[TRE-general] Repeat histories

[Rici Lake]

On 21-Feb-07, at 3:07 PM, Rici Lake wrote:

 > a bunch of handwaving

I forgot to mention, although I suppose it's obvious:

If we establish that a given sub-expression (R)* is
right-matchable, then we do not need ordinals at all;
we can get the Posix semantics by using the shortest-
non-empty-match rule for the last capture.

That subsumes the tests for constant-length R, amongst
others, because constant-length R are always right-matchable.
It also clearly covers the common case where all the
alternatives of R are distinguishable (eg., have distinct
prefixes) and the case where R is itself S*.