On Mon, Dec 12, 2016 at 9:30 PM, Bradford Larsen <brad.lar...@gmail.com> wrote: > An alternative possibility would be to revert to the pre-3.11 tokenizer on > EBCDIC systems. If I recall, the old tokenizer used a big switch statement > with character literals instead of the 'aiClass' table. I believe this > would avoid the EBCDIC tokenizing troubles, at the expense of lower > performance on those systems, and with the maintenance cost of keeping 2 > code paths around.
Or the build pass could compile a switch-based generator. It would make cross-compiling more complicated. Perhaps easier would be to have a set of unrolled table initializers which used the raw characters as indices, called as part of sqlite3_init(). -scott _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
