User Tools

Site Tools


features:pipes

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
features:pipes [2022/09/13 06:28] โ€“ old revision restored (2015/12/29 19:42) 135.181.78.182features:pipes [2022/09/16 18:47] (current) โ€“ old revision restored (2022/09/10 20:43) 123.219.209.121
Line 1: Line 1:
 ====== Pipes ====== ====== Pipes ======
 +
 +**Wikipedia: ** [[https://en.wikipedia.org/wiki/Pipeline_(Unix)|Pipeline (Unix)]]
 +
 +The idea of connecting an output from one process, via an in-kernel memory buffer, to the input of another process is entirely original to Unix.
 +
 +Doug McIlroy, in 1964, had proposed that "We should have some ways of connecting programs like garden hose โ€“ screw in another segment when it becomes necessary to massage data in another way. This is the way of IO also." (Dennis Ritchie, [[https://www.bell-labs.com/usr/dmr/www/mdmpipe.html|Advice from Doug McIlroy]])
 +
 +===== The Evolution of Pipes =====
 +
 +(taken from Dennis Ritchie, [[https://www.bell-labs.com/usr/dmr/www/hist.html|The Evolution of the Unix Time-sharing System]])
 +
 +One of the most widely admired contributions of Unix to the culture of operating systems and command languages is the //pipe//, as used in a pipeline of commands. Of course, the fundamental idea was by no means new; the pipeline is merely a specific form of coroutine. Even the implementation was not unprecedented, although we didn't know it at the time; the `communication files' of the Dartmouth Time-Sharing System [10] did very nearly what Unix pipes do, though they seem not to have been exploited so fully.
 +
 +Pipes appeared in Unix in 1972, well after the PDP-11 version of the system was in operation, at the suggestion (or perhaps insistence) of M. D. McIlroy, a long-time advocate of the non-hierarchical control flow that characterizes coroutines. Some years before pipes were implemented, he suggested that commands should be thought of as binary operators, whose left and right operand specified the input and output files. Thus a `copy' utility would be commanded by
 +
 +<code>
 +inputfile copy outputfile
 +</code>
 +
 +To make a pipeline, command operators could be stacked up. Thus, to sort //input,// paginate it neatly, and print the result off-line, one would write
 +<code>
 +input sort paginate offprint
 +</code>
 +
 +In today's system, this would correspond to
 +
 +<code>
 +sort input | pr | opr
 +</code>
 +
 +The idea, explained one afternoon on a blackboard, intrigued us but failed to ignite any immediate action. There were several objections to the idea as put: the infix notation seemed too radical (we were too accustomed to typing `cp x y' to copy //x// to //y//); and we were unable to see how to distinguish command parameters from the input or output files. Also, the one-input one-output model of command execution seemed too confining. What a failure of imagination!
 +
 +Some time later, thanks to McIlroy's persistence, pipes were finally installed in the operating system (a relatively simple job), and a new notation was introduced. It used the same characters as for I/O redirection. For example, the pipeline above might have been written
 +
 +<code>
 +sort input>pr>opr>
 +</code>
 +
 +The idea is that following a `>' may be either a file, to specify redirection of output to that file, or a command into which the output of the preceding command is directed as input. The trailing `>' was needed in the example to specify that the (nonexistent) output of //opr// should be directed to the console; otherwise the command //opr// would not have been executed at all; instead a file //opr// would have been created.
 +
 +The new facility was enthusiastically received, and the term `filter' was soon coined. Many commands were changed to make them usable in pipelines. For example, no one had imagined that anyone would want the //sort// or //pr// utility to sort or print its standard input if given no explicit arguments.
 +
 +Soon some problems with the notation became evident. Most annoying was a silly lexical problem: the string after `>' was delimited by blanks, so, to give a parameter to //pr// in the example, one had to quote:
 +<code>
 +sort input>"pr -2">opr>
 +</code>
 +
 +Second, in attempt to give generality, the pipe notation accepted `<' as an input redirection in a way corresponding to `>'; this meant that the notation was not unique. One could also write, for example,
 +
 +<code>
 +opr <pr<"sort input"<
 +</code>
 +
 +or even
 +
 +<code>
 +pr <"sort input"<>opr>
 +</code>
 +
 +The pipe notation using `<' and `>' survived only a couple of months; it was replaced by the present one that uses a unique operator to separate components of a pipeline. Although the old notation had a certain charm and inner consistency, the new one is certainly superior. Of course, it too has limitations. It is unabashedly linear, though there are situations in which multiple redirected inputs and outputs are called for. For example, what is the best way to compare the outputs of two programs? What is the appropriate notation for invoking a program with two parallel output streams?
 +
 +I mentioned above in the section on IO redirection that Multics provided a mechanism by which IO streams could be directed through processing modules on the way to (or from) the device or file serving as source or sink. Thus it might seem that stream-splicing in Multics was the direct precursor of Unix pipes, as Multics IO redirection certainly was for its Unix version. In fact I do not think this is true, or is true only in a weak sense. Not only were coroutines well-known already, but their embodiment as Multics spliceable IO modules required that the modules be specially coded in such a way that they could be used for no other purpose. The genius of the Unix pipeline is precisely that it is constructed from the very same commands used constantly in simplex fashion. The mental leap needed to see this possibility and to invent the notation is large indeed.
 +
 +===== Dating the Creation of Pipes =====
 +
 +According to the [[:anecdotes:pipes|anecdotes]], Ken Thompson added pipes to Unix in one day.
 +
 +The Second Edition of Unix, dated June 1972, didn't have pipes. By January 15, 1973, Unix did have pipes: Doug McIlroy put out the notice for a [[http://www.tuhs.org/Archive/Documentation/Papers/Unix_Users_Talk_Notes_Jan73.pdf|talk which described the state of UNIX]] at that time. Page 4 describes SYS PIPE and its implementation. This is as narrow as we can get on the date when Ken added pipes to the system.
  
features/pipes.txt ยท Last modified: 2022/09/16 18:47 by 123.219.209.121