module SearchEngine

[adrian]

Thanks @teppo for the new dev version - been playing around with that and the new auto desc and highlighting changes seem great!

[adrian]

Sorry @teppo - just noticed a character encoding issue. I have an excerpt (plain textarea) field which is being indexed and I am seeing this in the results:

https://ian.umces.edu/search/?q=dennison

That � is a normal single quote character which you can see here: 

https://ian.umces.edu/blog/page2/

Let me know if there is any other info I can provide.

[adrian]

@teppo - just took a better look at this and it is related to the HTML Entity Decoder textformatter that I have applied to the excerpt field. If I dump the content of $index in the Renderer::getResultAutoDesc method, I see this:

'Tony Larkum's 80th Birthday Party ... After attending a former colleague's birthday party, Bill Dennison recounts fond memories of time he spent with Tony Larkum in Australia and the United States that spanned from the 80's through early 2000's.

What is interesting to me is that the ' in "former colleague's birthday" looks as expected, but the one in "the 80's through" is converted to that � icon.

If I add this line: 

$index = $this->wire('sanitizer')->unentities($index);

then it works as expected, although I don't think we want to do that. I feel like this is somehow related to one of the regexes and a " 's " right at the end of the autodesc. Maybe it's truncating the 's to something like &#03 which it doesn't know how to display?

[adrian]

I just changed the $desc_max_length from 255 to 265 and that fixed it, so I think my last suggestion about it truncating in the middle of the 's is correct.

Here is the output of $value from ___renderResultDesc() which confirms it!

<div class="{classes.result_desc}">Tony Larkum's 80th Birthday Party ... After attending a former colleague's birthday party, Bill <strong class="{classes.result_highlight}">Dennison</strong> recounts fond memories of time he spent with Tony Larkum in Australia and the United States that spanned from the 80&#0...</div>

 

[teppo]

2 hours ago, adrian said:

I just changed the $desc_max_length from 255 to 265 and that fixed it, so I think my last suggestion about it truncating in the middle of the 's is correct.

Seems reasonable. In fact I had similar issues earlier, but those were fixed by switching from substr() to mb_substr(). This case seems a bit different, though. I'll try to set up a test case and see what I can do about that; so far I've only managed to end up with a partial encoded entity, 80& instead of 80' and so on ?

Meanwhile I noticed that in some cases the new auto desc feature generates odd looking "......" character sequences and other weird results, so there's more to tweak anyway. I'll try to dedicate some time to this in the next few days.

Quote

HTML Entity Decoder textformatter

Just to make sure: did you mean HTML Entity Encoder?

[teppo]

On 4/13/2021 at 12:49 PM, ngrmm said:

hi @teppo, thanks for the module. it's great!

i have pages which are selected through a page reference field from inside a table which is inside a repeater which is inside a repeater matrix item
(matrix_repeater) -> (matrix_repeater_item)  -> (repeater) -> (table) -> (table column [page reference field]) -> title or text_field

the search_index has already the title of those pages in it. is there an easy way to also get the other fields of the reference page inside the search_index.

Thanks!

Currently "id", "name", and "title" are indexed automatically for page reference fields. If a programmatic approach is easy enough, you can modify this behaviour and add more fields by hooking before Indexer::getPageReferenceIndexValue() and modifying the $indexed_fields argument. Alternatively you could hook into this method and in specific cases replace the behaviour completely, for an example by returning the output of the indexPage() method for each individual page instead (here's a loosely related example).

Let me know if you need additional pointers and I can provide some sample code.

[ngrmm]

14 minutes ago, teppo said:

Thanks!

Currently "id", "name", and "title" are indexed automatically for page reference fields. If a programmatic approach is easy enough, you can modify this behaviour and add more fields by hooking before Indexer::getPageReferenceIndexValue() and modifying the $indexed_fields argument. Alternatively you could hook into this method and in specific cases replace the behaviour completely, for an example by returning the output of the indexPage() method for each individual page instead (here's a loosely related example).

