Re: Re: unusual type conversion in invoke

[Vladimir Tsichevski <wowa1 at online dot ru>]

Hi Per,

here is more annoying situation:

wowa at src$ cat Test.scm

(invoke(make <java.lang.StringBuffer>) 'append (as <int> 10))

wowa at src$ kawa -C Test.scm
(compiling Test.scm)
Test.scm:3:2: warning - more than one definitely applicable method 
`append' in java.lang.StringBuffer
Test.scm:3:2: warning - candidate: java.lang.StringBuffer 
java.lang.StringBuffer.append(double)
Test.scm:3:2: warning - candidate: java.lang.StringBuffer 
java.lang.StringBuffer.append(float)
Test.scm:3:2: warning - candidate: java.lang.StringBuffer 
java.lang.StringBuffer.append(long)
Test.scm:3:2: warning - candidate: java.lang.StringBuffer 
java.lang.StringBuffer.append(int)
Test.scm:3:2: warning - candidate: java.lang.StringBuffer 
java.lang.StringBuffer.append(char)
Test.scm:3:2: warning - candidate: java.lang.StringBuffer 
java.lang.StringBuffer.append(boolean)
Test.scm:3:2: warning - candidate: java.lang.StringBuffer 
java.lang.StringBuffer.append(java.lang.Object)
Test.scm:3:2: warning - candidate: java.lang.StringBuffer 
java.lang.StringBuffer.append(java.lang.String)
wowa at src$

Kawa refuses to accept a method which signature *exactly* matches the 
type of its argument.
the problem is that this time I do not know how to work this around :-(

Regards,

Vladimir

>Vladimir Tsichevski wrote:
>
>> Hi all,
>>
>> just discovered an unusual time conversion kawa implicitly does when 
looking for
>> appropriate method to invoke.
>>
>> the following line:
>>
>> (invoke 'abcd 'indexOf 10)
>>
>> compiles to:
>>
>> "abcd".indexOf("10");
>>
>> i.e. the number 10 becomes the string "10".
>>
>> a workaround is to explicitly defining the type of constant:
>>
>> (invoke 'abcd 'indexOf (as <int> 10))
>>
>> but I cannot imagine the situation when such implicitly conversion 
might be useful :-(
>
>
>What's happening is that String.indexof(String) is considered
>"definitely applicable" (any object can be converted to a
>string using toString), while String.indexOf(int) applied
>to a gnu.math.IntNum is "conditionally applicable" (since
>the runtime value may be too large.  And the method selection
>code prefers methods that are definitely applicable.
>
>
>One could argue about whether the toString conversion should
>be automatic, but it should certainly be lower priority.
>We might want some types (as least Scheme strings) to be
>automatically converted to Java Strings.
>--
>    --Per Bothner
>per at bothner dot com   http://per.bothner.com/
>