On 8 Apr 2007, at 01:11, Fergal Daly wrote:
- it is 1-level deep grouping and so it requires another layer for degrouping. Separating the degrouping from the TAP parser confines it to a single level of grouping. If you're going to do grouping, you should go all the way and make it nestable and let a block be just another kind of test that happens to have subtests and possibly a plan. This requires putting it in the TAP parser.
It is nestable. begin begin ok end begin not ok end end
then if you go and make it nestable: - the white-space indentation is not backwards compatible but if you remove it then the whole things becomes far less human-readable.
I added that to the example only for readability.
- it's difficult to identify a test. You could say "the 2nd test of the the 3rd block of the 5th block" but it would be much nicer to say test 5.3.2 in which case you may as well put that identifier in the stream as the uique identifier of the test ie the test number, this has the added benefit that when I want to find it in the output, I can go straight to it.
What does that leading 5 tell you? How do you relate the (presumably synthetic) numbering to a specific test?
The begin/end syntax allows both begin and end to be followed by descriptive text. Since that text can say things like which file the tests came from it should make it pretty easy to go to a particular test.
-- Andy Armstrong, hexten.net
