Binary Serach of rows of M-by-N Matrix
Search is based on sum of column values of rows of input M-by-N Matrix
i.e. Rows of input Matrix must already be sorted in ascending order
based on sum of their column values. i.e.
row1sum<row2sum<...rowjsum<...rowmsum
where rowjsum=sum(mat(j,:),2) (i.e sum of column values)
If mat is not sorted use following two statments to sort rows of mat
according to sum of column values
matwithsum=sortrows([sum(mat,2),mat],1);
mat=matwithsum(:,2:end);
INPUT:
mat: Matrix contains values such as
[x1, y1,..., zn;
x2, y2,..., zn;
...
xm, ym,..., zn];
srow: row that is to be searched in mat
OUTPUT:
Using following different value of parameter 'type' following output can be return
1. 'first': returns 'indvec1' i.e index of first occured row in mat that matches to srow.
2. 'all': returns 'indvec1' i.e indices of all rows in mat that matches to srow.
3. 'last': returns 'indvec1' i.e index of last row in mat that matches to srow.
4. 'allcolsum': returns 'indvec1' i.e indices of all k rows of mat such that
sum(mat(j,:),2)=sum(srow,2),
where 1<=j<=k
5. 'colsumandval': return two index vector first is similar to return by "allcolsum" and second is similiar to return by "all' option.
Note that if we are searching e.g. [5 4 2] by option 'first' then there may be many rows whose column sum is equal to given row e.g. [2 4 5],[0 0 11],[4 2 5], [5 4 2] but only
last row i.e. [5,4,2] matches in all columns to given row to be searched.
-------------------------
bsearchmatrows.m (main program)
bsearchmatrowsTest1.m (simple test program)
bsearchmatrowsTest2.m (compare performance with MATLAB builtin metod index=find(ismember(mat,srow, 'rows'));
-------------------
For large data this search is extremely faster than MATLAB built in methods |
