Suggestion Backlog Slog, 8/22/2018

[RyanCavanaugh]

• Equality operator with unknown does not narrow the type of the unknown #25172 Narrowing unknown with === or friends
• Allow to use null as discriminator for tagged union #24193 null should be an allowable discriminant value
• type.isLiteral() returns false for boolean literals #26075 Should type.isLiteral() return true for boolean literals?
• Proposal: allow type arguments of generic being named #22631 Named type arguments
• Error when accessing private property in object destructuring assignment with computed name #26572 Destructuring computed private class property names
• Consider allowing access to UMD globals from modules #10178 Access UMD globals from modules
• type has no index signature #14951 Index signatures on fresh object literals
• Option for strictPropertyInitialization to check even potentially undefined properties #26015 Even stricter class property initialization
• Enhance DOM API completion by sourcing descriptions from MDN #26404 Scrape MDN for DOM/core documentation comments
• esModuleInterop does not implicitly enable allowSyntheticDefaultImports #26193 esModuleInterop should enable allowSyntheticDefaultImports
• Operator to ensure an expression is contextually typed by, and satisfies, some type #7481 "soft" type assertion operator
• Assume arity of tuples when declared as literal #24350 Should array literals be tuples?
• Proposal: new "invalid" type to indicate custom invalid states #23689 "invalid" type

[RyanCavanaugh]

• Equality operator with unknown does not narrow the type of the unknown #25172 Narrowng unknown with === or friends
  • OP is not super compelling
    • If you have a known RHS, why aren't you using the RHS?
  • Last case with Response is more compelling
    • Reminder that minimal examples often lose their compellingness
  • 👍 assuming no bad side effects
• Allow to use null as discriminator for tagged union #24193 null should be an allowable discriminant value
  • Comes up with node packages a lot
  • Didn't we do something here recently?
    • Allowing mixed-unit-type discriminants?
  • Treat null as a primitive for the purpose of detecting discriminated unions
  • 👍
• type.isLiteral() returns false for boolean literals #26075 Should type.isLiteral() return true for boolean literals?
  • Yes!
• Proposal: allow type arguments of generic being named #22631 Named type arguments
  • No
• Error when accessing private property in object destructuring assignment with computed name #26572 Destructuring computed private class property names
  • 👍 stop the madness, Won't Fix this bug
  • Daniel will add a testcase
• Consider allowing access to UMD globals from modules #10178 Access UMD globals from modules
  • Open question: If you import from that module, is it poisoned then ?
    • NOBODY CARES
  • No satisfactory workaround because there's no namespace import
  • What do we call this flag?
    • Bikesheeeeeed
  • 👍
• type has no index signature #14951 Index signatures on fresh object literals
  • It's clever? Or is it. Definitely!
  • Proposal: Fresh object literals get an index signature that widens away
    • Kind of expensive and weird
  • Proposal: Do something special when indexing an object literal
    • Walk through numeric/string names as appropriate and make a union with undefined
    • Cheap and targeted
  • 👍
• Option for strictPropertyInitialization to check even potentially undefined properties #26015 Even stricter class property initialization
  • Can't break the world - yet another flag
  • Effectively a syntactic enforcement
  • The control flow checker is kind of half-off in this case
  • 👎
• Enhance DOM API completion by sourcing descriptions from MDN #26404 Scrape MDN for DOM/core documentation comments
  • 👍 Community
• esModuleInterop does not implicitly enable allowSyntheticDefaultImports #26193 esModuleInterop should enable allowSyntheticDefaultImports
  • 👍 sure
• Operator to ensure an expression is contextually typed by, and satisfies, some type #7481 "soft" type assertion operator
  • Desired: "as" with one-way assignability semantics but still have contextual typing
  • What terrible operator name would we come up with?
  • We could use Flow's type assertion syntax - it's reserved
  • We want it but all available syntax (cough) is terrible
  • Needs a super clever syntactic insertion point
  • Related: Block-syntax type assertion
    • Speculating on how to make this happen in a performant way
  • Keep noodling on it
• Assume arity of tuples when declared as literal #24350 Should array literals be tuples?
  • Is the breakage not as bad as we thought?
  • General literal-assertion problem
    • Literal types are really problematic today
    • What would a smarter world look like?
  • ! should be the "do the other thing" operator?
    • ... no
  • We shouldn't hurt non-tuple-users to help tuple users, arguably
  • Keep noodling on it

[mmkal]

@RyanCavanaugh what about #23689?? There doesn't seem to be a 👍 or 👎 for it (very much hoping for 👍)!