No, no... don't try to reimplement i. . That's a lifetime of work. Or
more.
I will return the limit on the size of an array back to what it was,
2^48. That will be enough for all of us!
The flaw in my thinking was that I forgot that i. on characters and
2-byte characters doesn't use the exotic forms of i. - they just use
tricky lightweight ones (this is all Roger's stuff).
Thus, very large character vectors should be fine, and we will support
them as we did before.
Henry Rich
On 12/17/2019 11:53 AM, Raul Miller wrote:
If the /: and I. family do not have this limitation, it seems like it
would be straightforward to include a fallback for large arrays. As
in: literally: check the length of the array and if it's too large
evaluate a J explicit definition which implements indexof.
But we'd need a version without the limit error (and some patience) to
test the code. (So... j807 for that??)
Thanks,
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
