Design Meeting Notes, 11/18/2016

[mhegazy]

Narrowing any and --noImplicitAny

• Compiler should allow assignment of new property in empty object
• Endless waiting for 2.1.1 compiling (unuseable)
  • consider caching resolved signatures for non-generic non-overloaded functions. this should save a lot of a lot of types
• it now behaves consistently with var x; x = {}; x.a = 0; as var x = {}; x.a = 0;
• the fix is to add : any
• ultimately support object construction pattern
• Proposals:
  • make assignment to {} to be any for control flow purposes
  • do nothing
• Conclusion: do it only in noImplicitAny mode

Async Iterators

• Stage 3 feature:
  • Proposal: https://github.com/tc39/proposal-async-iteration
  • Request for TS: Asynchronous Iterations #11326
• Downlevel support
  • Generator support already requested: ES6 Generator support #1564
  • Used by redux-saga: https://yelouafi.github.io/redux-saga/docs/introduction/BeginnerTutorial.html, ember-concurrency
  • Users already asking for down level iterator protocol support:
    • Generator with Array Spread #11970
    • Spread operator is not correctly translated into JS #8856
  • Perfromance:
    • Ron: Perf measurements, on the compiler code base, seems like for .. of is whiten the margin of error from generator Downlevel
  • concerns:
    • clean emit code for for .. of
      • we pride ourselves on the cleanness of the output
      • we believe our users find this an asset
    • performance of the output of all other
    • how Symbol.Iterator and Symbol.AsyncIterator be decided?
      • assume the existence of a polyfill
    • This needs a global Promise to be defined to work, there is no way to specify a specific promise type
  • Proposal:
    • either opt-in or opt-out
      • opt-in is better, so that not every one has to be
      • a new flag to allow generators
    • add a typescript-specific syntax for for each (x of y)
      • No. no reason to add new constructs to the language