use "relational.sml"; results in errors.

[DavidPratten]

Hi,

Running https://github.com/julianhyde/morel/blob/main/src/test/resources/script/relational.sml using "use".

Does is this .sml file have dependencies?

use "relational.sml";

Appears to have multiple levels of issues.

1. Fatal. Keeping the multi-line comments, the interpreter gives up with:

net.hydromatic.morel.parse.TokenMgrError: Lexical error at line 3, column 27.  Encountered: <EOF> after : ""
        at net.hydromatic.morel.parse.MorelParserImplTokenManager.getNextToken(MorelParserImplTokenManager.java:1434)
        at net.hydromatic.morel.parse.MorelParserImpl.jj_ntk_f(MorelParserImpl.java:2787)
        at net.hydromatic.morel.parse.MorelParserImpl.statement(MorelParserImpl.java:2174)
        at net.hydromatic.morel.parse.MorelParserImpl.statementSemicolon(MorelParserImpl.java:2212)
        at net.hydromatic.morel.Shell$SubShell.extracted(Shell.java:477)
        at net.hydromatic.morel.Shell$SubShell$Use.use(Shell.java:541)
        at net.hydromatic.morel.eval.Session.use(Session.java:61)
        at net.hydromatic.morel.eval.Codes$21.apply(Codes.java:323)
        at net.hydromatic.morel.eval.Codes$ApplyCode.eval(Codes.java:2657)
        at net.hydromatic.morel.compile.Compiler.lambda$compileValDecl$17(Compiler.java:610)
        at net.hydromatic.morel.compile.Compiler$1.lambda$eval$0(Compiler.java:99)
        at net.hydromatic.morel.util.ThreadLocals.let(ThreadLocals.java:37)
        at net.hydromatic.morel.compile.Compiler$1.eval(Compiler.java:95)
        at net.hydromatic.morel.Shell$SubShell.lambda$extracted$0(Shell.java:484)
        at net.hydromatic.morel.eval.Session.withShell(Session.java:50)
        at net.hydromatic.morel.Shell$SubShell.extracted(Shell.java:483)
        at net.hydromatic.morel.Shell.run(Shell.java:264)
        at net.hydromatic.morel.Shell.main(Shell.java:88)

2. Negative Tests? After removing the multi-line comments the interpreter runs to completion but generates multiple instances like this:

Encountered "<EOF>" at line 2, column 18.
Was expecting one of:
    "datatype" ...
    "fun" ...
    "in" ...
    "val" ...

Encountered " "in" "in "" at line 1, column 1.
Was expecting one of:
    "case" ...
    "datatype" ...
    "fn" ...
    "fun" ...
    "if" ...
    "let" ...
    "val" ...
    "from" ...
    <NON_NEGATIVE_INTEGER_LITERAL> ...
    <NEGATIVE_INTEGER_LITERAL> ...
    <REAL_LITERAL> ...
    <SCIENTIFIC_LITERAL> ...
    <QUOTED_STRING> ...
    <CHAR_LITERAL> ...
    "(" ...
    "{" ...
    "[" ...
    "~" ...
    <IDENTIFIER> ...
    <LABEL> ...

[julianhyde]

I'm not at all surprised by this. The problem is that we use the .sml file suffix for two different languages - Morel scripts (such as can be run from the morel shell, and included from the shell via the use function) and test scripts (such as can be run using ScriptTest).

The requirements are different. We want tests to echo their input and output, and we want them to print an error and carry on if a command fails.

Both languages support the use function. But when you use use, they assume that the script is in the same language. Currently, the two languages each have their own implementation of use.

It might be possible to unify the languages by adding some built-in functions to control whether scripts echo their input, whether they carry on in the event of errors. These would have to be Morel functions, not merely directives to the shell. We have a few examples of such functions already: Sys.plan, Interact.use, Sys.set, Sys.show.

Also we would need to unify how we treat block comments (* ... *) and blank lines.

In summary, this is a bug, but it is low priority. Contributions welcome.

[DavidPratten]

OK so there are some negative test cases in relational.sml?

That just leaves comment handling to sort out - right?

[julianhyde]

Sounds about right. We definitely need negative tests (and they can't cause the script to bomb out). I have not analyzed it in detail. As I said, it did not occur to me to run test scripts via use from the shell.