Hi,
Looking for occurrences of a vector in an array is a general task that
can be useful in many situations.
vectorfind() currently does it, but with some cumbersome limitations.
A work was started in last February to extend vectorfind() capabilities.
It was announced @
http://mailinglists.scilab.org/Scilab-users-Search-for-a-subvector-in-a-vector-tp4035450p4035452.html
An extended version is now available for tests and comments, at the
unchanged URL:
https://codereview.scilab.org/#/c/19055/
A copy of its documentation page in PDF is attached.
This is a Scilab Enhancement Proposal (SEP).
Its history section summarizes new proposed features.
They are reminded and illustrated herebelow.
All comments are welcome.
Best regards
Samuel
=======================
*New features proposed*:
* |vectorfind(H,[])| returns an error.
=> Returning |[]| is proposed instead.
* When the needle is longer than the haystack side size, an error is
yielded.
=> Returning |[]| is proposed instead
* vectorfind() can't presently be used with a needle shorter than the
haystack size.
This was the limitation pointed by Christophe in February.
=> Supporting any *short needle* is proposed.
example:
m = [
1 0 0 1
1 0 0 1
0 0 1 0 ];
vectorfind(m, [0 0], "r") // => [3 4 5] instead of an error
vectorfind(m, [1 1], "c") // => [1 10] instead of an error
* When the haystack array has some *%nan* values, these ones are
presently never matching.
=> Considering %nan values as regular ones is proposed.
Example:
m = [
1 %nan 0 1
1 0 0 1
0 0 1 %nan ];
vectorfind(m, [1 %nan], "r") // => [1 9] instead of []
* It is presently not possible to use *wildcards* in the vector to
search. For instance, if in the previous matrix m we search all
columns starting with 0 and ending with 1, there is no way to
specify a [0 * * 1]-like vector as needle.
o A new *joker* option is proposed in input. It declares a value
that can be used in the vector and has the status of a wildcard:
It is matched by any value of the haystack array.
o In addition, a new *matching* output argument is proposed. It
allows to retrieve actual matching ranges. Example with the m
matrix hereabove:
--> [ind, matching] = vectorfind(m, [1 *-1*0], "r", *-1*)
ind =
1. 2.
matching =
1. Nan 0.
1. 0. 0.
* It is presently not possible to search the vector inside an
*hypermatrix*. Yet, N-Dimensional data are not rare -- like any RGB
image --, and hypermatrices are now natively encoded in Scilab 6.
=> Supporting hypermatrices as haystack is proposed. The attached
page shows 2 examples, like selecting all pixels of a RGB image (3D
array) having some given [r g b] values.
Another short example:
--> m = grand(2,4,3,"uin",0,2)
m =
(:,:,1)
2. 1. 2. 1.
1. 1. 0. 0.
(:,:,2)
0. 1. 0. 2.
2. 0. 1. 0.
(:,:,3)
0. 1. 2. 2.
0. 0. 1. 2.
--> vectorfind(m, [2 0], "c")
ans =
3. 8.
--> [ind, matching] = vectorfind(m, [0 -1 -1 2], "r", -1)
matching =
0. 1. 0. 2.
0. 1. 2. 2.
0. 0. 1. 2.
ind =
3. 5. 6.
* Extending vectorfind() to hypermatrices requires to allow probing
extra directions beyond rows and columns. Hence, specifying the
*probing direction as an integer 1 , 2, 3*,.. is proposed. However,
as indicated in the page (Description section), the equivalences 1
<=> "r" and 2 <=> "c" are questionable. The present proposal is
implemented with them.
Example with the m hypermatrix hereabove:
--> vectorfind(m, [2 0], 3)
ans =
1. 5. 10.
* Interpreting and using "matrix-like" default indices returned by
vectorfind() when the haystack is an hypermatrix may be hard and not
handy. A new option *|indType|* is proposed to allow returning
explicit or truly linear indices. It is proposed as well when
searching in a simple matrix.
Back to the last example hereabove, we will get, instead of [1 5 10]:
--> vectorfind(m, [2 0], 3,,"headIJK")
ans =
1. 1. 1.
1. 3. 1.
2. 1. 2.
_______________________________________________
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users
