On 08/01/15 02:30, Pádraig Brady wrote: > On 08/01/15 00:24, Assaf Gordon wrote: >> Hello, >> >> On 01/07/2015 03:09 PM, Eric Blake wrote: >>> On 01/07/2015 11:49 AM, Markus Elfring wrote: >>>> Do I read the current documentation correctly in >>>> the way that zero-terminated lines are not supported >>>> so far? >>>> http://www.gnu.org/software/coreutils/manual/html_node/split-invocation.html >>> >>> Correct. Patches welcome. In the meantime: >> >> attached is a quick attempt at splitting with custom line-separator '-t=SEP' >> and NUL '-z' . >> >> It's not complete, but could be improved if this is the right direction. > > Yes this is probably worth adding. > Note join(1) had both -t and -z as they were setting separate delimeters. > split(1) would only need one of those. > No other tools allow arbitrary delimiter chars, so I'd > go with just the -z, --zero-terminated option.
Actually thinking a bit more, the other commands using --zero are line oriented, whereas split is not, and so could use other chars in more cases Consider using ^L to mark split points in text for example. Though then you might want to support --suppress-matched like csplit does. Ah I just now see a TODO comment at the top of split.c that suggests -t CHAR or -t REGEX. Now BSD already has -p REGEX so we'd use that syntax for compat, though let's not worry about that now of course. Let's just go with -t CHAR for now. thanks, Pádraig.
