[parenscript-devel] Bug: Latest PS breaks how quoted symbols were handled before

[Vladimir Sedach]

The two issues with translating CL symbols to JS identifiers are
package prefixes, and name-mangling for special characters. The former
is not a problem with keywords by their nature, and I think we can
forgo the latter since the use of non-alphanumeric characters in
keywords (except for the dash #\-) is extremely low.

Vladimir

On Thu, May 7, 2009 at 11:08 PM, Daniel Gackle <danielgackle at gmail.com> wrote:
> That is a good and unfortunate point you make.
>
> I'm going to wait until the other bugs I posted have been fixed, then will
> upgrade, revisit this issue and try to think of a workaround. I'd rather not
> use strings for the operators, but perhaps keywords would do.
>
> Regards,
> Daniel
>
> On Wed, May 6, 2009 at 3:48 PM, Red Daly <reddaly at gmail.com> wrote:
>>
>> On Wed, May 6, 2009 at 1:34 PM, Daniel Gackle <danielgackle at gmail.com>
>> wrote:
>>>
>>> One thing I forgot to point out: notice that, in the latest PS, the minus
>>> operator '- is actually converted into an empty string in JS, which is
>>> obviously wrong.
>>>
>>>
>>> On Wed, May 6, 2009 at 2:27 PM, Daniel Gackle <danielgackle at gmail.com>
>>> wrote:
>>>>
>>>> The latest PS interprets QUOTE of a symbol to mean, "first convert this
>>>> symbol to a Javascript identifier, then a string". This breaks our code,
>>>> because our formula parser relies on symbols to represent operators. Here's
>>>> a line of code that PS was generating for us nicely before:
>>>>
>>>> (defparameter *op-symbols* '(| | + - * / ^ < > = <> <= >= & % ! { } [ ]
>>>> |(| |)| |:| |,| |"| |'|))
>>>> =>
>>>> var OPSYMBOLS = [' ', '+', '-', '*', '/', '^', '<', '>', '=', '<>',
>>>> '<=', '>=', '&', '%', '!', '{', '}', '[', ']', '(', ')', ':', ',', '"',
>>>> '\''];
>>>>
>>>> In the new PS, it looks like this:
>>>>
>>>> var OPSYMBOLS = [' ', 'plus', '', 'star', 'slash', '^', '<', '>',
>>>> 'equals', '<>', '<equals', '>equals', '&', 'percent', 'bang', '{', '}', '[',
>>>> ']', '(', ')', 'colon', ',', '"', '\''];
>>>>
>>>> I think this is a mistake. The only reason for translating symbols like
>>>> + to "plus" is if you want to use them in a JS identifier. But quoted
>>>> symbols are not meant to turn into JS identifiers, only JS strings.
>>>>
>>>> Looking at the git log, this change was introduced into
>>>> special-forms.lisp by commit dd4442b8973fe8b2c19b44f94f244934aa418ae8. This
>>>> was submitted at the time as a bug fix. From our point of view, this "fix"
>>>> introduces a bug, and quite a serious one, since it means we can't run our
>>>> formula parser (and a few other things) in the browser.
>>
>> The problem was that (slot-value object 'quoted-slot) had inconsistent
>> semantics compared to other uses of quoted symbols.  This was fixed by
>> making a quoted symbols behave as you describe: first convert this symbol to
>> a Javascript identifier, then a string.
>>
>> For an example of where string-translation can go wrong, see this code:
>>
>> (let* ((our-slot 'foo-bar)
>>          (val1 (slot-value object our-slot))
>>          (val2 (slot-value object 'foo-bar))
>>    ...)
>>
>> The fix made this code execute as you would believe it should: so that
>> val1 and val2 are equal.  If we were to translate symbols to their
>> STRING-VALUE, val1 would = object["foo-bar"] and val2 would =
>> object.fooBar.  That is, unless you change the semantics of slot-value.
>>
>> I am not sure what the context of your *op-symbols* code is, but it seems
>> like you should be able to avoid this problem by using strings.  I don't
>> know what the alternative is for the foo-bar example above, if we adopt a
>> symbol translator like you suggest.
>>
>> Best,
>> Red
>>
>>
>>
>>
>>>
>>> _______________________________________________
>>> parenscript-devel mailing list
>>> parenscript-devel at common-lisp.net
>>> http://common-lisp.net/cgi-bin/mailman/listinfo/parenscript-devel
>>>
>>
>>
>> _______________________________________________
>> parenscript-devel mailing list
>> parenscript-devel at common-lisp.net
>> http://common-lisp.net/cgi-bin/mailman/listinfo/parenscript-devel
>>
>
>
> _______________________________________________
> parenscript-devel mailing list
> parenscript-devel at common-lisp.net
> http://common-lisp.net/cgi-bin/mailman/listinfo/parenscript-devel
>
>