On Mon, May 12, 2014 at 12:06 PM, John Cowan <xxxxxx@mercury.ccil.org> wrote:

Alex Shinn scripsit:

> Normalization was in the early issues and dismissed because of lack
> of implementation support and unclear costs in new implementations.
> I think good recommended practice for now is to just normalize both
> inputs and patterns separately.

Okay, I can live with that. But normalizing an SRE is not a matter of
normalizing the strings in the SRE: indeed, that will break it.

This is just recommended practice. If all of your string

literals within the SRE are in NFC, and the input strings

are all in NFC, then you minimize the cases which

failed to match because of a normalization difference.

So at the very least I think a normalize-sre procedure must be provided that
takes an SRE and does the nitty-gritty of selectively expanding charsets
into disjunctions of sequences.

Sure, but as the details and implementation don't exist

yet I'm leaving this for future work.

Alex