* Clamp calculated sourcemap positions rather than throwing, to allow the underlying file to drift out of date with the sourcemap without a crash
* Clamp line as well
* constructor functions infer from call+construct
Also fix an incorrect combining of inferences for rest parameters: the
inferred types will be arrays in the body of the function and the
arguments from outside the function will be the element type.
* All functions infer from call+construct contexts
* Initial version, doesn't work for primitives yet.
Need to find out why.
* Primitives now work, plus improve error message
null and void don't even parse without parameter names so they are not
tested.
* Codefix: Add names to nameless parameters
* Improve error wording
* Add detail to error message
* string|number inferences are low priority
Also, refactor unifyFromContext to explicitly handle priorities
* string/number/strnum are not mutually exclusive
* Assert that high/low can't apply to same element
* Parameters infer from body usage and call sites
* Function expressions infer from variable decl usages
If the function expression is the initialiser of a variable declaration.
* Update isApplicableFunctionForInference too
* Update baseline
* module.exports aliases have correct flags
They are marked both as (1) alias and (2) assignment declaration. This
fixes alias resolution in cases where multiple module.exports
assignments exist, but differ in whether they are aliases or not:
```js
function f() { }
module.exports = f
module.exports = 23
```
Previously, this construct would fail to resolve the alias `f` because
the `export=` symbol would be marked as Alias | Value but not
Assignment. This change just adds Assignment so that the assignment
declaration alias-following rules apply: you should always follow the
alias, regardless of other flags.
Also, isAliasSymbolDeclaration needed to be tightened up. Previously, I
missed the condition that `module.exports =` aliases required an
EntityNameDeclaration on right-hand-side, just like `export default` and
`export =` aliases.
* Address PR comments
1. Rename test to be more accurate.
2. Always mark module.exports assignments with SymbolFlags.Assignment.
* infer from usage's unification uses multiple passes
Previously, the unification step of infer-from-usage codefix would stop
as soon an answer was found. Now it continues if the result is
*incomplete*, with the idea that later passes may provide a better
inference.
Currently, an *incomplete* inference is
1. The type any.
2. The empty object type `{}` or a union or intersection that contains
`{}`.
In the checker, any takes priority over other types since it basically
shuts down type checking. For type inference, however, any is one of the least
useful inferences.
`{}` is not a good inference for a similar reason; as a parameter
inference, it doesn't tell the caller much about what is expected, and
it doesn't allow the function author to use an object as expected. But
currently it's inferred whenever there's an initialisation with the
value `{}`. With this change, subsequent property assignments to the
same parameter will replace the `{}` with a specific anonymous type. For
example:
```js
function C(config) {
if (config === undefined) config = {};
this.x = config.x;
this.y = config.y;
this.z = config.z;
}
```
* Unify all passes of inference from usage
In the previous commit, I changed inference from usage to continue
inference if a the result was *incomplete*. This commit now runs all 4
inference passes and combines them in a unification step. Currently the
unification step is simple, it:
1. Gathers all inferences in a list.
2. Makes properties of anonymous types optional if there is an empty
object in the inference list.
3. Removes *vacuous* inferences.
4. Combines the type in a union.
An inference is *vacuous* if it:
1. Is any or void, when a non-any, non-void type is also inferred.
2. Is the empty object type, when an object type that is not empty is
also inferred.
3. Is an anonymous type, when a non-nullable, non-any, non-void,
non-anonymous type is also inferred.
I think I might eventually want a concept of priorities, like the
compiler's type parameter inference, but I don't have enough examples to
be sure yet.
Eventually, unification should have an additional step that examines the
whole inference list to see if its contents are collectively
meaningless. A good example is `null | undefined`, which is not useful.
* Remove isNumberOrString
* Unify anonymous types
@andy-ms pointed out that my empty object code was a special case of
merging all anonymous types from an inference and making properties
optional that are not in all the anonymous type. So I did that instead.
* Use getTypeOfSymbolAtLocation instead of Symbol.type!
* Unify parameter call-site inferences too
Because they still have a separate code path, they didn't use the new
unification code.
Also some cleanup from PR comments.
* Add object type unification test
Also remove dead code.
* Only use fallback if no inferences were found
Instead of relying on the unification code to remove the fallback.
Codefix for diagnostic error 2352: "Conversion of type '{0}' to type '{1}' may be
a mistake because neither type sufficiently overlaps with the other. If this was
intentional, convert the expression to 'unknown' first."
Add codefix for both AsExpressions and TypeAssertions
For parameters, the infer-from-usage codefix uses a substantially
different codepath that previously only looked at call site uses. When
this resulted in no inferences, or bad inferences, for a single
parameter, the codefix would just use any. Only if no usages of a
function were found would the codefix use the body-inference
code.
This commit makes parameter inference fall back to body-inference code
for individual parameters when there is no inference or inference to
any.
