5.10 Speeding Up Pipe Output ¶

This section describes a gawk-specific feature.

Normally, when you send data down a pipeline to a command with print or printf, gawk flushes the output down the pipe. That is, output is not buffered, but written directly. This assures, that pipeline output intermixed with gawk’s output comes out in the expected order:

print "something"                       # goes to standard output
print "something else" | "some-command" # also to standard output
print "more stuff"                      # and this too

There can be a price to pay for this; flushing data down the pipeline uses more CPU time, and in certain environments this can become expensive.

You can tell gawk not to flush buffered data in one of two ways:

• Set PROCINFO["BUFFERPIPE"] to any value. When this is done, gawk will buffer data for all pipelines.
• Set PROCINFO["command", "BUFFERPIPE"] to any value. In this case, only command’s data will be fully buffered.

You must create one or the other of these elements in PROCINFO before the first print or printf to the pipeline. Doing so after output has already been sent is too late.

Be aware that using this feature may change the output behavior of your programs, so exercise caution.