selector.

Using the same example as before:

$selector = "selector1=(title|field_a|field_b~=foo), selector2=(title|field_a|field_b~=bar)";

What the selector is saying at its most basic level:

FIND BOTH selector1 AND selector2

Or, to expand on this, it is saying:

(FIND "foo" IN title OR field_a OR field_b)
AND
(FIND "bar" IN title OR field_a OR field_b)

Crucially, "foo" and "bar" do not have to be in the same field to match.

Practical Method

In this example code, I am actually allowing the search for phrases (using "quoted text") as well as individual terms, so a person could enter...

"foo bar" baz

... and it will keep "foo bar" together aa one term and "baz" as a separate term and match them as an exact phrase.

// Keywords obtained from $input->get and cleaned (multiple spaces removed)/sanitized etc.
$keywords = '"foo bar" baz';

// Split into individual search terms by space (preserve spaces in quoted text)
$terms = str_getcsv($keywords, " "); // array("foo bar", "baz")

// Build up named selectors
$ns = ""; // named selectors string
$i=1; // named selector count

foreach ($terms as $term) {
    // operator '*=' - contains the exact word or phrase
    $ns .= ", ns{$i}=(title|field_a|field_b*=" . trim($term) . ")";
    $i++;
}
//$ns = ", ns1=(title|field_a|field_b*=foo bar), ns2=(title|field_a|field_b*=baz)"

// Construct the whole selector (modify/add other general selectors as needed)
$selector = "template=my-template, limit=20, sort=-date" . $ns;

// Find pages based on selector
$results = $pages->find($selector);

 

DISCLAIMER

I haven't done any tests to see if this method is more efficient than running queries on each field separately and combining the results, I just wanted to see if it was possible!

[Tom.]

I think you can also do:

$selector = "title|field_a|field_b~=foo|bar";

 

[LMD]

1 hour ago, Tom. said:

I think you can also do:

$selector = "title|field_a|field_b~=foo|bar";

 

Not quite, that would return pages where only ONE of the search terms is present in any of the fields ("foo" OR "bar").

FIND EITHER "foo" OR "bar" IN title
OR
FIND EITHER "foo" OR "bar" IN field_a
OR
FIND EITHER "foo" OR "bar" IN field_b

So it would match a page where only "foo" is contained in any (or all) of the fields -- but "bar" must also be present somewhere.

Likewise, this would not work either:

$selector = "title~=foo|bar, field_a~=foo|bar, field_b~=foo|bar";

FIND EITHER "foo" OR "bar" IN title
AND
FIND EITHER "foo" OR "bar" IN field_a
AND
FIND EITHER "foo" OR "bar" IN field_b

In this case, "foo" or "bar" must be present in all the fields, which is also not what we want

[szabesz]

Thanks for the tip @LMD

It is worth to note the new "tags for fields" search "shortcut" of ProcessWire 3.0.106:
https://processwire.com/blog/posts/pw-3.0.106/

I guess title|field_a|field_b in your example could be simplified too by using this new feature. Provided it supports "named selectors".

[dragan]

@LMD Thanks for this post. I'll definitely keep this bookmarked.