Am 09.03.2024 um 20:17 schrieb Ole Tange: > Yep. --null is another way to write --delimiter '\0'. It is different > from --recend '\0'. > > --delimiter is rarely used with --pipe, and --recend is always used > with --pipe/--pipepart. > > Here is an example where -d is used with --pipe: > > $ seq 330000 | parallel --tag --pipe -d , wc {} ::: -l,-w,-c > -l 165668 > -c 101752 > -w 149796 > > In other words: It is not a bug. Instead it is a feature of limited use. > > You have not provided us with the file 'source' so we do not know what > you are trying to do. > > My guess is that you are looking for --recend '\0' > > perl -e 'print map { "$_\0" } 1..1000000' | parallel --recend '\0' --pipe wc > > If that is the case please repay by explaining what should be changed > in the documentation so it would have been obvious to you. > > > /Ole Thank you for initiating further investigations from my side, as i did abandon my project at the time. English not being my first language, i thought, that --null and --recend '\0' and -d '\0' would all mean the same thing, thus my surprise, that --pipe was not handled the way, i did expect.
But your example proved, that --recend yould indeed make parallel do what i want. In my example, the file named "source" was just the output from some "find --print0" command, that i caught in order to play with a pipe for a while. And resuming this play/investigation, i found, that i can solve my initial problem (of distributing work of an enormous amount of files onto the processor cores in parallel) just fine. To meet your request, i have been rereading the man pages attentively. I found them to be clear and concise enough. As i am using the --pipe construct for the first time i was very much interested in SPREADING BLOCKS OF DATA. But my confusion concerning --null, --delimiter and --recend was not based on bad documentation, but on my failure to notice, that recend only concerns --pipe, whereas the other 2 do not apply in that context. Can you construct a case, where both could be useful? I fail to imagine one, and thus thought, they were identical. I still like (and use) parallel quite a bit, especially due to its versatility and ease of use compared to xargs or bash. Good job creating and evolving it! /Ecks