Impressive! Thanks for sharing this.

By using string_view and reading the whole file into memory in one go (as assumed it will fit), I’ve managed to get the timing from the winning entry from 16993 to 3061 – which is 5 times faster! (using the same 395Mb file). See http://coliru.stacked-crooked.com/a/83471b5b25357b40

The code is not that C++17 oriented, but I was looking at timing more than anything. Reading a line into a string and then splitting that string into other strings is sure time expensive! It doesn’t matter for small files – but for large files of hundred of thousands of lines, it sure does.

Very good point. I missed that one. Code review saves the day!

I notice that the csv class allows an arbitrary set of delimiters but the output stream always uses a comma. It seems like you would want to preserve the original delimiters.

Man. Disqus can use some improvements to their UI/UX 🙂 I mean, it should not be that hard to show comments moderation to the poster (with indication that it is pending).

Thanks, I’ll clean up the duplication a bit then. [EDIT no I won’t, Disqus doesn’t allow me to…]

Hi, the comments module on Disqus hides some comments until approbation. I’ve approved your comments so they should show now. I’m sorry for the inconvenience.

https://uploads.disquscdn.com/images/1b2333c76bb7b4885fde66c3de3d2f7cb173c477eb94dc85a1e20ec554bfd217.png