Let me know if you need additional pointers and I can provide some sample code.

ok, I'm a noob but I'll give it a try. I will use my loops and add the needed content to the searchindex.
But where should i place this hook to run it after a page is saved in the backend?

[adrian]

36 minutes ago, teppo said:

I'll try to dedicate some time to this in the next few days.

Thanks @teppo

This exact text in a textarea with a HTML Entity Encoder textformatter should allow you to replicate the issue.

After attending a former colleague's birthday party, Bill Dennison recounts fond memories of time he spent with Tony Larkum in Australia and the United States that spanned from the 80's through early 2000's.

 

37 minutes ago, teppo said:

Just to make sure: did you mean HTML Entity Encoder?

Yes, sorry about that ?

[teppo]

26 minutes ago, ngrmm said:

ok, I'm a noob but I'll give it a try. I will use my loops and add the needed content to the searchindex.

Great — let me know if you run into any trouble and I'll be happy to help!

26 minutes ago, ngrmm said:

But where should i place this hook to run it after a page is saved in the backend?

If you mean which file, for this purpose (saving pages in the admin) you could put the hook into site/templates/admin.php, but site/init.php is likely just as good (and also catches page saves that occur outside the admin).

[ngrmm]

5 minutes ago, teppo said:

Great — let me know if you run into any trouble and I'll be happy to help!

I found another solution and it works for me. First I loop through my pages and save the needed content into a helper-textarea-field. I added this helper-field into the indexed fields and this works. I guess this is not the best for the performance but it works for me.

Thanks again

[adrian]

@teppo - I looked into the issue with truncation of entities a little more and it seems like there are a couple of different approaches to fix this. The simplest seems to be to simply remove any trailing bits of entity that is left over, eg: 

  $value = rtrim(preg_replace('/(?:<(?!.+>)|&(?!.+;)).*$/us', '', $value));

as mentioned here: https://www.drupal.org/project/drupal/issues/2279655#comment-8843201 - there is some other useful discussion there about this issue.

Another approach I've come across is to calculate the length by treating the entities as a single character like they detail here: https://gist.github.com/andykirk/b304a3c84594515677e6 and https://alanwhipple.com/2011/05/25/php-truncate-string-preserving-html-tags-words/

I tried the first approach and it seems to work fine. Hopefully this might save you a bit of time. This is where I've put it for now:

    protected function formatResultAutoDesc(string $match, string $index, string $desc): string {
        if ($match !== '') {
            $match_length = mb_strlen($match);
            $add_prefix = (empty($desc) || mb_substr($desc, -3) !== '...') && (mb_strpos($match, '...') === 0 || mb_substr($index, 0, $match_length) !== $match);
            $add_suffix = mb_substr($index, -$match_length) !== $match || mb_strrpos($match, '.') !== $match_length;
            // Remove scraps of HTML entities from the end of a strings
            $match = rtrim(preg_replace('/(?:<(?!.+>)|&(?!.+;)).*$/us', '', $match));
            $match = ($add_prefix ? '...' . $match : $match) . ($add_suffix ? '...' : '');
        }
        return $match;
    }

 

[teppo]

11 hours ago, adrian said:

I tried the first approach and it seems to work fine. Hopefully this might save you a bit of time.

Thanks, @adrian. I agree that this is the simplest solution, and probably the most sensible one right now. It's now included in the dev branch of the module ?

[adrian]

Hi @teppo - I've been playing around with multiple operators (https://processwire.com/docs/selectors/operators/#using-more-than-one-operator-at-a-time). I think you mentioned that these aren't supported, so please feel free to ignore this request, but what I discovered when using this operator combination: *=~= is that the grouping feature no longer works if PW fails to find any results via *= and moves on to ~=

All results end up being listed under the All tab with no other tabs shown. For now I've gone back to my old approach to adding the tabs which works with this operator. 

I have also played around with the #= operator which I think will need some special handling to support because of the need to additional additional quotes and escaping. 

