binary encoding vs. asBuffer

[timkuijsten]

I'm wondering what the exact purpose of the asBuffer option is (and it's cousins keyAsBuffer and valueAsBuffer). To me it feels redundant since one can use keyEncoding binary and/or valueEncoding binary.

I wonder what to do when using IndexedDB as a backend which has support for native JS types as values and supports some different types for keys (Level/level-js#48). I know in level.js they even have a raw option. I can't really tell the difference between encoding, asBuffer and raw.

/cc @nolanlawson

[vweevers]

I'm wondering what the exact purpose of the asBuffer option is (and it's cousins keyAsBuffer and valueAsBuffer).

The asBuffer option tells the store to fetch data as a Buffer, and avoid unnecessary conversion. Stores like leveldown have the ability to return either a Buffer or string, both sourced from the same byte array. See encoding format for more.

To me it feels redundant since one can use keyEncoding binary and/or valueEncoding binary.

The binary encoding is just one of the encodings (which is user-specified) that result in asBuffer: true (which is derived from the encoding).

I wonder what to do when using IndexedDB as a backend which has support for native JS types as values and supports some different types for keys (Level/level-js#48)

TBD, to be discussed elsewhere.

I know in level.js they even have a raw option.

IMO the non-standard raw option should be removed from level.js. And if asBuffer is false, level.js should return the value as-is. Then, by using the identity encoding, one can achieve the same effect as raw: true.

[vweevers]

See also Level/level-js#58