Prof.
in the book "Numerical methods in structural mechanics" p80.
     | dh1/dx    0                    0              dh2/dx     0                    0                |
B= | 0           dh3/dx          dh4/dx         0           dh5/dx           dh6/dx         |   
     | 0            h3+dh7/dx   h4+dh8/dx  0            h5+dh9/dx   h6+dh10/dx  |

but in the OOFEM code LIBeam2d :: computeBmatrixAt()
l    = this->giveLength();
    ksi  = aGaussPoint->giveCoordinate(1);
    n1x   = -1.0 / l;
    n4x   =  1.0 / l;
    n3xx  = -1.0 / l;
    n6xx  =  1.0 / l;
    n2xxx = -1.0 / l;
    n3xxx =  0.5 * ( 1 - ksi );
    n5xxx =  1.0 / l;
    n6xxx =  0.5 * ( 1. + ksi );

    // answer = new FloatMatrix(3,6) ;
    answer.resize(3, 6);
    answer.zero();

    answer.at(1, 1) =  n1x;
    answer.at(1, 4) =  n4x;
    answer.at(2, 3) =  n3xx;
    answer.at(2, 6) =  n6xx;
    answer.at(3, 2) =  n2xxx;
    answer.at(3, 3) =  n3xxx;
    answer.at(3, 5) =  n5xxx;
    answer.at(3, 6) =  n6xxx;

some question:
(1) why the answer.at(2,2) and answer.at(2,5) =0 ?
(2) how to derived the following resule according table 2.1:
    n3xx  = -1.0 / l;
    n6xx  =  1.0 / l;
    n2xxx = -1.0 / l;
    n3xxx =  0.5 * ( 1 - ksi );
    n5xxx =  1.0 / l;
    n6xxx =  0.5 * ( 1. + ksi );

thanks