Just rambling out loud here - I also played around with adding basic support for swapping between *= and ~= depending on whether the user's search term includes double quotes. I almost stuck with that approach but I am honestly not sure how many non-tech users know about quoting search terms, so I thought the *=~= was probably better for most people.

[adrian]

@teppo - sorry for the barrage. Again, don't worry about this if it's not important to you, but I noticed an issue with highlighting in the auto description if there is punctuation in the search, eg:
https://ian.umces.edu/search/?q=bird's+head

Note the last result: "Science communication training for West Papua, Indonesia" doesn't highlight "Bird’s Head" because it's not a regular single quote character used for the apostrophe. I wonder if highlight matching can / should ignore punctuation characters?

[2hoch11]

I had this moster on my search result page:

$selector = "template=produkt, check2|check_temp=1, (title|text|text_editor_minimal|text_editor|text_editor2|text_editor3|search_index~=$q), (produkt_hersteller.title%=$q), (produkt_bereiche.title%=$q), (produkt_anwendungen.title%=$q), (produkt_verwendungen.title%=$q), (produkt_geratetypen.title%=$q), (produkt_gruppen.title%=$q), (produkt_verdichter.title%=$q), (produkt_kaltemittel.title%=$q), (produkt_gwp.title%=$q)";

It worked so far. After I updated PW to 3.0.165 I got an Internal Server error. So far so good. I updated SearchEngine too. No change. I removed piece by piece parts of the selector. Finally I realised that the subfield selectors (like produkt_hersteller.title) caused the error.

Now I ended with this selector:

$selector = "template=produkt, check2|check_temp=1, (title|text|text_editor_minimal|text_editor|text_editor2|text_editor3|search_index%=$q)";

Has anybody an idea why subfield selectors don't work anymore?

[mlfct]

Hey @teppo,

i got a quick question - maybe it is even trivial:

I just upgraded from SE 0.26 to 0.30.1 and i am running into some issues with the JSON output. Up until now it was working flawlessly, when i was querying my search endpoint via AJAX. The output was just dead simple:

$searchEngine = \ProcessWire\modules()->get('SearchEngine');
header('Content-Type: application/json');
return $searchEngine->renderResultsJSON();

But somehow this endpoint returns no results on 0.30.1, when using the same query as on 0.26:

{
  "query": "xyz"
}

I also did a quick debug query via the SE admin panel, which returns all the expected results.

Am i overlooking some changes?

[teppo]

On 5/21/2021 at 2:48 PM, 2hoch11 said:

I had this moster on my search result page:

$selector = "template=produkt, check2|check_temp=1, (title|text|text_editor_minimal|text_editor|text_editor2|text_editor3|search_index~=$q), (produkt_hersteller.title%=$q), (produkt_bereiche.title%=$q), (produkt_anwendungen.title%=$q), (produkt_verwendungen.title%=$q), (produkt_geratetypen.title%=$q), (produkt_gruppen.title%=$q), (produkt_verdichter.title%=$q), (produkt_kaltemittel.title%=$q), (produkt_gwp.title%=$q)";

It worked so far. After I updated PW to 3.0.165 I got an Internal Server error. So far so good. I updated SearchEngine too. No change. I removed piece by piece parts of the selector. Finally I realised that the subfield selectors (like produkt_hersteller.title) caused the error.

Now I ended with this selector:

$selector = "template=produkt, check2|check_temp=1, (title|text|text_editor_minimal|text_editor|text_editor2|text_editor3|search_index%=$q)";

Has anybody an idea why subfield selectors don't work anymore?

This seems like an issue with ProcessWire itself, not SearchEngine. If you have a chance you might want to give the latest dev version of ProcessWire a try, though if it's a live site then I'd recommend doing it locally / in a development environment first, or alternatively backing your site up before the update (just in case).

If that doesn't solve it, you may want to open an issue for this at https://github.com/processwire/processwire-issues. "Internal Server error" alone isn't very informative, so you may also want to dig into your log files etc. to see if there's anything more specific there.

