Hoenders and Slump[7]
          describe a method for determining number and multiplicities of zeros of a function
          based on a numerical quadrature technique, but as they show in their Tables 4-5, this
          technique is unstable.  It seems unlikely to us that any numerical technique
          could decisively solve this problem.  Symbolic methods can come to our aid in
          some instances.  First, if
           
          is a rational function over a suitable computable extension
           
          of
           , then square-free factorization is applicable.  Second, if
           
          where
           
          is a purely transcendental extension of a computable extension
           
          of
           ,
           
          is a multiple zero, and
           
          where
           
          is a rational function, then
         
        
            
         
        
            
         
        
          are rational functions in
           
          and
           
          identical to zero.  Since
           
          is a common root of both
           
          and
           , the resultant
         
        
            
         
        
          must be zero.  Hence
           
          is a root of
           
          and is algebraic over
           .  The equation
           
          can be solved symbolically for
           
          and then this solution substituted into
           
          and
           
          to ascertain if
           
          proving
           
          is a multiple zero.  Third, if
           
          can be decomposed as a composition
           ,
           
          is a root of
           
          with multiplicity
           ,
           
          is a root of
           
          with multiplicity
           , then
           
          is a root of
           
          with multiplicity
           . 
         
        A piecewise holomorphic function defined by 
        
            
         
        
          consisting of finitely many holomorphic
           
          may be treated as
           
          separate inputs to our algorithm. 
         
        
          An expression
           
          containing non-holomorphic subexpressions can sometimes be rewritten to become holomorphic
          or piecewise holomorphic.  As an example, suppose
           
          where
           
          and
           
          are holomorphic.  Then solve
           
          and replace
           
          by better expressions
           . 
         
        
          Our algorithm is restricted to a finite interval
           .  In some cases, such as
           , this is necessary, for otherwise there would be an infinite number of solutions. 
          In other cases, such as
           , there are only finitely many solutions even on
           .  The strategy proposed in this case is to use some asymptotic analysis to find
           
          such that
           
          is non-zero on
           
          and
           
          leaving only
           
          to contend with.
          |