Re: When converting tests...

2004-12-20 Thread Patrick R. Michaud
On Mon, Dec 20, 2004 at 05:27:37PM +0200, Markus Laire wrote:
> How should I convert dot, $, ...
> 
> 1# p5: "abc"  =~ /a.c/;  (match)
> 2# p5: "a\nc" =~ /a.c/;  (no match)
> 
> Equivalent code for '.' would now be '\N'. Still there are tests where I 
> could just leave the dot alone (e.g. all tests where there is no \n in 
> target-string.)

These are cases where I think the p5 tests each need to become
multiple tests in the p6rules suite.  The above tests probably should 
go into the p6rules suite as

   "abc"   /a.c/   (match)
   "a\nc"  /a.c/   (match)
   "abc"   /a\Nc/  (match)
   "a\nc"  /a\Nc/  (no match)

so that we get all of the cases implied by #1 and #2 above.

I think this also points to why we might want to just do as much
autoconverting of 're_tests' as we can at the beginning, and then
decide we've done enough and maintain things manually from there.

Pm


When converting tests...

2004-12-20 Thread Markus Laire
How should I convert dot, $, ...
Some examples:
1# p5: "abc"  =~ /a.c/;  (match)
2# p5: "a\nc" =~ /a.c/;  (no match)
Equivalent code for '.' would now be '\N'. Still there are tests where I 
could just leave the dot alone (e.g. all tests where there is no \n in 
target-string.)

In test-1 I could leave dot alone to get a test which tests similar 
concept (dot matching single char).

In test-2 I must change dot to \N to retain the idea of the test.
Still I could just change dot to \N in both tests. But then I wouldn't 
get a test for a dot...

Similar problem exists for $. When there is no //m modifier, and no \n 
in the end, I could just keep $.

When there is \n in the end (but no //m), I could change $ to be \n?$
Of course I could change $ to be \n?$ even when there is no \n in the 
string. But should I?

So should I convert items like dot or $ depending on the string I know 
test will match the rule against (like whether it contains \n or not) - 
or should I convert these items allways in the same way.

(Of course once perl6-rules starts working a lot better than now, we 
anyway need totally new tests to consider all the new possibilities.)

--
Markus Laire