Suggestion Backlog Slog, 4/20/2016

[RyanCavanaugh]

• negating type constraints #7993 Negative type constraints
• Better display representation for intersection types #7705 Better display of intersection types
• Destructured parameter should not enforce presence / non-nullability from type annotation on a property that's default-initialized #7678 Destructured parameter should not enforce presence / non-nullability from type annotation on a property that's default-initialized
• Allow ArrayLike as a spread source #7596 Allow ArrayLike as a spread source
• Suggestion: readonly constructor parameters #7590 readonly constructor parameter properties
• Excess properties are not checked e.g. in callbacks #7547 Excess properties not checked in callbacks
• Suggestion: Implied Promise return type in async functions #7284 Implied Promise return type in async functions
• Allow super constrains for type parameters #7265 Allow super constraints for type parameters
• JSX: a way to make all the properties of value-based elements optional #7004 A way to make all properties of value-based elements optional
• More specific JSON parse reviver argument in lib.d.ts #6955 More specific reviver type in JSON#parse
• Readonly indexer and constructor usage: question #6781 Readonly indexer and constructor usage
• Allow defining static index signature #6480 Allow defining static index signature
• Leak of noImplicitAny checking with private static property #6415 Leak of noImplicitAny checking with private static property
• Provide an official set of Coding Guidelines #878 Provide official coding guidelines
• Partial classes  #563 Partial classes
• Special bottom type #3076 Special 'bottom' type
• Default generic type variables #2175 Default generic types

[RyanCavanaugh]

Postponed to Monday

[RyanCavanaugh]

• negating type constraints #7993 Negative type constraints
  • Ryan's 'workaround' is disgusting 😄
  • See if other scenarios appear
• Better display representation for intersection types #7705 Better display of intersection types
  • Agree the proposed behavior is desirable
  • There are potentially some differences between T & U and its decomposed form?
    • We can't think of any
  • What do we do if one is an object type and one isn't?
    • Just do this for the anonymous types of the intersection
  • It's actually really hard and dangerous
    • Because anonymous types can be self-referentially recursive
    • Potential for a lot of runaway bugs
  • Close to Too Complex given the risk/reward tradeof
  • Accepting PRs but they need a giant test suite
• Destructured parameter should not enforce presence / non-nullability from type annotation on a property that's default-initialized #7678 Destructured parameter should not enforce presence / non-nullability from type annotation on a property that's default-initialized
  • Sympathetic that this is annoying
  • Real problem is that there's no syntactic room in Component1 to put a type annotation on just propC
  • Axiom: Type annotations always win
  • 👎 Too Complex
• Allow ArrayLike as a spread source #7596 Allow ArrayLike as a spread source
  • If we look for ArrayLike, this would work
    • Compat problem with ES6 and surrogate pairs on strings
  • Also enables other good behavior like DOM collections
  • Does anyone spread a string on purpose?
  • 👍 Accepting PRs and live with surrogate pair compat issue
• Suggestion: readonly constructor parameters #7590 readonly constructor parameter properties
  • Why didn't this work before? Unsure (keyword soup fear?)
  • Committed for 2.0
  • 👍
• Excess properties are not checked e.g. in callbacks #7547 Excess properties not checked in callbacks
  • Agree this is not the ideal behavior
  • Caused by widening the return type of function expressions
  • Looking at Don't widen return types of function expressions #241 trying to figure out its status
  • Why do we have this behavior today?
    • The types that go in to widen only come from checkExpression, which have a limited domain of non-circular types
    • So we'd need to modify the widening logic to handle self-referential recursive types
  • Confusing but workaround is simple (add a return type annotation to function expression)
  • If Don't widen return types of function expressions #241 keeps happening, address it
  • Too Complex but keep an eye on function expression return type widening
• Suggestion: Implied Promise return type in async functions #7284 Implied Promise return type in async functions
  • Working as intended - Spooky action at a distance hinders readability
  • 👎
• Allow super constrains for type parameters #7265 Allow super constraints for type parameters
  • Dupe JSX: a way to make all the properties of value-based elements optional #7004
• JSX: a way to make all the properties of value-based elements optional #7004 A way to make all properties of value-based elements optional
  • Architectural/perf concerns about the amount of indirection this requires
    • e.g. what is the internal behavior of Optionalize<string> ?
    • Generics prevent assumptions here
    • Important is that union/intersection types are always top-level
  • Very large work item, but this appears in a lot of frameworks
    • e.g. potential to construct previously unconstructable things (e.g. union types which have sub-union types as members)
  • Think about it some more
• More specific JSON parse reviver argument in lib.d.ts #6955 More specific reviver type in JSON#parse
  • Breaking change for people who got a contextual any and indexed into an array
    • Seems unlikely and fix is very simple
  • 👍
• Readonly indexer and constructor usage: question #6781 Readonly indexer and constructor usage
  • What's the intent of a readonly index signature? Disagreement
  • What's the actual code? Need to understand the scenario better
• Allow defining static index signature #6480 Allow defining static index signature
  • 👍 Accepting PRs, should be reasonably straightforward
• Leak of noImplicitAny checking with private static property #6415 Leak of noImplicitAny checking with private static property
  • A type query should never generate an implicit any
  • 👍 but fix is that it should be illegal to use the typeof a private member to annotate a public member
• Provide an official set of Coding Guidelines #878 Provide official coding guidelines
  • Nope
• Partial classes  #563 Partial classes
  • Just sugar for interface A { methods; }, A.prototype.methods = ...
  • Allow property declarations and methods, but not property initializers (no way to emit)
  • Allow statics as well (including initializers, since we can emit those)
  • People are already sort of doing this already
  • Bikeshed partial keyword
    • extension class is the name C# would have called it but there isn't a clear "master" declaration like there is here
    • Confuse people by making it almost the same as C#, but not?
    • Or have a new keyword that we have to explain is 'close to' partial
    • Which will be preferable?
• Special bottom type #3076 Special 'bottom' type
  • Out of time
• Default generic type variables #2175 Default generic types
  • Out of time