[Scilab-users] question on makecell

[Samuel Gougeon]

Le 05/12/2019 à 14:25, Federico Miyara a écrit :
>
> Dear all,
>
> When using makecell such as in
>
> --> u = makecell([2 2], cos, 'hello', 1+%s^2, [1 2; 3 4])
>  u  =
>
>   [  1 fptr      ]  [1x1 string  ]
>   [1x1 polynomial]  [2x2 constant]
>
> we get a cell array where the unidimensional ordering of the cells 
> goes first along columns and then ows. However, when extracting 
> components unidimensionally, it behaves the other way around:
>
> --> u(1:4)
>  ans  =
>
>   [  1 fptr      ]
>   [1x1 polynomial]
>   [1x1 string    ]
>   [2x2 constant  ]
>
> The same happens wen applying matrix:
>
> --> matrix(u,4,1)
>  ans  =
>
>   [  1 fptr      ]
>   [1x1 polynomial]
>   [1x1 string    ]
>   [2x2 constant  ]
>
> --> matrix(u,1,4)
>  ans  =
>
>   [1 fptr]  [1x1 polynomial]  [1x1 string]  [2x2 constant]
>
> Isn't this inconsistent?
> Shouldn't makecell create the cell array going along rows and then 
> columns?


You are right, Federico.
We could have expected to list cells elements in linearized indices 
order, as for all other native Scilab arrays.

This irregularity/exception is completely useless and misleading:

--> c = { "abcd" %T ; %pi %z}
  c  =
   [1x1 string  ]  [1x1 boolean   ]
   [1x1 constant]  [1x1 polynomial]

--> makecell([2 2], c{:}(:))
  ans  =
   [1x1 string ]  [1x1 constant  ]
   [1x1 boolean]  [1x1 polynomial]

The fact that it is documented does not attenuate it.

Please do not hesitate to report a wish about removing this exception.
It would not be complicated to remove it, and tracking and updating it 
in existing codes would be easy, through the /makecell/ keyword.

Regards
Samuel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/users/attachments/20191205/89ba66cd/attachment.htm>