A Pipe has two ends. Using APL in a multiprocess/multiprocessor environment. A proposal for a flexible but easy to use syntax.

A proposal for a flexible but easy to use syntax.

Many mechanisms have been proposed for accessing data, and running processes external to APL. For some of these, the syntax is bizarre, for others a detailed knowledge of some other language (sometimes even assembler) is necessary to do the simplest thing. Some, such as shared variables, offer a highly efficient method of passing data between two processes in the same processor, without having to copy the data. A number of methods either allow the passing of data and/or commands to an external shell (such as the C-shell under Unix), or allow the results of commands to the shell to be imported into APL. These in effect connect APL to one end (but not both ends) of a pipe of shell commands. I offer for general consideration a device that allows data to be piped out of APL, through (a series of) shell commands, and back into APL. It works by connecting both ends of a pipeline of shell commands to an APL statement. The output of the first part of the APL expression fills the pipe connected to the standard input of the first command in the pipeline of shell commands; the next part of the APL expression receives the output of the pipe connected to the standard output of the the last command in the pipeline. The commands may be executed in the same processor, or remotely, with no change in syntax.