[teppo]

On 5/27/2021 at 5:51 PM, mlfct said:

I just upgraded from SE 0.26 to 0.30.1 and i am running into some issues with the JSON output. Up until now it was working flawlessly, when i was querying my search endpoint via AJAX. The output was just dead simple:

$searchEngine = \ProcessWire\modules()->get('SearchEngine');
header('Content-Type: application/json');
return $searchEngine->renderResultsJSON();

But somehow this endpoint returns no results on 0.30.1, when using the same query as on 0.26:

{
  "query": "xyz"
}

I also did a quick debug query via the SE admin panel, which returns all the expected results.

Am i overlooking some changes?

This seems like a bug. So far I've traced it back to version 0.29.0, but will have to dig in further to see what's causing it. This version introduced major changes behind the scenes, so I'm not entirely surprised that something went wrong.

I'll get back to you once I figure this out.

[teppo]

@mlfct, the issue mentioned above should now be fixed in SearchEngine version 0.30.2. Thanks for letting me know about this and sorry for taking so long to solve it ?

[gornycreative]

@adrian How are you able to extract image information along with the rest of the rendered results? I've tried adding item.image_field to the templates in a number of configurations and I can get the raw filename but not any of the other page image fields, e.g. httpUrl, height/width, etc.

[adrian]

19 minutes ago, gornycreative said:

@adrian How are you able to extract image information along with the rest of the rendered results? I've tried adding item.image_field to the templates in a number of configurations and I can get the raw filename but not any of the other page image fields, e.g. httpUrl, height/width, etc.

I added this to my search.php template file. Hope that helps.

$searchEngine->addHookAfter('Renderer::renderResult', function($event) {
    $p = $event->arguments[0];
    $image = $p->getunFormatted('pdf_images|image|images|video_images')->first();
    if($image) {
        $thumb = $image->size(0, 260, array('upscaling' => false));
        $event->return = '
        <div class="row">
            <div class="small-12 medium-4 large-5 xlarge-4 xxlarge-3 columns teaser__image">
                <a href="'.$p->url.'">
                    <img style="max-height: 300px" src = "'.$thumb->url.'" />
                </a>
            </div>
            <div class="small-12 medium-8 large-7 xlarge-8 xxlarge-9 columns teaser__content">
            ' . $event->return . '
        </div>';
    }
});

 

[gornycreative]

@adrian Thanks, it does. I wasn't sure if I'd need to dig into hooks on this or not.

@teppo Just to confirm, looking at Renderer it also seems like there isn't a method off the bat to get the paginationString from the query... just count and total.

Looks like I'd need to hook into renderResultsListSummary with a pagination check for that? I think I could get it from the magic get method?

[teppo]

18 hours ago, gornycreative said:

@teppo Just to confirm, looking at Renderer it also seems like there isn't a method off the bat to get the paginationString from the query... just count and total.

Looks like I'd need to hook into renderResultsListSummary with a pagination check for that? I think I could get it from the magic get method?

Depends on where and how you want to use this ?

The query object has "pager" property that returns a rendered pager, and Renderer has public method (Renderer::renderPager(array $args, Query $query)) that does the same. Whether or not these will be useful depends, again, largely on your context.

[gornycreative]

This has been great. I just wanted to confirm one thing I noticed which is that _auto_desc doesn't seem to highlight/load a summary on partial word matches, although the search itself picks them up.

So for example if I search for 'business' using a partial match operator, any article search_index that includes the whole word 'business' will create a summary with the word 'business' marked... and results will show up where businesses and businessmen are in the index, but no summary with a highlighted partial word appears. The summary for these entries is blank.

Is this the intended behavior?

 

[tires]

I chose the Operator:
~=
(i.E. All given words appear in compared value, in any order. Matches whole words. Uses “fulltext” index.)

But if i enter "everything 2021" i am not finding the page with the titel "everything is great in 2021".