TC is too strict in TypedVariableBecomes pattern

[PaulKlint]

The following gives a type error, but is ok and should only lead to a failure to match certain list elements at runtime:

public value main(list[value] args) =  [ <s,r,L> | list[int] L:[*str s, *str r] <- [ [1,2], ["3","4"] ]];

gives

error("Cannot assign pattern of type list[str] to non-inferred variable L of type list[int]",|rascal:///experiments/Compiler/Examples/Tst.rsc|(96,28,<3,51>,<3,79>))

[swatbot]

No, this can never match! List of lists versus sublists. But the error message is confusing. —
Jurgen J. Vinju
CWI SWAT
INRIA Lille
UvA master software engineering
http://jurgen.vinju.org

On Tue, Dec 31, 2013 at 9:39 AM, Paul Klint notifications@github.com
wrote:

The following gives a type error, but is ok and should only lead to a failure to match certain list elements at runtime:

public value main(list[value] args) =  [ <s,r,L> | list[int] L:[*str s, *str r] <- [ [1,2], ["3","4"] ]];

gives

error("Cannot assign pattern of type list[str] to non-inferred variable L of type list[int]",|rascal:///experiments/Compiler/Examples/Tst.rsc|(96,28,<3,51>,<3,79>))

---

Reply to this email directly or view it on GitHub:
#473

[PaulKlint]

well .. it is in the test set :-)

Indeed, this particular example (there are more of these) will never match, but it should not give a type error.

[tvdstorm]

I don't see the problem, is it because it says *str s intead of *list[str] s?

[PaulKlint]

The "problem" (= reason for not matching) is that L is declared with type list[int]

[mahills]

It isn't just a reason for not matching -- you are declaring L as a list[int] and then binding it to a list of strings. That seems like a type problem to me. @jurgenvinju, you are right that this can never match, but that isn't what is generating the error message here.

[PaulKlint]

@mahills you convinced me, this should be a static error; so there is no work involved and we can close this issue.

I am still (too much) inclined to assume that a test case is correct, but on the other hand I have already found dozens of type-incorrect tests.

[swatbot]

An issue is that these literals used in tests have a too specific types for testing purposes. Normally you would never know these specific types and normally you would not have such error messages. I saw this has led too more confusion in other tests as well, so please consider this subtlety, because we are not testing the normal rascal programs like this. —
Jurgen J. Vinju
CWI SWAT
INRIA Lille
UvA master software engineering
http://jurgen.vinju.org

On Tue, Dec 31, 2013 at 4:27 PM, Paul Klint notifications@github.com
wrote:

Closed #473.

Reply to this email directly or view it on GitHub:
#473