Fetching a cached Eloquent collection of 10 items times out - laravel

Question

I'm building a search functionality that returns large collections which are paginated using a LengthAwarePaginator. I'm trying to cache results using a key called $searchFilter_$query_$offsetPages for a single page of returned results (10 items). It goes into the cache just fine. However, it times out when I try to check using Cache::has($key) or fetch using Cache::get($key).
The same problem occurs in the browser as well as in artisan Tinker. Strangely, when I put a random set of 10 items into the cache in Tinker and fetch them back, everything works fine. I'm using Redis as the cache driver.
Here is my controller method:
public function search($filter, $query, $layout, Request $request) {
if($layout == "list-map") {
return view("list-map")->with(['filter' => $filter, 'query' => $query, 'layout' => 'list-map']);
} else {
$offsetPages = $request->input('page', 1) - 1;
$cacheKey = $filter . "_" . $query . "_" . $offsetPages;
if(Cache::has($cacheKey)) {
\Log::info("fetching results from cache");
$data = Cache::get($cacheKey);
$totalCt = $data[0];
$results = $data[1];
} else {
$results = $this->getResults($filter, $query);
$totalCt = $results->count();
$results = $results->slice($offsetPages, $this->resultsPerPage);
\Log::info("caching results");
Cache::put($cacheKey, [$totalCt, $results], 5);
}
$results = new LengthAwarePaginator($results,
$totalCt,
$this->resultsPerPage,
$request->input('page', 1),
['path' => LengthAwarePaginator::resolveCurrentPath()]
);
return view($layout)->with(['filter' => $filter, 'query' => $query, 'layout' => $layout, 'results' => $results]);
}
}

Answer 1

So, the issue was that many of the models in the collection returned from my getResults() method were obtained via relationship queries. When I would dd($results) on the single page of 10 results, I could see that there was a "relations" field on each model. Inside that array were thousands of recursively related models based on the relationship I originally queried. I was unable to find any information about an option to not eager load these related models. Instead I came up with a bit of a hacky workaround to fetch the models directly:
$results = $results->slice($offsetPages, $this->resultsPerPage);
//load models directly so they don't include related models.
$temp = new \Illuminate\Database\Eloquent\Collection;
foreach($results as $result) {
if(get_class($result) == "App\Doctor") {
$result = Doctor::find($result->id);
} else if(get_class($result == "App\Organization")) {
$result = Organization::find($result->id);
}
$temp->push($result);
}
$results = $temp;
\Log::info("caching results");
Cache::put($cacheKey, [$totalCt, $results], 5);
If anyone knows the best practice in this situation, please let me know. Thanks!
Edit:
I've found a better solution instead of the above workaround. If I query my relationships like this: $taxonomy->doctors()->get() rather than $taxonomy->doctors, it does not load in the huge recusive relations.

Answer 2

I dont really see why your code doesn't work. The only potential problems I see are the cache keys, which could contain problematic characters, as well as the way you check for a cached value. As you are using Cache::has($key) before Cache::get($key), you could end up with a race condition where the first call returns true and the latter null because the cached value timed out just between the two calls.
I tried to address both issues in the following snippet:
public function search($filter, $query, $layout, Request $request)
{
if($layout == "list-map") {
return view("list-map")->with(['filter' => $filter, 'query' => $query, 'layout' => 'list-map']);
} else {
$offsetPages = $request->input('page', 1) - 1;
$cacheKey = md5("{$filter}_{$query}_{$offsetPages}");
$duration = 5; // todo: make this configurable or a constant
[$totalCount, $results] = Cache::remember($cacheKey, $duration, function () use ($filter, $query) {
$results = $this->getResults($filter, $query);
$totalCount = $results->count();
$filteredResults = $results->slice($offsetPages, $this->resultsPerPage);
return [$totalCount, $filteredResults];
});
$results = new LengthAwarePaginator($results,
$totalCount,
$this->resultsPerPage,
$request->input('page', 1),
['path' => LengthAwarePaginator::resolveCurrentPath()]
);
return view($layout)->with(compact('filter', 'query', 'layout', 'results'));
}
}
The inbuilt function Cache::remember() doesn't use Cache::has() under the hood. Instead, it will simply call Cache::get(). As this function will return null as default if no cache was hit, the function can easily determine if it has to execute the closure or not.
I also wrapped the $cacheKey in md5(), which gives a consistently valid key.
Looking at the following part of your code
$results = $this->getResults($filter, $query);
$totalCount = $results->count();
$filteredResults = $results->slice($offsetPages, $this->resultsPerPage);
I am quite sure the whole search could be improved (independently of the caching). Because it seems you are loading all results for a specific search into memory, even if you throw away most parts of it. There is certainly a better way to do this.