Re^2: Pipes (for interprocess communication)

[penguin_asylum]

Sorry, one more question.
Why is it that "print ''.<$reader>" works, but not just "print <$reader>" ?

[Celada]

.<$reader> forces scalar context while <$reader> uses the context supplied by the print, which is array context. In array context, the <> operator sucks in the whole rest of the file.

Why does this not work? Because it is going to wait indefinately the the end of the file to come. On pipes, the end of file comes when the other end of the pipe is closed. In robin's example, the pipe is never closed.

Mostly for this reason, the usual paradigm is this:

1. Create the pipe
2. fork
3. in the parent, close the reader and use only the writer
4. in the child, close the writer and use only the reader.
or, of course,
1. Create the pipe
2. fork
3. in the parent, close the writer and use only the reader
4. in the child, close the reader and use only the writer.

Trying to use both the reader and the writer end of the same pipe in the same process leads to deadlock. See the documentation of things like IPC::Open2 for how this can occur with a longer chain of pipes too.