getBuffer() does not match the behavior of ioredis

[ryanwitt]

Consider the following example:

const Redis = require('ioredis-mock')
const redis = new Redis()
const value = Buffer.from('1T+HJWDNyTa4jJXwoBbV6Q==', 'base64')
await redis.set('foo', value)
const returnedValue = await redis.getBuffer('foo')
assert.equal(value.toString('base64'), returnedValue.toString('base64'))

Result:

Uncaught:
AssertionError [ERR_ASSERTION]: '1T+HJWDNyTa4jJXwoBbV6Q==' == '77+9P++/vSVg77+977+9Nu+/ve+/ve+/ve+/vRbvv73vv70='
    at REPL37:1:8
    at Script.runInThisContext (vm.js:134:12)
    at REPLServer.defaultEval (repl.js:566:29)
    at bound (domain.js:421:15)
    at REPLServer.runBound [as eval] (domain.js:432:12)
    at REPLServer.onLine (repl.js:909:10)
    at REPLServer.emit (events.js:412:35)
    at REPLServer.emit (domain.js:475:12)
    at REPLServer.Interface._onLine (readline.js:434:10)
    at REPLServer.Interface._line (readline.js:791:8) {
  generatedMessage: true,
  code: 'ERR_ASSERTION',
  actual: '1T+HJWDNyTa4jJXwoBbV6Q==',
  expected: '77+9P++/vSVg77+977+9Nu+/ve+/ve+/ve+/vRbvv73vv70=',
  operator: '=='
}

What's going on here? Well, it looks like the implementation for get always calls Buffer.toString() on values that were originally stored as buffers. This encodes the data returned as utf8, which breaks on bytes that don't have representation in unicode. From the node Buffer docs:

When decoding a Buffer into a string that does not exclusively contain valid UTF-8 data, the Unicode replacement character U+FFFD � will be used to represent those errors.

To bring this into compliance with ioredis, we likely need to change the getBuffer implementation to detect that a key was stored as a buffer and return it as such without interpretation. I think this will yield the same behavior as ioredis get and getBuffer.

Will send PR.

[stipsan]

🎉 This issue has been resolved in version 7.1.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