BACKEND auto-selection of function backend per-function

the user can specify a backend, or else this class automatically determines suitable backends for the system.

'sys' is usually the slowest, but most compatible.
'python' or 'dotnet' are among the fastest, but not always available
'java' and 'perl' are usually available and medium speed
'native' uses the latest Matlab syntax available
'legacy' is Matlab syntax for older Matlab versions. Sometimes legacy is faster than native, but native is more robust/general.

classdef Backend < matlab.mixin.SetGet

properties (Constant)
optionalBackends = ["java", "perl", "python", "dotnet"]
namespace = "stdlib"
end

properties
backends string = ["native", "legacy", "sys"]
func
backend
end

methods

function self = Backend(functionName, backendReq)
arguments
  functionName (1,1) string = ""
  backendReq (1,:) string = string.empty
end

if isempty(backendReq) || ~isscalar(backendReq)
  for i = 1:numel(self.optionalBackends)
    f = str2func(sprintf('%s.has_%s', self.namespace, self.optionalBackends(i)));
    if f()
      self.backends = [self.optionalBackends(i), self.backends];
    end
  end
end

if strlength(functionName)
  self.func = self.getFunc(functionName, backendReq);
end

end


function backendAvailable = select(self, functionName, backendReq, firstOnly)
arguments
  self
  functionName (1,1) string
  backendReq (1,:) string = string.empty
  firstOnly (1,1) logical = false
end

backendAvailable = string.empty;

if ~any(strlength(backendReq))
  backendReq = self.backends;
end

for m = backendReq
  switch m
    case 'dotnet'
      if ~any(ismember(self.backends, 'dotnet'))
        continue
      end

      switch functionName
        case {'create_symlink', 'ram_total', 'read_symlink'}
          if stdlib.dotnet_api() < 6, continue, end
        case {'get_owner', 'get_uid', 'is_admin'}
          if isunix(), continue, end
      end
    case 'java'
      if ~any(ismember(self.backends, 'java'))
        continue
      end

      switch functionName
        case {'device', 'hard_link_count' , 'inode', 'is_admin'}
          if ispc(), continue, end
      end
    case 'perl'
      switch functionName
        case {'get_uid'}
          if ispc(), continue, end
      end
    case 'python'
      if ~any(ismember(self.backends, 'python'))
        continue
      end

      switch functionName
        case {'filesystem_type', 'is_removable', 'ram_total', 'ram_free'}
          if ~stdlib.python.has_psutil(); continue, end
        case {'cpu_load', 'get_owner', 'get_process_priority', 'get_uid'}
          if ispc(), continue, end
        case 'is_admin'
          if ispc() || stdlib.matlabOlderThan('R2024a'), continue, end
        case 'is_dev_drive'
          pyv = stdlib.python_version();
          if any(pyv(1:2) < [3, 12]), continue, end
      end
    case 'native'

      switch functionName
        case 'create_symlink'
          % Some Windows R2025a give error 'MATLAB:io:filesystem:symlink:NeedsAdminPerms'
          % 25.1.0.2973910 (R2025a) Update 1 gave this error for example.
          if stdlib.matlabOlderThan('R2024b') || ispc()
            continue
          end
        case {'is_symlink', 'read_symlink'}
          if stdlib.matlabOlderThan('R2024b'), continue, end
        case {'get_permissions', 'set_permissions'}
          if stdlib.matlabOlderThan('R2025a'), continue, end
      end
  end

  if ~isempty(which(sprintf('%s.%s.%s', self.namespace, m, functionName)))
    backendAvailable(end+1) = m; %#ok<AGROW>
    if firstOnly
      return
    end
  end

end

end


function func = getFunc(self, functionName, backendReq)
arguments
  self
  functionName (1,1) string
  backendReq (1,:) string = string.empty
end

if isscalar(backendReq)
  self.backend = backendReq;
else
  self.backend = self.select(functionName, backendReq, true);
end

if isempty(self.backend)
  error('No backend found for %s.%s', self.namespace, functionName);
else
  func = str2func(sprintf('%s.%s.%s', self.namespace, self.backend, functionName));
end

end

end
end