[19:11:44] <Krinkle>	 paladox: Regarding using PHP arrays for large key value files at Miraheze... Yeah, I'd say go for it. Especially if you need it in the critical path on every request. It's by far the fastest. Gets compiled into opcache and accessed directory from shared memory. So "require foo.php" won't actually open any files.
[19:13:16] <Krinkle>	 If it starts to get multiple megabytes large, then it might be worth looking into LLDB, or CDB files. These will involve some file open and read ops, so not as fast, but they'll cost you no memory and beat JSON files you have now because they won't have to be fully read and then fully decoded. It's a constant time access to a single key in the file.
[19:14:31] <Krinkle>	 If you're benchmarking locally ok the CLI, remember to turn on opcache to get a fair and realistic comparison. Example: https://gerrit.wikimedia.org/g/mediawiki/libs/less.php/+/c6922de56885cf01edd07d70478b71b6967a8a80/composer.json#62
[21:05:37] <paladox>	 Thanks!!