Description
Hello.
First of all thanks for the library.
My use case is that, using sizeCalculation, i want to be more or less strictly ensured that the LruCache instance is not consuming more memory then expected from sizeCalculation.
The problem is that i can't just do something like:
sizeCalculation: (k,v) => Buffer.byteLength(k, 'utf8') + Buffer.byteLength(v, 'utf-8').
because internally LruCache instance doesnt keep only one Map or Object for the key-values, It also save the data in bunch of arrays for indexing, like those ones from internals:
'starts', 'ttls',
'sizes', 'keyMap',
'keyList', 'valList',
'next', 'prev',
'head', 'tail',
'free', 'isBackgroundFetch',
'backgroundFetch', 'moveToTail',
'indexes', 'rindexes',
'isStale'
]
My question is: is there any correct and verified way to write the correct sizeCalculation implementation, to be more or less in sync with real memory consumption?
I tried something like this:
export const calcSize =
(hasTtl: boolean) =>
(val: any, key: string): number => {
let sum = 0;
const keySize = Buffer.byteLength(key, 'utf8');
let valSize;
if (val === undefined) {
valSize = 0;
} else {
valSize =
typeof val === 'string'
? Buffer.byteLength(val, 'utf8')
: Buffer.byteLength(JSON.stringify(val), 'utf-8');
}
// calc KeysMap key + index(int)
sum += keySize + INT_SIZE_BYTES;
// calc keyList key size
sum += keySize;
// cacl valList val size
sum += valSize;
// calc ttl start + ttl
// next arr
sum += INT_SIZE_BYTES;
// prev arr
sum += INSPECT_MAX_BYTES;
// size of the size itself
sum += INT_SIZE_BYTES;
// ttl start arr
if (hasTtl) {
sum += INT_SIZE_BYTES;
}
return sum;
};
but with the calculation it still shows me around 1.5x unsync with real memory