ArmedBear / J

On Mon, 18 Oct 2004 at 21:23:55 +1300, Adam Warner wrote:
> Hi Peter,
>
> (setf (readtable-case *readtable*) :invert)
> '|CL|:eval-when
>
> This should return the symbol EVAL-WHEN in the package COMMON-LISP.
>
> And this shouldn't return the symbol EVAL-WHEN in the package COMMON-LISP:
> '|CL|:EVAL-WHEN
>
> (The COMMON-LISP nickname is escaped so the package name is unaffected by
> the readtable mode. The symbol's characters are unescaped and all of one
> case so they are inverted. Thus eval-when should be interned as
> "EVAL-WHEN" in package COMMON-LISP).

This is now fixed in CVS.

Thanks!

-Peter

Hi Peter Graves,

>> (setf (readtable-case *readtable*) :invert)
>> '|CL|:eval-when
>>
>> This should return the symbol EVAL-WHEN in the package COMMON-LISP.
>>
>> And this shouldn't return the symbol EVAL-WHEN in the package COMMON-LISP:
>> '|CL|:EVAL-WHEN
>>
>> (The COMMON-LISP nickname is escaped so the package name is unaffected by
>> the readtable mode. The symbol's characters are unescaped and all of one
>> case so they are inverted. Thus eval-when should be interned as
>> "EVAL-WHEN" in package COMMON-LISP).
> 
> This is now fixed in CVS.

Fantastic! I've tested the symbol reader extensively and I've managed
to reproduce an additional bug.

(setf (readtable-case *readtable*) :invert)
(make-package :another-package)
(in-package :another-package)
'cl:loop => COMMON-LISP:loop
            [the package name is printed in the wrong case]

I haven't located the fix but this section of Symbol.java appears to be
insufficient (I'd add the two lines with the +'s in the order below
because invert is a more popular readtable mode than downcase):

        // Package prefix is necessary.
        String packageName = pkg.getName();
        if (needsEscape(packageName, readtableCase, thread))
            packageName = multipleEscape(packageName);
+	else if (printCase == Keyword.INVERT)
+	    packageName = invert(packageName);
        else if (printCase == Keyword.DOWNCASE)
            packageName = packageName.toLowerCase();
        StringBuffer sb = new StringBuffer(packageName);
        if (((Package)pkg).findExternalSymbol(n) != null)
            sb.append(':');
        else
            sb.append("::");
        sb.append(s);
        return sb.toString();

Regards,
Adam

On Tue, 30 Nov 2004 at 23:53:44 +1300, Adam Warner wrote:
> Fantastic! I've tested the symbol reader extensively and I've managed
> to reproduce an additional bug.
>
> (setf (readtable-case *readtable*) :invert)
> (make-package :another-package)
> (in-package :another-package)
> 'cl:loop => COMMON-LISP:loop
>             [the package name is printed in the wrong case]

I've just checked in a fix for this.

The ANSI test suite does not appear to catch this failure; ABCL passed
all the tests in print-symbols.lsp even without the latest fix.

> I haven't located the fix but this section of Symbol.java appears to be
> insufficient (I'd add the two lines with the +'s in the order below
> because invert is a more popular readtable mode than downcase):
>
>         // Package prefix is necessary.
>         String packageName = pkg.getName();
>         if (needsEscape(packageName, readtableCase, thread))
>             packageName = multipleEscape(packageName);
> +	else if (printCase == Keyword.INVERT)
> +	    packageName = invert(packageName);
>         else if (printCase == Keyword.DOWNCASE)
>             packageName = packageName.toLowerCase();
>         StringBuffer sb = new StringBuffer(packageName);
>         if (((Package)pkg).findExternalSymbol(n) != null)
>             sb.append(':');
>         else
>             sb.append("::");
>         sb.append(s);
>         return sb.toString();

*PRINT-CASE* is specified to be "one of the symbols :UPCASE, :DOWNCASE,
or :CAPITALIZE", so I don't think we need to handle :INVERT there.

Thanks!

-Peter

Peter Graves wrote:

> The ANSI test suite does not appear to catch this failure; ABCL passed
> all the tests in print-symbols.lsp even without the latest fix.

The reader is one part of the spec that has not yet been tested
(except incidentally to testing other parts.)

	Paul

Hi Peter Graves,

> On Tue, 30 Nov 2004 at 23:53:44 +1300, Adam Warner wrote:
>> Fantastic! I've tested the symbol reader extensively and I've managed
>> to reproduce an additional bug.
>>
>> (setf (readtable-case *readtable*) :invert)
>> (make-package :another-package)
>> (in-package :another-package)
>> 'cl:loop => COMMON-LISP:loop
>>             [the package name is printed in the wrong case]
> 
> I've just checked in a fix for this.

Thanks Peter!

> *PRINT-CASE* is specified to be "one of the symbols :UPCASE, :DOWNCASE,
> or :CAPITALIZE", so I don't think we need to handle :INVERT there.

I didn't realise where the odd :CAPITALIZE option came from! Thank
goodness that read case sensitivity has priority over print case:
"In order to avoid complex interactions between the case setting of the
readtable and *PRINT-CASE*, this proposal specifies a significance for
*PRINT-CASE* only when the case setting is :UPCASE or :DOWNCASE"
[Issue READ-CASE-SENSITIVITY Writeup]

Regards,
Adam