* Make the non relative name resolution cache generic
* Add cache for non relative type reference resolution
* Add getter only methods so we can use these in future for sharing resolutions across the projects
* Mark existing non relative module name cache getOrCreate as deprecated
* Prefer getters for getting directory results
* Fix missed errors in switch when using union of literal and non-literal types (#38686)
This commit makes it so we don’t use the base type of literals when checking comparability in switch. The comparability checks handle that case already, is my understanding, so we don’t need to clobber the type before actually doing the check, causing missed errors.
When comparing the types in switch, if a union with a literal and a non-literal was used, the compiler in `checker.ts` would automatically get the base type of all parts of the union, resulting in missed errors. For example, if the union of the non-literal `number` and literal `"hello"` was compared to the literal `"world"` in a switch case, the compiler would miss that they’re actually not comparable.
Maybe someone can tell me why we were getting the base type before checking comparability, rather than relying on the logic within the comparability checks to handle literal/base type comparability?
* Fix lint (whitespace) issue in checker.ts by running lint with fix flag.
* Refactoring so CacheWithRedirects has Key and Value type parameters
* ModuleResolutionCache or TypeRefDirectiveCache will look in directory before solving, so ResolutionCache doesnt need this check
* Test showing module resolution is not shared because resolution cache doesnt update own options
* Enable traceResolution on some of the project reference tests
* Simplify CacheWithRedirects and ensure the options are set in all common scenarios so cache can be shared between redirects
* Make failedlookup etc optional in ResolvedModule/TypeRefefWithFailedLookupLocations
Also make accidental public failed lookup internal
* Add new API for module and type ref resolution
* Store auto type reference resolutions
* Modify test to show how using program partially doesnt report resolution diagnostics
* Ensure that resolution diagnostics are reported in filePreocessingDiagnostics so they can be reused when program is reused
* Some cleanup
* Remove the newly added ReoslutionInfo in favor of new APIs
* update
* fix services' type's isLiteral
* update literal completions tests
* initial prototype
* use symbol to expression. TODO: filter existing, replace import nodes
* WIP
* WIP
* remove booleans from literals
* trigger at case keyword positions
* clean up tests
* fix element access expression case
* refactor dealing with existing values into a tracker
* fix merge errors
* cleanup and more tests
* fix lint errors
* more merge conflict fixes and cleanup
* use appropriate quotes
* small indentation fix
* refactor case clause tracker
* experiment: support tabstops after each case clause
* address small CR comments
* fix completion entry details; add test case
* fix lint errors
* remove space before tab stops; refactor
* Refactoring so CacheWithRedirects has Key and Value type parameters
* ModuleResolutionCache or TypeRefDirectiveCache will look in directory before solving, so ResolutionCache doesnt need this check
* Test showing module resolution is not shared because resolution cache doesnt update own options
* Enable traceResolution on some of the project reference tests
* Simplify CacheWithRedirects and ensure the options are set in all common scenarios so cache can be shared between redirects
* When fsEvent for change is repeated
* When trying to check if program is uptodate, read the files from disk to determine the version instead of delaying so that new program is not created if file contents have not changed
* Fixed an issue with `in` not being able to be used on narrowed down expression of a generic nullable type
* Add another test case from a new issue
* Move the fix to `hasEmptyObjectIntersection`
* Numeric literal assignable to enum literal only when values match
* Accept new baselines
* Update compiler sources
* Accept new baselines
* Fix test runner
* Any numeric literal type is assignable to a computed numeric enum type
* Refactor `Extensions`, fix lookup priorities
* Update baselines
* Clean up
* Temporarily apply same fix to `node` for testing
* Revert ancestor node_modules lookups back to two passes
* Revert "Temporarily apply same fix to `node` for testing"
This reverts commit 4014f3005a.
* Update baselines
* Add new test for node_modules searching behavior
* Update new baseline
* Add tests where typeRef resolution is not in the cache
* Ensure that type ref reuse is actually handled since type ref resolution only sends containingSourceFileMode and not containingSourceFile
* Add known keys of the mapped type constraint to its members
* Avoid second pass of adding known members by instead passing `noReductions` to `mapType`
This debug code was added quite a while ago, constructed such that we
wouldn't have to ship this code to our users.
However, this is the sole place in the compiler project where the ts
namespace "escapes" the bundle. By moving this debug code into the
compiler itself, we no longer have any references to the ts namespace
itself for our bundles that don't export anything (tsc,
typingsInstaller). This lets bundlers tree shake the compiler, reducing
the size of our output by _5.7 MB_ (a ridiculous improvement for
_adding_ code).
This eliminates a significant number of dependencies, eliminating all
npm audit issues, speeding up `npm ci` by 20%, and overall making the
build faster (faster startup, direct code is faster than streams, etc)
and clearer to understand.
I'm finding it much easier to make build changes for the module
transform with this; I can more clearly indicate task dependencies and
prevent running tasks that don't need to be run.
Given we're changing our build process entirely (new deps, new steps),
it seems like this is a good time to change things up.