Suggestion Backlog Slog, 10/31/2016

[RyanCavanaugh]

👻 "no one has a costume" edition 👻

• Suggestion: Allow classes to implement type aliases #11207 Allow classes to implement type aliases
• [Suggestion]: Add --strictReadonlyChecks option to catch runtime errors modifying frozen objects at compile time #11481 Add --strictReadonlyChecks option
• No error when attempting to read a write-only property #11596 Write-only properties
• Annotate immediately-invoked functions for inlining flow control analysis #11498 Annotate immediately-invoked functions for inlining flow control analysis
• Type guards do not work on array access expressions #11483 Type guards on const-accessed array index expressions
• Suggestion: Let null assertion (!) operator block "used before assigned" errors #11463 Let null assertion (!) operator block "used before assigned" errors
• Custom type guard fails to account for assignment within the argument expression #11271 Handle assignment in UDTG argument position
• TS1093: Can't specify return types on constructors #11588 Return type annotations on constructors
• Deleting from readonly dictionary does not give a type error #11480 delete operator on readonly properties / index signatures
• Interactive init #10964 Interactive tsc --init
• Start using ES5 functionality in tsc #10125 Use ES5 in tsc
• --strictNullChecks does not gaurd against optional arguments assignment #9450 Enforce callback optionality under --strictNullCheck
• Contextual typing does not flow backward through function return types to the function arguments #5256 Reverse-flow contextual typing

[RyanCavanaugh]

• Suggestion: Allow classes to implement type aliases #11207 Allow classes to implement type aliases
  • Why not? 👍
• [Suggestion]: Add --strictReadonlyChecks option to catch runtime errors modifying frozen objects at compile time #11481 Add --strictReadonlyChecks option
  • Would require the readonly T type operator
  • ... which would require variance annotations
  • .... which is enormous
  • e.g. Array / ReadOnlyArray
• No error when attempting to read a write-only property #11596 Write-only properties
  • Setters without getters are insane
  • Why do we even allow you to write set-only properties?
    • Make this an error
• Annotate immediately-invoked functions for inlining flow control analysis #11498 Annotate immediately-invoked functions for inlining flow control analysis
  • Could just be a flag on a function type?
  • Do we assume that it is invoked, or may be invoked?
    • Assume that it maybe isn't invoked
  • Establish a taxonomy of what kinds of function expressions and side effects we need to think about
• Type guards do not work on array access expressions #11483 Type guards on const-accessed array index expressions
  • Currently: Type guards don't apply to indexed access
  • Why not write const j = list[j]; ?
  • Too Complex for now due to array access
• Suggestion: Let null assertion (!) operator block "used before assigned" errors #11463 Let null assertion (!) operator block "used before assigned" errors
  • Problem is that we don't have any block-flowed assertions
  • We should come up with a way to flow in a statement-level assertion
  • 👍 does the same thing as existing behavior
• Custom type guard fails to account for assignment within the argument expression #11271 Handle assignment in UDTG argument position
  • Requires deeper syntactic analysis
  • How deep into the tree are we willing to walk?
  • Could just write if (x = foo.field, isDefined(x)) {
  • AMF
• TS1093: Can't specify return types on constructors #11588 Return type annotations on constructors
  • TC39 rejected async constructors and we do too
  • 👎
• Deleting from readonly dictionary does not give a type error #11480 delete operator on readonly properties / index signatures
  • 👍
• Interactive init #10964 Interactive tsc --init
  • Duplicately discussed
• Start using ES5 functionality in tsc #10125 Use ES5 in tsc
  • 👍
• --strictNullChecks does not gaurd against optional arguments assignment #9450 Enforce callback optionality under --strictNullCheck
  • Seems like a good idea, but is a BC break for --SNC users
  • But parameter bivariance, so oops
  • 👎 for rampant inconsistency
• Contextual typing does not flow backward through function return types to the function arguments #5256 Reverse-flow contextual typing
  • Fwd dupe to clearer issue Use return type as an inference location #11152
  • Think about it more