Hi Tim, Thanks for the explanation.
In this example, $dataWithOnlyIdAndEmail = \array_intersect_key($data, ['id' => 0, 'email' => 1]); the second argument is an associative array. But in my *array_only(array $input, array $keys) *function, *$keys* is a *numeric array.* Working with a *numeric array* may be slightly faster than building an associative array first. Just wanted to clarify that point. Best regards, Muhammed Arshid KV On Sun, 1 Mar 2026, 7:40 pm Tim Düsterhus, <[email protected]> wrote: > Hi > > Am 2026-02-22 19:58, schrieb Muhammed Arshid KV: > > Here is a simple example of an `array_only()` implementation using > > existing > > PHP functions: > > > > ```php > > function array_only(array $input, array $keys): array { > > return array_intersect_key($input, array_flip($keys)); > > } > > ``` > > > > This works, but `array_flip($keys)` creates an extra temporary hash > > table. > > So peak memory becomes: input + keys + flipped array + result. > > That is true for this specific implementation of a “wrapper function”. > It is not necessarily true for other implementations, e.g. simply using > the functions directly instead of creating a wrapper for two method > calls. `array_flip()` supports compile-time evaluation, so you can just > write: > > $dataWithOnlyIdAndEmail = \array_intersect_key($data, > \array_flip(['id', 'email'])); > > and OPcache will make sure to rewrite it to: > > $dataWithOnlyIdAndEmail = \array_intersect_key($data, ['id' => 0, > 'email' => 1]); > > avoiding the intermediate array. > > Best regards > Tim Düsterhus > > PS: It appears that the RFC is not listed in the overview at > https://wiki.php.net/rfc. >
