Re: std in NDArrays

[John W. Eaton]

On 27-Aug-2004, David Bateman <address@hidden> wrote:

| As I believe the bitcmp, bitget, bitset issues are completely
| addressed by my last patch (patch.bitfcns20040826-3),

I will look at that one next.

| I attacked the
| problem of "format hex", etc. I also at the same time fixed up the
| column alignment of the integer types..
| 
| Please find the patch and change log attached...

OK, I applied the most recent version of this patch that you sent
after our discussions.

| Since you are
| attacking the fread issues and together with my patch for
| bit{cmp|get|set}, the only issue remaining on your todo list for a
| 2.1.58 release is the mixed type double scalar and integer
| operations. Not sure what you want here though..

For compatibility, we need to make it possible to write things like

  x = uint8 (16);
  y = 2 * x;

and have the result be a uint8 object.  Matlab only does this for
double scalars, presumably because all literal scalar constants in the
language are converted to double values on input, but it is somewhat
inconvenient to be forced to write

  y = uint8 (2) * x;

and also a bit of a problem since mixed-type integer ops are not
supported, so you then have to know the type of the rest of the
expression to be able to multiply by a literal scalar...

I will take a shot at implementing this next.

Thanks,

jwe