I think we need to restrict the range of valid unicode identifiers somehow.
For example, whitespace probably should not be in variable names: https://www.cs.tut.fi/~jkorpela/chars/spaces.html Or, for example, right-to-left marks should not be in variable names: https://en.wikipedia.org/wiki/Right-to-left_mark (Also, we should be aware that we currently do not support all of unicode - we only [roughly] support the ucs2 characters.) Thanks, -- Raul On Sun, Jun 12, 2016 at 4:30 PM, Marshall Lochbaum <mwlochb...@gmail.com> wrote: > Unbox has code to allow unicode identifiers in J, with the following > rules: > > - All code must be UTF-8. Invalid UTF-8 causes a spelling error. > - Any non-ASCII character is treated as alphabetic. Identifiers can use > these characters freely. > > This is completely backwards-compatible with existing J, and allows us > to use things like greek characters and code in other languages: > > π > |value error: π > π =: 1p1 > π > 3.14159 > π_1 > |value error: π_1 > > What do people think about this? Should it be added to jsource? Should > the rules be changed for some characters? > > Marshall > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
