You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For some use cases, it would be beneficial to retain the original input order.
(It's understood that some speed may need to be sacrified. It's also understood that the increased code complexity may be more than you want to maintain!)
Well, yeah ... but it also drastically reduces performance (of course). 😁
But ... combining -t 1 with GNU parallel's --keep-order flag, parallel will do the work of reassembling the outputs in the correct order. So that's a workaround that preserves performance.
This is something I previously looked at, but performance was not what I wanted it to be, so a keep order option was scrapped. I'll put this back on the todo list.
Activity
cyclone-github commentedon Sep 9, 2025
For now, using flag -t 1 (single thread) will maintain the original 1:1 order.
roycewilliams commentedon Sep 9, 2025
Well, yeah ... but it also drastically reduces performance (of course). 😁
But ... combining
-t 1with GNU parallel's--keep-orderflag,parallelwill do the work of reassembling the outputs in the correct order. So that's a workaround that preserves performance.cyclone-github commentedon Sep 9, 2025
This is something I previously looked at, but performance was not what I wanted it to be, so a keep order option was scrapped. I'll put this back on the todo list.
roycewilliams commentedon Sep 9, 2025
That's generous, but maybe the workaround is good enough, and your time would be better spent on something with higher payoff. 😅
cyclone-github commentedon Sep 9, 2025
About to push a PR which adds indexing to keep the output order the same as input. Negligible performance hit.
added feature: "keep-order" from #7
roycewilliams commentedon Sep 11, 2025
Slick! Thanks very much!
cyclone-github commentedon Sep 11, 2025
You're welcome!