Reconstruct. This could only be utilised if you have a single dtype for all columns. Or else, you are going to end up with dtype item for all columns and converting them again involves additional dictionary do the job. But there are other solutions to say the same point; for https://garrettxqdmx.blog4youth.com/37151600/the-smart-trick-of-stack-ball-speed-mode-that-nobody-is-discussing