asked on

putting a buffer into a pipeline

I have two apps, writer and reader. Writer writes to stdout and reader reads from stdin. Currently, they are pipped together `writer | reader`.

Reader batch reads about 10k lines at a time then processes. Writer needs to use this processing time to build up the next batch of 10k lines. The writer doesn't have a buffer in it, and I can't reasonably change this. I'd like to changes this to `writer | buffer | reader` such that the buffer can collect the output of the writer and make it available for the reader when it wants it.

Is there something pre-fab in linux that can be this buffer? Or do I need to write something?

ASKER CERTIFIED SOLUTION

medvedd

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

Brian Utterback

Since you said that there were 10K lines, be careful when you choose your buffer sizes. The buffer sizes are specified in bytes/blocks etc., not lines. Also since the dd command is not threaded, if you make the buffer too small it will block writing to the pipe and the writer will back up and probably block writing to its pipe. If you make the buffer too large then the data may not be available at all when the reader asks for it.

Each pipe has its own buffer, so you can leverage that get the effect you want.

Do this:

writer | cat | cat | ... | cat | reader

Its ugly, but simple. Just add more "cat" stages until you get the effect you want.