On Sun, Jan 28, 2024 at 08:15:33PM +0000, Gavin Smith wrote: > Below is a patch to introduce a new variable to avoid using the > Unicode::Collate module. Turning the module off shortens run times by > about 5% (depending on the size of indices in the document). > > Users can use this if they don't find texi2any fast enough, if they > don't care about having the indices sorted correctly, if they don't > have many non-ASCII characters in index entry text, or just for working > on a manual. > > I propose that USE_UNICODE_COLLATE is on by default, as is currently > the case, to have correct index sorting by default, as the impact is > relatively small.
With collation also possible with XS/C, but with a different result than in perl, I think that there should be a way to use perl unicode collation from C too, in addition to using a unicode collation or not. Should it be a separate customization variable, or should USE_UNICODE_COLLATION be replaced by a variable with a textual value taking more possibilities, for example USE_COLLATION with possible values: unicode unicodeperl basic ? -- Pat
